{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.7730962504831852,
  "eval_steps": 500,
  "global_step": 16000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 110.1875,
      "epoch": 4.831851565519907e-05,
      "grad_norm": 15.142305917979977,
      "learning_rate": 5e-07,
      "loss": -0.0,
      "reward": 1.25,
      "reward_std": 0.4387562870979309,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 1
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 114.75,
      "epoch": 9.663703131039814e-05,
      "grad_norm": 14.809241765918944,
      "learning_rate": 4.999758407421724e-07,
      "loss": -0.0,
      "reward": 1.375,
      "reward_std": 0.6346027851104736,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 124.21875,
      "epoch": 0.00014495554696559721,
      "grad_norm": 13.379259552927518,
      "learning_rate": 4.999516814843448e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.4457654356956482,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 0.96875,
      "step": 3
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 111.75,
      "epoch": 0.00019327406262079628,
      "grad_norm": 601.7144861871252,
      "learning_rate": 4.999275222265172e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.49760180711746216,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 4
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 108.53125,
      "epoch": 0.00024159257827599537,
      "grad_norm": 21.376868981952253,
      "learning_rate": 4.999033629686895e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.5625313520431519,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 0.96875,
      "step": 5
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 88.5625,
      "epoch": 0.00028991109393119443,
      "grad_norm": 26.78357902234386,
      "learning_rate": 4.998792037108619e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.5770788192749023,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 0.96875,
      "step": 6
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 103.21875,
      "epoch": 0.0003382296095863935,
      "grad_norm": 21.93142809072171,
      "learning_rate": 4.998550444530344e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.6671271324157715,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 106.28125,
      "epoch": 0.00038654812524159255,
      "grad_norm": 17.084756671844687,
      "learning_rate": 4.998308851952068e-07,
      "loss": -0.0,
      "reward": 1.671875,
      "reward_std": 0.36999404430389404,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 124.5,
      "epoch": 0.00043486664089679164,
      "grad_norm": 9.136773149857643,
      "learning_rate": 4.998067259373792e-07,
      "loss": -0.0,
      "reward": 1.40625,
      "reward_std": 0.4124833345413208,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 0.9375,
      "step": 9
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 118.40625,
      "epoch": 0.00048318515655199073,
      "grad_norm": 16.978362922926575,
      "learning_rate": 4.997825666795516e-07,
      "loss": -0.0,
      "reward": 1.171875,
      "reward_std": 0.3039700984954834,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 0.875,
      "step": 10
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 119.59375,
      "epoch": 0.0005315036722071898,
      "grad_norm": 13.325574803978368,
      "learning_rate": 4.99758407421724e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.5848016142845154,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 0.9375,
      "step": 11
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 141.40625,
      "epoch": 0.0005798221878623889,
      "grad_norm": 8.431823329804283,
      "learning_rate": 4.997342481638963e-07,
      "loss": 0.0,
      "reward": 1.078125,
      "reward_std": 0.5179688930511475,
      "rewards/accuracy_reward": 0.125,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 0.90625,
      "step": 12
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 131.96875,
      "epoch": 0.000628140703517588,
      "grad_norm": 15.139865179316045,
      "learning_rate": 4.997100889060688e-07,
      "loss": 0.0,
      "reward": 1.5,
      "reward_std": 0.7260431051254272,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 0.875,
      "step": 13
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 121.34375,
      "epoch": 0.000676459219172787,
      "grad_norm": 31.315129005665327,
      "learning_rate": 4.996859296482411e-07,
      "loss": -0.0,
      "reward": 1.671875,
      "reward_std": 0.7513567209243774,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 0.96875,
      "step": 14
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 113.53125,
      "epoch": 0.0007247777348279861,
      "grad_norm": 16.955756035560174,
      "learning_rate": 4.996617703904136e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.5013028383255005,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 113.84375,
      "epoch": 0.0007730962504831851,
      "grad_norm": 31.2749828529687,
      "learning_rate": 4.99637611132586e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.5164258480072021,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 16
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 120.71875,
      "epoch": 0.0008214147661383842,
      "grad_norm": 15.085501072157163,
      "learning_rate": 4.996134518747584e-07,
      "loss": -0.0,
      "reward": 1.59375,
      "reward_std": 0.6612019538879395,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 0.90625,
      "step": 17
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 107.9375,
      "epoch": 0.0008697332817935833,
      "grad_norm": 45.92182770891859,
      "learning_rate": 4.995892926169308e-07,
      "loss": -0.0,
      "reward": 1.59375,
      "reward_std": 0.6018744707107544,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 18
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 104.0,
      "epoch": 0.0009180517974487824,
      "grad_norm": 15.040346630138659,
      "learning_rate": 4.995651333591032e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.589191198348999,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 0.90625,
      "step": 19
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 118.125,
      "epoch": 0.0009663703131039815,
      "grad_norm": 10.62594232723054,
      "learning_rate": 4.995409741012756e-07,
      "loss": 0.0,
      "reward": 1.109375,
      "reward_std": 0.38258251547813416,
      "rewards/accuracy_reward": 0.125,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 0.9375,
      "step": 20
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 135.28125,
      "epoch": 0.0010146888287591806,
      "grad_norm": 12.744415092020956,
      "learning_rate": 4.99516814843448e-07,
      "loss": -0.0,
      "reward": 1.25,
      "reward_std": 0.6149259805679321,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 0.96875,
      "step": 21
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 108.65625,
      "epoch": 0.0010630073444143795,
      "grad_norm": 25.881178503231165,
      "learning_rate": 4.994926555856203e-07,
      "loss": -0.0,
      "reward": 1.734375,
      "reward_std": 0.6611828804016113,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 22
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 119.75,
      "epoch": 0.0011113258600695787,
      "grad_norm": 19.971276670866047,
      "learning_rate": 4.994684963277928e-07,
      "loss": -0.0,
      "reward": 1.4375,
      "reward_std": 0.7392380833625793,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 0.9375,
      "step": 23
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 124.28125,
      "epoch": 0.0011596443757247777,
      "grad_norm": 16.117275734448256,
      "learning_rate": 4.994443370699652e-07,
      "loss": -0.0,
      "reward": 1.453125,
      "reward_std": 0.5126951932907104,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 24
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 120.71875,
      "epoch": 0.0012079628913799767,
      "grad_norm": 24.267471456858143,
      "learning_rate": 4.994201778121376e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.5850760340690613,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 25
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 132.21875,
      "epoch": 0.001256281407035176,
      "grad_norm": 26.129810617385587,
      "learning_rate": 4.9939601855431e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.6284842491149902,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 0.96875,
      "step": 26
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 131.625,
      "epoch": 0.0013045999226903749,
      "grad_norm": 12.375416378730701,
      "learning_rate": 4.993718592964824e-07,
      "loss": -0.0,
      "reward": 1.4375,
      "reward_std": 0.4619767367839813,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 0.9375,
      "step": 27
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 132.40625,
      "epoch": 0.001352918438345574,
      "grad_norm": 15.883091067120562,
      "learning_rate": 4.993477000386548e-07,
      "loss": 0.0,
      "reward": 1.453125,
      "reward_std": 0.5068542957305908,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 0.96875,
      "step": 28
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 126.28125,
      "epoch": 0.001401236954000773,
      "grad_norm": 24.65376614518306,
      "learning_rate": 4.993235407808271e-07,
      "loss": -0.0,
      "reward": 1.359375,
      "reward_std": 0.6136289834976196,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 29
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 120.59375,
      "epoch": 0.0014495554696559723,
      "grad_norm": 19.198227257409375,
      "learning_rate": 4.992993815229997e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.7768711447715759,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 0.96875,
      "step": 30
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 124.3125,
      "epoch": 0.0014978739853111712,
      "grad_norm": 23.564346204174132,
      "learning_rate": 4.99275222265172e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.5434433221817017,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 0.96875,
      "step": 31
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 109.125,
      "epoch": 0.0015461925009663702,
      "grad_norm": 18.965712049950234,
      "learning_rate": 4.992510630073444e-07,
      "loss": -0.0,
      "reward": 1.59375,
      "reward_std": 0.39408785104751587,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 32
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 107.875,
      "epoch": 0.0015945110166215694,
      "grad_norm": 13.109149671059429,
      "learning_rate": 4.992269037495168e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.4944065809249878,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 33
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 132.0625,
      "epoch": 0.0016428295322767684,
      "grad_norm": 55.92671260394731,
      "learning_rate": 4.992027444916892e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.2783603370189667,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 34
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 115.28125,
      "epoch": 0.0016911480479319676,
      "grad_norm": 23.34111317826194,
      "learning_rate": 4.991785852338616e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.46113359928131104,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 35
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 119.65625,
      "epoch": 0.0017394665635871666,
      "grad_norm": 15.449278065323663,
      "learning_rate": 4.991544259760339e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.5970016121864319,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 36
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 109.09375,
      "epoch": 0.0017877850792423658,
      "grad_norm": 103.1214159899848,
      "learning_rate": 4.991302667182064e-07,
      "loss": 0.0,
      "reward": 1.484375,
      "reward_std": 0.41897958517074585,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 0.9375,
      "step": 37
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 135.09375,
      "epoch": 0.0018361035948975648,
      "grad_norm": 28.72953308705984,
      "learning_rate": 4.991061074603788e-07,
      "loss": 0.0,
      "reward": 1.546875,
      "reward_std": 0.7149031162261963,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 38
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 117.40625,
      "epoch": 0.0018844221105527637,
      "grad_norm": 37.42503498336226,
      "learning_rate": 4.990819482025512e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.6485819816589355,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 0.96875,
      "step": 39
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 107.25,
      "epoch": 0.001932740626207963,
      "grad_norm": 25.43316736939667,
      "learning_rate": 4.990577889447236e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.5434433221817017,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 40
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 117.28125,
      "epoch": 0.001981059141863162,
      "grad_norm": 13.062697842731525,
      "learning_rate": 4.99033629686896e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.5167425274848938,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 0.96875,
      "step": 41
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 134.6875,
      "epoch": 0.002029377657518361,
      "grad_norm": 12.81500137797492,
      "learning_rate": 4.990094704290684e-07,
      "loss": -0.0,
      "reward": 1.34375,
      "reward_std": 0.4670553505420685,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 42
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 103.15625,
      "epoch": 0.00207769617317356,
      "grad_norm": 18.12674509599202,
      "learning_rate": 4.989853111712408e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.6580699682235718,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 43
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 147.25,
      "epoch": 0.002126014688828759,
      "grad_norm": 11.465857654460452,
      "learning_rate": 4.989611519134132e-07,
      "loss": -0.0,
      "reward": 1.234375,
      "reward_std": 0.4368523061275482,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 44
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 132.71875,
      "epoch": 0.002174333204483958,
      "grad_norm": 22.676437698294823,
      "learning_rate": 4.989369926555855e-07,
      "loss": -0.0,
      "reward": 1.671875,
      "reward_std": 0.6620141267776489,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 0.96875,
      "step": 45
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 130.28125,
      "epoch": 0.0022226517201391575,
      "grad_norm": 12.242025461478375,
      "learning_rate": 4.98912833397758e-07,
      "loss": -0.0,
      "reward": 1.734375,
      "reward_std": 0.5468531847000122,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 46
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 121.40625,
      "epoch": 0.0022709702357943565,
      "grad_norm": 13.22296327516224,
      "learning_rate": 4.988886741399304e-07,
      "loss": -0.0,
      "reward": 1.6875,
      "reward_std": 0.521652102470398,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 0.96875,
      "step": 47
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 128.625,
      "epoch": 0.0023192887514495554,
      "grad_norm": 21.12891784443542,
      "learning_rate": 4.988645148821028e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.5390850305557251,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 48
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 145.09375,
      "epoch": 0.0023676072671047544,
      "grad_norm": 11.45415164792251,
      "learning_rate": 4.988403556242752e-07,
      "loss": -0.0,
      "reward": 1.671875,
      "reward_std": 0.546349287033081,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 0.96875,
      "step": 49
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 141.03125,
      "epoch": 0.0024159257827599534,
      "grad_norm": 14.66939609177466,
      "learning_rate": 4.988161963664476e-07,
      "loss": -0.0,
      "reward": 1.453125,
      "reward_std": 0.5728911757469177,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 0.96875,
      "step": 50
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 132.625,
      "epoch": 0.002464244298415153,
      "grad_norm": 18.14584333116273,
      "learning_rate": 4.9879203710862e-07,
      "loss": -0.0,
      "reward": 1.578125,
      "reward_std": 0.7105082869529724,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 0.96875,
      "step": 51
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 133.21875,
      "epoch": 0.002512562814070352,
      "grad_norm": 16.94940098594036,
      "learning_rate": 4.987678778507924e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.7372558116912842,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 0.90625,
      "step": 52
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 143.21875,
      "epoch": 0.0025608813297255508,
      "grad_norm": 10.308990985030967,
      "learning_rate": 4.987437185929649e-07,
      "loss": -0.0,
      "reward": 1.90625,
      "reward_std": 0.5363268852233887,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 0.9375,
      "step": 53
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 115.53125,
      "epoch": 0.0026091998453807498,
      "grad_norm": 13.668327387972326,
      "learning_rate": 4.987195593351372e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.6138861179351807,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 0.96875,
      "step": 54
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 122.1875,
      "epoch": 0.002657518361035949,
      "grad_norm": 48.46388714909905,
      "learning_rate": 4.986954000773096e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.7344688177108765,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 0.96875,
      "step": 55
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 124.125,
      "epoch": 0.002705836876691148,
      "grad_norm": 22.271513707551513,
      "learning_rate": 4.98671240819482e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.6098296642303467,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 0.90625,
      "step": 56
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 149.0625,
      "epoch": 0.002754155392346347,
      "grad_norm": 16.90105625567009,
      "learning_rate": 4.986470815616544e-07,
      "loss": -0.0,
      "reward": 1.328125,
      "reward_std": 0.49564939737319946,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 57
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 142.15625,
      "epoch": 0.002802473908001546,
      "grad_norm": 19.82870783589864,
      "learning_rate": 4.986229223038268e-07,
      "loss": -0.0,
      "reward": 1.5,
      "reward_std": 0.694413423538208,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 58
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 137.9375,
      "epoch": 0.002850792423656745,
      "grad_norm": 17.1155362510915,
      "learning_rate": 4.985987630459992e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.5624590516090393,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 59
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 133.625,
      "epoch": 0.0028991109393119445,
      "grad_norm": 19.499885629652614,
      "learning_rate": 4.985746037881716e-07,
      "loss": -0.0,
      "reward": 1.421875,
      "reward_std": 0.7001520395278931,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 0.96875,
      "step": 60
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 157.5625,
      "epoch": 0.0029474294549671435,
      "grad_norm": 8.568393819395883,
      "learning_rate": 4.985504445303441e-07,
      "loss": -0.0,
      "reward": 1.578125,
      "reward_std": 0.46929195523262024,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 0.96875,
      "step": 61
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 124.78125,
      "epoch": 0.0029957479706223425,
      "grad_norm": 15.380871523491207,
      "learning_rate": 4.985262852725164e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.6490083336830139,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 62
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 136.90625,
      "epoch": 0.0030440664862775415,
      "grad_norm": 20.448479173663266,
      "learning_rate": 4.985021260146889e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.36722296476364136,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 63
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 113.3125,
      "epoch": 0.0030923850019327404,
      "grad_norm": 5.547087122378533,
      "learning_rate": 4.984779667568612e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 64
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 133.0625,
      "epoch": 0.00314070351758794,
      "grad_norm": 10.134440937265644,
      "learning_rate": 4.984538074990336e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.4953651428222656,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 65
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 120.1875,
      "epoch": 0.003189022033243139,
      "grad_norm": 23.985136592518543,
      "learning_rate": 4.98429648241206e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.5041035413742065,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 66
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 117.0625,
      "epoch": 0.003237340548898338,
      "grad_norm": 20.00012601415987,
      "learning_rate": 4.984054889833784e-07,
      "loss": -0.0,
      "reward": 2.09375,
      "reward_std": 0.40411296486854553,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 67
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 128.75,
      "epoch": 0.003285659064553537,
      "grad_norm": 14.592271619412681,
      "learning_rate": 4.983813297255508e-07,
      "loss": -0.0,
      "reward": 1.375,
      "reward_std": 0.6309332847595215,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 0.96875,
      "step": 68
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 128.3125,
      "epoch": 0.0033339775802087358,
      "grad_norm": 19.77834913125633,
      "learning_rate": 4.983571704677232e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.6018744707107544,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 69
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 131.15625,
      "epoch": 0.003382296095863935,
      "grad_norm": 11.208535990776223,
      "learning_rate": 4.983330112098957e-07,
      "loss": -0.0,
      "reward": 1.90625,
      "reward_std": 0.46516215801239014,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 70
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 134.5625,
      "epoch": 0.003430614611519134,
      "grad_norm": 18.20625270224352,
      "learning_rate": 4.98308851952068e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.6730197668075562,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 0.96875,
      "step": 71
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 136.125,
      "epoch": 0.003478933127174333,
      "grad_norm": 34.885353084240165,
      "learning_rate": 4.982846926942404e-07,
      "loss": -0.0,
      "reward": 1.5,
      "reward_std": 0.35479748249053955,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 72
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 125.5625,
      "epoch": 0.003527251642829532,
      "grad_norm": 20.340489494969237,
      "learning_rate": 4.982605334364128e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.6237938404083252,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 73
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 133.875,
      "epoch": 0.0035755701584847316,
      "grad_norm": 11.943538566727833,
      "learning_rate": 4.982363741785852e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.5649483799934387,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 74
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 148.4375,
      "epoch": 0.0036238886741399305,
      "grad_norm": 12.965633529603748,
      "learning_rate": 4.982122149207576e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.6983485221862793,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 0.9375,
      "step": 75
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 122.71875,
      "epoch": 0.0036722071897951295,
      "grad_norm": 34.14204567454802,
      "learning_rate": 4.981880556629299e-07,
      "loss": -0.0,
      "reward": 1.484375,
      "reward_std": 0.6826589107513428,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 76
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 134.625,
      "epoch": 0.0037205257054503285,
      "grad_norm": 12.169318846228924,
      "learning_rate": 4.981638964051025e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.6018790006637573,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 0.9375,
      "step": 77
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 129.03125,
      "epoch": 0.0037688442211055275,
      "grad_norm": 15.113186118180767,
      "learning_rate": 4.981397371472748e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.5727479457855225,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 78
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 116.34375,
      "epoch": 0.003817162736760727,
      "grad_norm": 25.250662722307826,
      "learning_rate": 4.981155778894472e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.6652344465255737,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 0.96875,
      "step": 79
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 108.5625,
      "epoch": 0.003865481252415926,
      "grad_norm": 11.046320410614,
      "learning_rate": 4.980914186316196e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.46244120597839355,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 80
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 114.6875,
      "epoch": 0.003913799768071125,
      "grad_norm": 29.58552750737202,
      "learning_rate": 4.98067259373792e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.42873120307922363,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 81
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 105.1875,
      "epoch": 0.003962118283726324,
      "grad_norm": 18.903766642842676,
      "learning_rate": 4.980431001159644e-07,
      "loss": -0.0,
      "reward": 2.078125,
      "reward_std": 0.4742468595504761,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 82
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 132.65625,
      "epoch": 0.004010436799381523,
      "grad_norm": 11.662191089169022,
      "learning_rate": 4.980189408581368e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.5518776178359985,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 83
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 119.5625,
      "epoch": 0.004058755315036722,
      "grad_norm": 13.479971811505692,
      "learning_rate": 4.979947816003092e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.3937711715698242,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 84
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 137.46875,
      "epoch": 0.004107073830691921,
      "grad_norm": 14.628264594731768,
      "learning_rate": 4.979706223424816e-07,
      "loss": -0.0,
      "reward": 1.625,
      "reward_std": 0.45782607793807983,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 85
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 127.1875,
      "epoch": 0.00415539234634712,
      "grad_norm": 16.897759663558148,
      "learning_rate": 4.97946463084654e-07,
      "loss": -0.0,
      "reward": 1.40625,
      "reward_std": 0.6314173936843872,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 86
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 111.625,
      "epoch": 0.00420371086200232,
      "grad_norm": 17.84324372355478,
      "learning_rate": 4.979223038268264e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.4198554754257202,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 87
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 126.65625,
      "epoch": 0.004252029377657518,
      "grad_norm": 9.798162353175833,
      "learning_rate": 4.978981445689988e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.39774754643440247,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 88
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 131.59375,
      "epoch": 0.004300347893312718,
      "grad_norm": 27.429920302182882,
      "learning_rate": 4.978739853111712e-07,
      "loss": 0.0,
      "reward": 1.390625,
      "reward_std": 0.6602194309234619,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 0.96875,
      "step": 89
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 128.15625,
      "epoch": 0.004348666408967916,
      "grad_norm": 14.570386938263049,
      "learning_rate": 4.978498260533436e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.5281927585601807,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 90
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 119.1875,
      "epoch": 0.0043969849246231155,
      "grad_norm": 13.67439252214839,
      "learning_rate": 4.97825666795516e-07,
      "loss": -0.0,
      "reward": 1.59375,
      "reward_std": 0.5859901309013367,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 91
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 111.125,
      "epoch": 0.004445303440278315,
      "grad_norm": 9.17931463690914,
      "learning_rate": 4.978015075376885e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.44018012285232544,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 92
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 110.78125,
      "epoch": 0.0044936219559335135,
      "grad_norm": 398.5941792328713,
      "learning_rate": 4.977773482798609e-07,
      "loss": -0.0,
      "reward": 1.34375,
      "reward_std": 0.5207738280296326,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 93
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 110.6875,
      "epoch": 0.004541940471588713,
      "grad_norm": 7.587019196776143,
      "learning_rate": 4.977531890220332e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.36722296476364136,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 94
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 101.9375,
      "epoch": 0.0045902589872439115,
      "grad_norm": 17.715735048600738,
      "learning_rate": 4.977290297642056e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.5497657656669617,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 95
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 116.4375,
      "epoch": 0.004638577502899111,
      "grad_norm": 14.666357629157966,
      "learning_rate": 4.97704870506378e-07,
      "loss": -0.0,
      "reward": 1.625,
      "reward_std": 0.6466031074523926,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 96
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 108.65625,
      "epoch": 0.00468689601855431,
      "grad_norm": 14.177919317050742,
      "learning_rate": 4.976807112485504e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.4976017475128174,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 97
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 113.15625,
      "epoch": 0.004735214534209509,
      "grad_norm": 19.079724008713594,
      "learning_rate": 4.976565519907228e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.6612651944160461,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 98
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 106.65625,
      "epoch": 0.004783533049864708,
      "grad_norm": 20.785328581914964,
      "learning_rate": 4.976323927328952e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.6157901287078857,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 99
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 114.71875,
      "epoch": 0.004831851565519907,
      "grad_norm": 21.13645604025988,
      "learning_rate": 4.976082334750677e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.5848015546798706,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 100
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 104.34375,
      "epoch": 0.004880170081175106,
      "grad_norm": 27.58768221958009,
      "learning_rate": 4.975840742172401e-07,
      "loss": -0.0,
      "reward": 1.5625,
      "reward_std": 0.6792572736740112,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 101
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 107.375,
      "epoch": 0.004928488596830306,
      "grad_norm": 13.903395829891279,
      "learning_rate": 4.975599149594124e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.6147273778915405,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 102
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 103.34375,
      "epoch": 0.004976807112485504,
      "grad_norm": 11.10683627179353,
      "learning_rate": 4.975357557015849e-07,
      "loss": -0.0,
      "reward": 1.5625,
      "reward_std": 0.37868884205818176,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 103
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 129.75,
      "epoch": 0.005025125628140704,
      "grad_norm": 10.719563589498037,
      "learning_rate": 4.975115964437572e-07,
      "loss": -0.0,
      "reward": 1.5625,
      "reward_std": 0.3471825420856476,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 104
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 99.15625,
      "epoch": 0.005073444143795903,
      "grad_norm": 11.901545100214271,
      "learning_rate": 4.974874371859296e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.4822518825531006,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 105
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 101.5625,
      "epoch": 0.0051217626594511016,
      "grad_norm": 12.39769682154333,
      "learning_rate": 4.97463277928102e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.38675177097320557,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 106
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 112.75,
      "epoch": 0.005170081175106301,
      "grad_norm": 10.105285066531465,
      "learning_rate": 4.974391186702744e-07,
      "loss": 0.0,
      "reward": 1.5,
      "reward_std": 0.32082509994506836,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 107
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 110.75,
      "epoch": 0.0052183996907614995,
      "grad_norm": 15.353422064956565,
      "learning_rate": 4.974149594124469e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.5041035413742065,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 108
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 100.1875,
      "epoch": 0.005266718206416699,
      "grad_norm": 14.857606341402413,
      "learning_rate": 4.973908001546192e-07,
      "loss": -0.0,
      "reward": 1.75,
      "reward_std": 0.5723869204521179,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 109
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 103.0,
      "epoch": 0.005315036722071898,
      "grad_norm": 19.901185889316643,
      "learning_rate": 4.973666408967917e-07,
      "loss": 0.0,
      "reward": 1.546875,
      "reward_std": 0.6133956909179688,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 110
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 96.46875,
      "epoch": 0.005363355237727097,
      "grad_norm": 22.472987006046765,
      "learning_rate": 4.97342481638964e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.5854380130767822,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 111
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 101.6875,
      "epoch": 0.005411673753382296,
      "grad_norm": 11.192191720952609,
      "learning_rate": 4.973183223811364e-07,
      "loss": 0.0,
      "reward": 1.40625,
      "reward_std": 0.39879563450813293,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 0.96875,
      "step": 112
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 107.71875,
      "epoch": 0.005459992269037495,
      "grad_norm": 13.436474732765719,
      "learning_rate": 4.972941631233088e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.4419417381286621,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 113
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 105.78125,
      "epoch": 0.005508310784692694,
      "grad_norm": 21.79636664638492,
      "learning_rate": 4.972700038654812e-07,
      "loss": -0.0,
      "reward": 1.46875,
      "reward_std": 0.6550866365432739,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 114
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 95.0,
      "epoch": 0.005556629300347894,
      "grad_norm": 47.207663919295264,
      "learning_rate": 4.972458446076536e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.622909426689148,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 115
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 91.375,
      "epoch": 0.005604947816003092,
      "grad_norm": 128.23384669491804,
      "learning_rate": 4.97221685349826e-07,
      "loss": -0.0,
      "reward": 2.078125,
      "reward_std": 0.3831073045730591,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 116
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 88.96875,
      "epoch": 0.005653266331658292,
      "grad_norm": 20.84824403290395,
      "learning_rate": 4.971975260919985e-07,
      "loss": -0.0,
      "reward": 1.90625,
      "reward_std": 0.50025475025177,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 117
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 89.78125,
      "epoch": 0.00570158484731349,
      "grad_norm": 18.125149967501155,
      "learning_rate": 4.971733668341708e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.3039700984954834,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 118
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 94.9375,
      "epoch": 0.00574990336296869,
      "grad_norm": 19.657395141627504,
      "learning_rate": 4.971492075763432e-07,
      "loss": -0.0,
      "reward": 1.421875,
      "reward_std": 0.48572152853012085,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 119
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 92.40625,
      "epoch": 0.005798221878623889,
      "grad_norm": 19.00943158284224,
      "learning_rate": 4.971250483185156e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.5476284623146057,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 120
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 96.125,
      "epoch": 0.005846540394279088,
      "grad_norm": 30.14098538654748,
      "learning_rate": 4.97100889060688e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.5034401416778564,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 121
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 81.125,
      "epoch": 0.005894858909934287,
      "grad_norm": 15.406102874679373,
      "learning_rate": 4.970767298028604e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.5403927564620972,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 122
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 83.21875,
      "epoch": 0.0059431774255894855,
      "grad_norm": 43.36520702654935,
      "learning_rate": 4.970525705450329e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.6689563989639282,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 123
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 85.71875,
      "epoch": 0.005991495941244685,
      "grad_norm": 12.454457776636495,
      "learning_rate": 4.970284112872053e-07,
      "loss": -0.0,
      "reward": 1.4375,
      "reward_std": 0.35776790976524353,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 124
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.1875,
      "epoch": 0.006039814456899884,
      "grad_norm": 12.965878984347723,
      "learning_rate": 4.970042520293776e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.46707719564437866,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 125
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.6875,
      "epoch": 0.006088132972555083,
      "grad_norm": 16.038025769391894,
      "learning_rate": 4.9698009277155e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.25069957971572876,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 126
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 75.71875,
      "epoch": 0.006136451488210282,
      "grad_norm": 17.475705734881046,
      "learning_rate": 4.969559335137224e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.5591002106666565,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 127
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 82.96875,
      "epoch": 0.006184770003865481,
      "grad_norm": 16.368506013388973,
      "learning_rate": 4.969317742558948e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.5264461040496826,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 128
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 76.84375,
      "epoch": 0.00623308851952068,
      "grad_norm": 10.25065826308469,
      "learning_rate": 4.969076149980672e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.23243674635887146,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 129
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 80.40625,
      "epoch": 0.00628140703517588,
      "grad_norm": 12.207335518694682,
      "learning_rate": 4.968834557402396e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.3708576261997223,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 130
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 81.3125,
      "epoch": 0.006329725550831078,
      "grad_norm": 11.42501492228125,
      "learning_rate": 4.968592964824121e-07,
      "loss": -0.0,
      "reward": 1.4375,
      "reward_std": 0.3029785752296448,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 131
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 75.34375,
      "epoch": 0.006378044066486278,
      "grad_norm": 9.86276666507429,
      "learning_rate": 4.968351372245845e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.23400893807411194,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 132
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 84.15625,
      "epoch": 0.006426362582141476,
      "grad_norm": 19.73068468800391,
      "learning_rate": 4.968109779667569e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.6415244936943054,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 133
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 81.84375,
      "epoch": 0.006474681097796676,
      "grad_norm": 32.09290237364445,
      "learning_rate": 4.967868187089293e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.4797552227973938,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 134
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 77.875,
      "epoch": 0.006522999613451875,
      "grad_norm": 15.661439940439186,
      "learning_rate": 4.967626594511016e-07,
      "loss": -0.0,
      "reward": 1.6875,
      "reward_std": 0.47052502632141113,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 135
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 101.28125,
      "epoch": 0.006571318129107074,
      "grad_norm": 126.73872071172691,
      "learning_rate": 4.96738500193274e-07,
      "loss": -0.0,
      "reward": 1.40625,
      "reward_std": 0.6355680227279663,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 136
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 82.3125,
      "epoch": 0.006619636644762273,
      "grad_norm": 20.25276831539019,
      "learning_rate": 4.967143409354464e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.45992228388786316,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 137
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 99.4375,
      "epoch": 0.0066679551604174716,
      "grad_norm": 24.306284511977033,
      "learning_rate": 4.966901816776188e-07,
      "loss": -0.0,
      "reward": 1.34375,
      "reward_std": 0.5853191614151001,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 138
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 77.78125,
      "epoch": 0.006716273676072671,
      "grad_norm": 27.64270248712896,
      "learning_rate": 4.966660224197913e-07,
      "loss": -0.0,
      "reward": 1.40625,
      "reward_std": 0.6183040738105774,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 139
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.0,
      "epoch": 0.00676459219172787,
      "grad_norm": 5.530088363402859,
      "learning_rate": 4.966418631619637e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.23568855226039886,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 140
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 76.1875,
      "epoch": 0.006812910707383069,
      "grad_norm": 17.348402524079056,
      "learning_rate": 4.966177039041361e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.4330676198005676,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 141
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 70.25,
      "epoch": 0.006861229223038268,
      "grad_norm": 21.062438800827973,
      "learning_rate": 4.965935446463084e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.4947689175605774,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 142
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 70.5,
      "epoch": 0.006909547738693468,
      "grad_norm": 21.885350321578997,
      "learning_rate": 4.965693853884809e-07,
      "loss": -0.0,
      "reward": 1.375,
      "reward_std": 0.28104937076568604,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 143
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 61.0,
      "epoch": 0.006957866254348666,
      "grad_norm": 7.885651331156384,
      "learning_rate": 4.965452261306532e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 144
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 74.625,
      "epoch": 0.007006184770003866,
      "grad_norm": 15.704090085655844,
      "learning_rate": 4.965210668728256e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 145
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 74.21875,
      "epoch": 0.007054503285659064,
      "grad_norm": 41.93721667131592,
      "learning_rate": 4.96496907614998e-07,
      "loss": -0.0,
      "reward": 1.46875,
      "reward_std": 0.4783255457878113,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 146
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 71.25,
      "epoch": 0.007102821801314264,
      "grad_norm": 28.66501508156416,
      "learning_rate": 4.964727483571705e-07,
      "loss": -0.0,
      "reward": 1.6875,
      "reward_std": 0.5249190926551819,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 147
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 67.75,
      "epoch": 0.007151140316969463,
      "grad_norm": 33.9230996440207,
      "learning_rate": 4.964485890993429e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.6607464551925659,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 148
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 65.0625,
      "epoch": 0.007199458832624662,
      "grad_norm": 15.596411085036005,
      "learning_rate": 4.964244298415152e-07,
      "loss": -0.0,
      "reward": 1.546875,
      "reward_std": 0.5225197076797485,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 149
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 64.59375,
      "epoch": 0.007247777348279861,
      "grad_norm": 19.321762400987925,
      "learning_rate": 4.964002705836877e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.5220552682876587,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 150
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 74.21875,
      "epoch": 0.00729609586393506,
      "grad_norm": 24.376703513209282,
      "learning_rate": 4.9637611132586e-07,
      "loss": -0.0,
      "reward": 1.921875,
      "reward_std": 0.4441883862018585,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 151
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 63.90625,
      "epoch": 0.007344414379590259,
      "grad_norm": 13.964762878473822,
      "learning_rate": 4.963519520680324e-07,
      "loss": 0.0,
      "reward": 1.09375,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.0625,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 152
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 57.34375,
      "epoch": 0.0073927328952454584,
      "grad_norm": 34.56766285826505,
      "learning_rate": 4.963277928102048e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.28247907757759094,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 153
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 66.5,
      "epoch": 0.007441051410900657,
      "grad_norm": 25.88742857586725,
      "learning_rate": 4.963036335523773e-07,
      "loss": -0.0,
      "reward": 1.59375,
      "reward_std": 0.6108826398849487,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 154
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 58.25,
      "epoch": 0.007489369926555856,
      "grad_norm": 25.07426217642962,
      "learning_rate": 4.962794742945497e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.32666346430778503,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 155
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 68.5,
      "epoch": 0.007537688442211055,
      "grad_norm": 24.38381360636354,
      "learning_rate": 4.96255315036722e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 0.96875,
      "step": 156
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 78.75,
      "epoch": 0.007586006957866254,
      "grad_norm": 14.5122528851565,
      "learning_rate": 4.962311557788945e-07,
      "loss": -0.0,
      "reward": 1.46875,
      "reward_std": 0.5303300619125366,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 0.96875,
      "step": 157
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.34375,
      "epoch": 0.007634325473521454,
      "grad_norm": 0.0,
      "learning_rate": 4.962069965210668e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 158
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 60.84375,
      "epoch": 0.007682643989176652,
      "grad_norm": 39.008467545592765,
      "learning_rate": 4.961828372632392e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.3404400944709778,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 159
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 62.4375,
      "epoch": 0.007730962504831852,
      "grad_norm": 13.849874544128348,
      "learning_rate": 4.961586780054116e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.47617846727371216,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 160
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 72.9375,
      "epoch": 0.00777928102048705,
      "grad_norm": 15.727864445568422,
      "learning_rate": 4.96134518747584e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.46510523557662964,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 161
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 76.71875,
      "epoch": 0.00782759953614225,
      "grad_norm": 21.93078047053228,
      "learning_rate": 4.961103594897565e-07,
      "loss": -0.0,
      "reward": 1.359375,
      "reward_std": 0.47976505756378174,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 162
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 70.09375,
      "epoch": 0.007875918051797449,
      "grad_norm": 18.0960029826763,
      "learning_rate": 4.960862002319289e-07,
      "loss": -0.0,
      "reward": 1.921875,
      "reward_std": 0.5132404565811157,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 163
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 72.0625,
      "epoch": 0.007924236567452649,
      "grad_norm": 50.58954434341022,
      "learning_rate": 4.960620409741013e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.39310699701309204,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 164
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 64.0625,
      "epoch": 0.007972555083107846,
      "grad_norm": 28.820057750010005,
      "learning_rate": 4.960378817162737e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.45233404636383057,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 165
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 61.6875,
      "epoch": 0.008020873598763046,
      "grad_norm": 123.34244222360196,
      "learning_rate": 4.96013722458446e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.5497305393218994,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 0.96875,
      "step": 166
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 72.4375,
      "epoch": 0.008069192114418245,
      "grad_norm": 23.526477900841567,
      "learning_rate": 4.959895632006184e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.685854434967041,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 167
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 63.875,
      "epoch": 0.008117510630073444,
      "grad_norm": 14.806401660931433,
      "learning_rate": 4.959654039427908e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 168
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 66.71875,
      "epoch": 0.008165829145728644,
      "grad_norm": 18.852603907123132,
      "learning_rate": 4.959412446849632e-07,
      "loss": -0.0,
      "reward": 1.4375,
      "reward_std": 0.519324541091919,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 169
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 69.8125,
      "epoch": 0.008214147661383842,
      "grad_norm": 22.857143055845214,
      "learning_rate": 4.959170854271357e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.5591002106666565,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 170
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 61.3125,
      "epoch": 0.008262466177039041,
      "grad_norm": 12.52693814534448,
      "learning_rate": 4.958929261693081e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.41457754373550415,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 171
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 62.03125,
      "epoch": 0.00831078469269424,
      "grad_norm": 26.620086117811656,
      "learning_rate": 4.958687669114805e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.5062023997306824,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 172
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 60.59375,
      "epoch": 0.00835910320834944,
      "grad_norm": 25.24930590275252,
      "learning_rate": 4.958446076536529e-07,
      "loss": 0.0,
      "reward": 1.40625,
      "reward_std": 0.3503679633140564,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 173
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 55.34375,
      "epoch": 0.00840742172400464,
      "grad_norm": 32.31016173988275,
      "learning_rate": 4.958204483958253e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.636022686958313,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 174
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 61.15625,
      "epoch": 0.008455740239659837,
      "grad_norm": 19.10223258070406,
      "learning_rate": 4.957962891379976e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.6206084489822388,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 175
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 58.03125,
      "epoch": 0.008504058755315036,
      "grad_norm": 9.816356705118197,
      "learning_rate": 4.9577212988017e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 176
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 53.25,
      "epoch": 0.008552377270970236,
      "grad_norm": 14.697513526107805,
      "learning_rate": 4.957479706223424e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.38656681776046753,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 177
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 59.5,
      "epoch": 0.008600695786625435,
      "grad_norm": 32.734295036635764,
      "learning_rate": 4.957238113645149e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.5066196918487549,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 178
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 59.75,
      "epoch": 0.008649014302280635,
      "grad_norm": 12.486864826612576,
      "learning_rate": 4.956996521066873e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.42036867141723633,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 179
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 60.4375,
      "epoch": 0.008697332817935832,
      "grad_norm": 10.007151199690197,
      "learning_rate": 4.956754928488597e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.2226182520389557,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 180
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 54.1875,
      "epoch": 0.008745651333591032,
      "grad_norm": 17.185007259715615,
      "learning_rate": 4.956513335910321e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.5981680154800415,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 181
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 55.40625,
      "epoch": 0.008793969849246231,
      "grad_norm": 13.460187659703033,
      "learning_rate": 4.956271743332044e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.5613039135932922,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 182
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 45.96875,
      "epoch": 0.00884228836490143,
      "grad_norm": 19.304968573564,
      "learning_rate": 4.956030150753769e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.3234682083129883,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 183
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 52.40625,
      "epoch": 0.00889060688055663,
      "grad_norm": 147.2328805485989,
      "learning_rate": 4.955788558175492e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.4214168190956116,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 184
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 58.78125,
      "epoch": 0.008938925396211828,
      "grad_norm": 30.93743233087772,
      "learning_rate": 4.955546965597216e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.6346027851104736,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 185
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 53.21875,
      "epoch": 0.008987243911867027,
      "grad_norm": 10.308578635601007,
      "learning_rate": 4.955305373018941e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 186
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 54.59375,
      "epoch": 0.009035562427522226,
      "grad_norm": 12.604130881526228,
      "learning_rate": 4.955063780440665e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.3908922076225281,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 187
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 45.4375,
      "epoch": 0.009083880943177426,
      "grad_norm": 21.933727030812232,
      "learning_rate": 4.954822187862389e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.5057574510574341,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 188
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 49.84375,
      "epoch": 0.009132199458832625,
      "grad_norm": 23.426637132835726,
      "learning_rate": 4.954580595284112e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.6227300763130188,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 189
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 53.6875,
      "epoch": 0.009180517974487823,
      "grad_norm": 9.79675469315397,
      "learning_rate": 4.954339002705837e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.39347878098487854,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 190
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 50.125,
      "epoch": 0.009228836490143022,
      "grad_norm": 14.579096576502527,
      "learning_rate": 4.95409741012756e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.43768274784088135,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 191
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 55.125,
      "epoch": 0.009277155005798222,
      "grad_norm": 13.85587380058334,
      "learning_rate": 4.953855817549284e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.35520076751708984,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 192
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 63.03125,
      "epoch": 0.009325473521453421,
      "grad_norm": 16.201133826120035,
      "learning_rate": 4.953614224971009e-07,
      "loss": -0.0,
      "reward": 1.34375,
      "reward_std": 0.49760177731513977,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 193
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 54.71875,
      "epoch": 0.00937379203710862,
      "grad_norm": 27.95362561195346,
      "learning_rate": 4.953372632392733e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.5624581575393677,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 194
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 47.8125,
      "epoch": 0.009422110552763818,
      "grad_norm": 18.785145082308755,
      "learning_rate": 4.953131039814457e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.7009119987487793,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 195
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 39.5625,
      "epoch": 0.009470429068419018,
      "grad_norm": 20.199077870270866,
      "learning_rate": 4.95288944723618e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.4387563169002533,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 196
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 61.0,
      "epoch": 0.009518747584074217,
      "grad_norm": 23.94602480063214,
      "learning_rate": 4.952647854657905e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.4592459797859192,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 197
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 52.21875,
      "epoch": 0.009567066099729417,
      "grad_norm": 14.627821175124815,
      "learning_rate": 4.952406262079628e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.5571421980857849,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 198
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 44.03125,
      "epoch": 0.009615384615384616,
      "grad_norm": 5.786054598453081,
      "learning_rate": 4.952164669501352e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 199
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 53.5,
      "epoch": 0.009663703131039814,
      "grad_norm": 12.934618620052682,
      "learning_rate": 4.951923076923076e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.4037913680076599,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 200
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 45.4375,
      "epoch": 0.009712021646695013,
      "grad_norm": 33.963454608848856,
      "learning_rate": 4.951681484344801e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.5730265378952026,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 201
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 37.21875,
      "epoch": 0.009760340162350212,
      "grad_norm": 0.0,
      "learning_rate": 4.951439891766525e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 202
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 44.34375,
      "epoch": 0.009808658678005412,
      "grad_norm": 11.73151163992609,
      "learning_rate": 4.951198299188249e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.4500974118709564,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 203
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 43.21875,
      "epoch": 0.009856977193660611,
      "grad_norm": 10.518435196654039,
      "learning_rate": 4.950956706609973e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.3877224326133728,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 204
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 41.5625,
      "epoch": 0.00990529570931581,
      "grad_norm": 28.40789611271811,
      "learning_rate": 4.950715114031697e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.4019092619419098,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 205
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 50.625,
      "epoch": 0.009953614224971008,
      "grad_norm": 18.43282093800467,
      "learning_rate": 4.95047352145342e-07,
      "loss": -0.0,
      "reward": 1.59375,
      "reward_std": 0.49760180711746216,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 206
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 38.65625,
      "epoch": 0.010001932740626208,
      "grad_norm": 12.00064407581638,
      "learning_rate": 4.950231928875144e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.42466527223587036,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 207
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 37.625,
      "epoch": 0.010050251256281407,
      "grad_norm": 10.765234478995449,
      "learning_rate": 4.949990336296868e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.2550578713417053,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 208
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 44.59375,
      "epoch": 0.010098569771936607,
      "grad_norm": 13.49050316695399,
      "learning_rate": 4.949748743718593e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.4491388201713562,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 209
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 40.25,
      "epoch": 0.010146888287591806,
      "grad_norm": 13.1273477114429,
      "learning_rate": 4.949507151140317e-07,
      "loss": -0.0,
      "reward": 1.671875,
      "reward_std": 0.3676722049713135,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 210
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 39.4375,
      "epoch": 0.010195206803247004,
      "grad_norm": 8.685392645113213,
      "learning_rate": 4.949265558562041e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 211
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 39.96875,
      "epoch": 0.010243525318902203,
      "grad_norm": 0.0,
      "learning_rate": 4.949023965983765e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 212
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 43.8125,
      "epoch": 0.010291843834557403,
      "grad_norm": 13.360147786224104,
      "learning_rate": 4.948782373405489e-07,
      "loss": -0.0,
      "reward": 1.6875,
      "reward_std": 0.25646141171455383,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 213
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 36.15625,
      "epoch": 0.010340162350212602,
      "grad_norm": 0.0,
      "learning_rate": 4.948540780827213e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 214
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 36.15625,
      "epoch": 0.010388480865867801,
      "grad_norm": 0.0,
      "learning_rate": 4.948299188248936e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 215
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 42.5,
      "epoch": 0.010436799381522999,
      "grad_norm": 16.67864572265595,
      "learning_rate": 4.94805759567066e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 216
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 41.78125,
      "epoch": 0.010485117897178198,
      "grad_norm": 22.69390571024235,
      "learning_rate": 4.947816003092385e-07,
      "loss": -0.0,
      "reward": 1.6875,
      "reward_std": 0.49441635608673096,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 217
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 45.9375,
      "epoch": 0.010533436412833398,
      "grad_norm": 10.226550448978207,
      "learning_rate": 4.947574410514109e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.31983357667922974,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 218
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 39.40625,
      "epoch": 0.010581754928488597,
      "grad_norm": 10.598070024450228,
      "learning_rate": 4.947332817935833e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.27467289566993713,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 219
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 36.15625,
      "epoch": 0.010630073444143797,
      "grad_norm": 6.7297327303970365,
      "learning_rate": 4.947091225357557e-07,
      "loss": -0.0,
      "reward": 2.078125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 220
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 40.28125,
      "epoch": 0.010678391959798994,
      "grad_norm": 4.960869263767019,
      "learning_rate": 4.946849632779281e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 221
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 38.40625,
      "epoch": 0.010726710475454194,
      "grad_norm": 11.406375882301004,
      "learning_rate": 4.946608040201004e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 222
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 35.84375,
      "epoch": 0.010775028991109393,
      "grad_norm": 14.782281112513623,
      "learning_rate": 4.946366447622729e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.3650856614112854,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 223
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 40.4375,
      "epoch": 0.010823347506764593,
      "grad_norm": 4.930799983822777,
      "learning_rate": 4.946124855044453e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 224
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 43.28125,
      "epoch": 0.010871666022419792,
      "grad_norm": 30.003779136512517,
      "learning_rate": 4.945883262466177e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 225
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 36.09375,
      "epoch": 0.01091998453807499,
      "grad_norm": 5.476000446521705,
      "learning_rate": 4.945641669887901e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 226
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 48.03125,
      "epoch": 0.010968303053730189,
      "grad_norm": 16.115557415023318,
      "learning_rate": 4.945400077309625e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.584973156452179,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 227
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 41.46875,
      "epoch": 0.011016621569385389,
      "grad_norm": 4.5900353472223685,
      "learning_rate": 4.945158484731349e-07,
      "loss": 0.0,
      "reward": 1.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.03125,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 228
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 38.0625,
      "epoch": 0.011064940085040588,
      "grad_norm": 17.032957158967697,
      "learning_rate": 4.944916892153072e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.36278262734413147,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 229
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 40.8125,
      "epoch": 0.011113258600695787,
      "grad_norm": 17.05253566669222,
      "learning_rate": 4.944675299574797e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.493342787027359,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 230
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 34.0,
      "epoch": 0.011161577116350985,
      "grad_norm": 0.0,
      "learning_rate": 4.94443370699652e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 231
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 38.4375,
      "epoch": 0.011209895632006184,
      "grad_norm": 20.42618871766826,
      "learning_rate": 4.944192114418245e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 232
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 36.96875,
      "epoch": 0.011258214147661384,
      "grad_norm": 63.843669752037925,
      "learning_rate": 4.943950521839969e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.27841317653656006,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 233
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 46.375,
      "epoch": 0.011306532663316583,
      "grad_norm": 34.484414325107544,
      "learning_rate": 4.943708929261693e-07,
      "loss": -0.0,
      "reward": 2.328125,
      "reward_std": 0.31512534618377686,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 234
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 44.5,
      "epoch": 0.011354851178971783,
      "grad_norm": 17.511372225207666,
      "learning_rate": 4.943467336683417e-07,
      "loss": -0.0,
      "reward": 1.515625,
      "reward_std": 0.37992191314697266,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 235
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 42.375,
      "epoch": 0.01140316969462698,
      "grad_norm": 25.02437748967537,
      "learning_rate": 4.943225744105141e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.5872231721878052,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 236
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 40.84375,
      "epoch": 0.01145148821028218,
      "grad_norm": 8.414650595007945,
      "learning_rate": 4.942984151526865e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.3608423173427582,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 237
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 49.875,
      "epoch": 0.01149980672593738,
      "grad_norm": 18.761321341348655,
      "learning_rate": 4.942742558948588e-07,
      "loss": 0.0,
      "reward": 1.53125,
      "reward_std": 0.5239396095275879,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 238
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 48.53125,
      "epoch": 0.011548125241592579,
      "grad_norm": 18.795809874998945,
      "learning_rate": 4.942500966370312e-07,
      "loss": 0.0,
      "reward": 1.546875,
      "reward_std": 0.40724146366119385,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 239
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 44.75,
      "epoch": 0.011596443757247778,
      "grad_norm": 11.336757724009415,
      "learning_rate": 4.942259373792037e-07,
      "loss": -0.0,
      "reward": 1.75,
      "reward_std": 0.27102428674697876,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 240
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 40.25,
      "epoch": 0.011644762272902976,
      "grad_norm": 0.0,
      "learning_rate": 4.942017781213761e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 241
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 53.6875,
      "epoch": 0.011693080788558175,
      "grad_norm": 17.77154870945557,
      "learning_rate": 4.941776188635485e-07,
      "loss": 0.0,
      "reward": 1.265625,
      "reward_std": 0.49564939737319946,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 242
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 40.53125,
      "epoch": 0.011741399304213375,
      "grad_norm": 6.12616057097276,
      "learning_rate": 4.941534596057209e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.24251843988895416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 243
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 50.46875,
      "epoch": 0.011789717819868574,
      "grad_norm": 9.18912878710964,
      "learning_rate": 4.941293003478933e-07,
      "loss": -0.0,
      "reward": 1.25,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 244
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 41.75,
      "epoch": 0.011838036335523773,
      "grad_norm": 5.161698178636418,
      "learning_rate": 4.941051410900657e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 245
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 41.15625,
      "epoch": 0.011886354851178971,
      "grad_norm": 5.632367513429027,
      "learning_rate": 4.94080981832238e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 246
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 46.0,
      "epoch": 0.01193467336683417,
      "grad_norm": 0.0,
      "learning_rate": 4.940568225744106e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 247
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 43.125,
      "epoch": 0.01198299188248937,
      "grad_norm": 0.0,
      "learning_rate": 4.940326633165829e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 248
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 45.0625,
      "epoch": 0.01203131039814457,
      "grad_norm": 12.629732719670285,
      "learning_rate": 4.940085040587553e-07,
      "loss": 0.0,
      "reward": 1.546875,
      "reward_std": 0.49618881940841675,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 249
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 51.09375,
      "epoch": 0.012079628913799769,
      "grad_norm": 22.127613587087694,
      "learning_rate": 4.939843448009277e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.6370707750320435,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 250
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 46.40625,
      "epoch": 0.012127947429454966,
      "grad_norm": 7.313107969532202,
      "learning_rate": 4.939601855431001e-07,
      "loss": -0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 251
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 45.40625,
      "epoch": 0.012176265945110166,
      "grad_norm": 20.399768840267345,
      "learning_rate": 4.939360262852725e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.49705949425697327,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 252
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 47.125,
      "epoch": 0.012224584460765365,
      "grad_norm": 25.285212971535284,
      "learning_rate": 4.939118670274449e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.6492904424667358,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 253
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 42.125,
      "epoch": 0.012272902976420565,
      "grad_norm": 0.0,
      "learning_rate": 4.938877077696173e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 254
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 39.28125,
      "epoch": 0.012321221492075764,
      "grad_norm": 17.939186544627784,
      "learning_rate": 4.938635485117897e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.6387534141540527,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 255
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 38.96875,
      "epoch": 0.012369540007730962,
      "grad_norm": 12.549196917169176,
      "learning_rate": 4.938393892539621e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.27564918994903564,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 256
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 48.09375,
      "epoch": 0.012417858523386161,
      "grad_norm": 14.819787857767539,
      "learning_rate": 4.938152299961345e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.3945523202419281,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 257
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 46.46875,
      "epoch": 0.01246617703904136,
      "grad_norm": 16.231680621524344,
      "learning_rate": 4.937910707383069e-07,
      "loss": -0.0,
      "reward": 1.578125,
      "reward_std": 0.42553597688674927,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 258
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 39.75,
      "epoch": 0.01251449555469656,
      "grad_norm": 28.363719612752206,
      "learning_rate": 4.937669114804793e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.5928765535354614,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 259
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 33.6875,
      "epoch": 0.01256281407035176,
      "grad_norm": 10.276959007234938,
      "learning_rate": 4.937427522226517e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.2731459438800812,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 260
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 43.21875,
      "epoch": 0.012611132586006957,
      "grad_norm": 19.098469958546346,
      "learning_rate": 4.937185929648241e-07,
      "loss": -0.0,
      "reward": 1.546875,
      "reward_std": 0.6926257014274597,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 261
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 38.90625,
      "epoch": 0.012659451101662157,
      "grad_norm": 11.89989423095446,
      "learning_rate": 4.936944337069964e-07,
      "loss": -0.0,
      "reward": 2.390625,
      "reward_std": 0.2662131190299988,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 262
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 42.59375,
      "epoch": 0.012707769617317356,
      "grad_norm": 17.598835382602292,
      "learning_rate": 4.93670274449169e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.4945741891860962,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 263
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 40.9375,
      "epoch": 0.012756088132972555,
      "grad_norm": 12.082430989592877,
      "learning_rate": 4.936461151913413e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 264
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 42.59375,
      "epoch": 0.012804406648627755,
      "grad_norm": 15.258242372398131,
      "learning_rate": 4.936219559335137e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.5181578397750854,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 265
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 37.46875,
      "epoch": 0.012852725164282952,
      "grad_norm": 12.367517907808656,
      "learning_rate": 4.935977966756861e-07,
      "loss": -0.0,
      "reward": 1.734375,
      "reward_std": 0.35307908058166504,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 266
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 42.71875,
      "epoch": 0.012901043679938152,
      "grad_norm": 25.077173552743186,
      "learning_rate": 4.935736374178585e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.46561095118522644,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 267
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 53.5625,
      "epoch": 0.012949362195593351,
      "grad_norm": 10.29165312119517,
      "learning_rate": 4.935494781600309e-07,
      "loss": -0.0,
      "reward": 1.671875,
      "reward_std": 0.3463020324707031,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 268
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 48.9375,
      "epoch": 0.01299768071124855,
      "grad_norm": 9.618901096021755,
      "learning_rate": 4.935253189022032e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 269
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 43.09375,
      "epoch": 0.01304599922690375,
      "grad_norm": 8.615581289673255,
      "learning_rate": 4.935011596443757e-07,
      "loss": -0.0,
      "reward": 1.515625,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 270
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 46.96875,
      "epoch": 0.013094317742558948,
      "grad_norm": 0.0,
      "learning_rate": 4.934770003865481e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 271
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 43.6875,
      "epoch": 0.013142636258214147,
      "grad_norm": 15.972511988636418,
      "learning_rate": 4.934528411287205e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.31539323925971985,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 272
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 58.5,
      "epoch": 0.013190954773869347,
      "grad_norm": 25.32967037979268,
      "learning_rate": 4.934286818708929e-07,
      "loss": -0.0,
      "reward": 1.625,
      "reward_std": 0.7239210605621338,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 273
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 49.3125,
      "epoch": 0.013239273289524546,
      "grad_norm": 16.918094335035857,
      "learning_rate": 4.934045226130653e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 274
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 56.53125,
      "epoch": 0.013287591805179745,
      "grad_norm": 36.23082567894201,
      "learning_rate": 4.933803633552377e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.38415998220443726,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 275
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 51.375,
      "epoch": 0.013335910320834943,
      "grad_norm": 11.455836750986647,
      "learning_rate": 4.933562040974101e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.39408785104751587,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 276
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 48.8125,
      "epoch": 0.013384228836490143,
      "grad_norm": 12.605619198133944,
      "learning_rate": 4.933320448395825e-07,
      "loss": -0.0,
      "reward": 1.515625,
      "reward_std": 0.24280104041099548,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 277
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 47.375,
      "epoch": 0.013432547352145342,
      "grad_norm": 5.0491911430925045,
      "learning_rate": 4.93307885581755e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 278
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 53.3125,
      "epoch": 0.013480865867800541,
      "grad_norm": 9.679749406716727,
      "learning_rate": 4.932837263239273e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 279
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 56.9375,
      "epoch": 0.01352918438345574,
      "grad_norm": 16.175930588467093,
      "learning_rate": 4.932595670660997e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.46982741355895996,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 280
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.0625,
      "epoch": 0.013577502899110938,
      "grad_norm": 0.0,
      "learning_rate": 4.932354078082721e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 281
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 57.21875,
      "epoch": 0.013625821414766138,
      "grad_norm": 24.411982502160193,
      "learning_rate": 4.932112485504445e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.5581902861595154,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 282
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 57.65625,
      "epoch": 0.013674139930421337,
      "grad_norm": 28.25540386273274,
      "learning_rate": 4.931870892926169e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.41505181789398193,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 283
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 61.75,
      "epoch": 0.013722458446076537,
      "grad_norm": 9.280734281315711,
      "learning_rate": 4.931629300347893e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.2651395797729492,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 284
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 61.65625,
      "epoch": 0.013770776961731736,
      "grad_norm": 0.0,
      "learning_rate": 4.931387707769617e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 285
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 57.46875,
      "epoch": 0.013819095477386936,
      "grad_norm": 16.752816435383988,
      "learning_rate": 4.93114611519134e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.5214560031890869,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 286
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 60.09375,
      "epoch": 0.013867413993042133,
      "grad_norm": 25.041103645215856,
      "learning_rate": 4.930904522613066e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.32666346430778503,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 287
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 66.65625,
      "epoch": 0.013915732508697333,
      "grad_norm": 7.586912231160616,
      "learning_rate": 4.930662930034789e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 288
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 64.96875,
      "epoch": 0.013964051024352532,
      "grad_norm": 19.59084051447298,
      "learning_rate": 4.930421337456513e-07,
      "loss": 0.0,
      "reward": 1.453125,
      "reward_std": 0.30935919284820557,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 289
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 55.03125,
      "epoch": 0.014012369540007731,
      "grad_norm": 65.45408104187993,
      "learning_rate": 4.930179744878237e-07,
      "loss": -0.0,
      "reward": 2.21875,
      "reward_std": 0.4323658347129822,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 290
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 66.4375,
      "epoch": 0.014060688055662931,
      "grad_norm": 10.984255545942489,
      "learning_rate": 4.929938152299961e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 291
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 59.1875,
      "epoch": 0.014109006571318129,
      "grad_norm": 15.782691403591594,
      "learning_rate": 4.929696559721685e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.42429083585739136,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 292
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 62.8125,
      "epoch": 0.014157325086973328,
      "grad_norm": 23.570906233667483,
      "learning_rate": 4.929454967143409e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.5617531538009644,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 293
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 74.4375,
      "epoch": 0.014205643602628527,
      "grad_norm": 46.654200234846876,
      "learning_rate": 4.929213374565134e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.6895453929901123,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 294
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 63.8125,
      "epoch": 0.014253962118283727,
      "grad_norm": 10.570288484116292,
      "learning_rate": 4.928971781986857e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 295
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 64.71875,
      "epoch": 0.014302280633938926,
      "grad_norm": 33.01121847423372,
      "learning_rate": 4.928730189408581e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.5713133811950684,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 296
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 59.53125,
      "epoch": 0.014350599149594124,
      "grad_norm": 10.504849858290903,
      "learning_rate": 4.928488596830305e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.2302001267671585,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 297
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 68.09375,
      "epoch": 0.014398917665249323,
      "grad_norm": 17.855719203133482,
      "learning_rate": 4.928247004252029e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.47569912672042847,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 298
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 64.34375,
      "epoch": 0.014447236180904523,
      "grad_norm": 12.147370571261957,
      "learning_rate": 4.928005411673753e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.43556109070777893,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 299
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 66.9375,
      "epoch": 0.014495554696559722,
      "grad_norm": 8.065561337669322,
      "learning_rate": 4.927763819095477e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.3088849186897278,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 300
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.75,
      "epoch": 0.014543873212214922,
      "grad_norm": 0.0,
      "learning_rate": 4.927522226517201e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 301
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 68.78125,
      "epoch": 0.01459219172787012,
      "grad_norm": 6.044381520200691,
      "learning_rate": 4.927280633938925e-07,
      "loss": 0.0,
      "reward": 1.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.03125,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 302
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 58.9375,
      "epoch": 0.014640510243525319,
      "grad_norm": 7.9411756217437555,
      "learning_rate": 4.92703904136065e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 303
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 64.75,
      "epoch": 0.014688828759180518,
      "grad_norm": 6.107110943616828,
      "learning_rate": 4.926797448782373e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 304
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.28125,
      "epoch": 0.014737147274835717,
      "grad_norm": 0.0,
      "learning_rate": 4.926555856204097e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 305
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 65.625,
      "epoch": 0.014785465790490917,
      "grad_norm": 11.206460912305992,
      "learning_rate": 4.926314263625821e-07,
      "loss": -0.0,
      "reward": 2.375,
      "reward_std": 0.3230288028717041,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 306
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.96875,
      "epoch": 0.014833784306146115,
      "grad_norm": 0.0,
      "learning_rate": 4.926072671047545e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 307
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 68.46875,
      "epoch": 0.014882102821801314,
      "grad_norm": 9.100765547036318,
      "learning_rate": 4.925831078469269e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.28247910737991333,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 308
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 63.8125,
      "epoch": 0.014930421337456513,
      "grad_norm": 10.160690014650246,
      "learning_rate": 4.925589485890994e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.39774754643440247,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 309
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 65.25,
      "epoch": 0.014978739853111713,
      "grad_norm": 10.021697352850792,
      "learning_rate": 4.925347893312718e-07,
      "loss": -0.0,
      "reward": 1.90625,
      "reward_std": 0.31663835048675537,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 310
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 69.09375,
      "epoch": 0.015027058368766912,
      "grad_norm": 8.08006665000897,
      "learning_rate": 4.925106300734441e-07,
      "loss": -0.0,
      "reward": 2.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 311
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 67.46875,
      "epoch": 0.01507537688442211,
      "grad_norm": 7.182623788049111,
      "learning_rate": 4.924864708156165e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 312
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 69.28125,
      "epoch": 0.01512369540007731,
      "grad_norm": 13.133157326292022,
      "learning_rate": 4.924623115577889e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 313
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 70.0,
      "epoch": 0.015172013915732509,
      "grad_norm": 6.848705548570963,
      "learning_rate": 4.924381522999613e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 314
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.15625,
      "epoch": 0.015220332431387708,
      "grad_norm": 0.0,
      "learning_rate": 4.924139930421337e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 315
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 62.21875,
      "epoch": 0.015268650947042908,
      "grad_norm": 10.375225763362028,
      "learning_rate": 4.923898337843061e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.2852528691291809,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 316
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 70.34375,
      "epoch": 0.015316969462698105,
      "grad_norm": 13.525117920341618,
      "learning_rate": 4.923656745264786e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.43980440497398376,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 317
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 51.78125,
      "epoch": 0.015365287978353305,
      "grad_norm": 10.282845920801712,
      "learning_rate": 4.92341515268651e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.3629639744758606,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 318
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 59.71875,
      "epoch": 0.015413606494008504,
      "grad_norm": 11.070099911403988,
      "learning_rate": 4.923173560108233e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 319
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 64.15625,
      "epoch": 0.015461925009663703,
      "grad_norm": 4.693056536020468,
      "learning_rate": 4.922931967529958e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 320
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 61.59375,
      "epoch": 0.015510243525318903,
      "grad_norm": 9.32786319349995,
      "learning_rate": 4.922690374951681e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.4492208957672119,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 321
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 63.9375,
      "epoch": 0.0155585620409741,
      "grad_norm": 43.48193971489509,
      "learning_rate": 4.922448782373405e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.5640522241592407,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 322
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 69.75,
      "epoch": 0.0156068805566293,
      "grad_norm": 15.476442168711648,
      "learning_rate": 4.922207189795129e-07,
      "loss": 0.0,
      "reward": 1.53125,
      "reward_std": 0.4966200590133667,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 323
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.5,
      "epoch": 0.0156551990722845,
      "grad_norm": 0.0,
      "learning_rate": 4.921965597216853e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 324
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 60.40625,
      "epoch": 0.015703517587939697,
      "grad_norm": 6.659442107569846,
      "learning_rate": 4.921724004638578e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 325
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 61.5625,
      "epoch": 0.015751836103594898,
      "grad_norm": 12.935280042302688,
      "learning_rate": 4.921482412060301e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.2505248188972473,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 326
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 63.21875,
      "epoch": 0.015800154619250096,
      "grad_norm": 8.906305826795162,
      "learning_rate": 4.921240819482026e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 327
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 64.75,
      "epoch": 0.015848473134905297,
      "grad_norm": 6.162397000976902,
      "learning_rate": 4.920999226903749e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 328
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 74.28125,
      "epoch": 0.015896791650560495,
      "grad_norm": 27.19768020300501,
      "learning_rate": 4.920757634325473e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 329
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 63.53125,
      "epoch": 0.015945110166215692,
      "grad_norm": 16.951069358570283,
      "learning_rate": 4.920516041747197e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 330
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 69.0625,
      "epoch": 0.015993428681870894,
      "grad_norm": 22.4688040872021,
      "learning_rate": 4.920274449168921e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 331
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 61.78125,
      "epoch": 0.01604174719752609,
      "grad_norm": 15.16068479661949,
      "learning_rate": 4.920032856590645e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.2962459325790405,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 332
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.1875,
      "epoch": 0.016090065713181292,
      "grad_norm": 8.251729007982766,
      "learning_rate": 4.91979126401237e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.22152717411518097,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 333
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.625,
      "epoch": 0.01613838422883649,
      "grad_norm": 0.0,
      "learning_rate": 4.919549671434094e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 334
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 78.5,
      "epoch": 0.016186702744491688,
      "grad_norm": 23.409103560211083,
      "learning_rate": 4.919308078855817e-07,
      "loss": -0.0,
      "reward": 1.359375,
      "reward_std": 0.34957700967788696,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 335
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 85.34375,
      "epoch": 0.01623502126014689,
      "grad_norm": 18.512801896273764,
      "learning_rate": 4.919066486277541e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.32666346430778503,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 336
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 83.90625,
      "epoch": 0.016283339775802087,
      "grad_norm": 15.739780615341578,
      "learning_rate": 4.918824893699265e-07,
      "loss": 0.0,
      "reward": 1.140625,
      "reward_std": 0.39774754643440247,
      "rewards/accuracy_reward": 0.09375,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 337
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 63.5625,
      "epoch": 0.016331658291457288,
      "grad_norm": 23.867991691135316,
      "learning_rate": 4.918583301120989e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.2961486577987671,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 338
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.90625,
      "epoch": 0.016379976807112485,
      "grad_norm": 0.0,
      "learning_rate": 4.918341708542713e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 339
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 61.0,
      "epoch": 0.016428295322767683,
      "grad_norm": 11.228880851331523,
      "learning_rate": 4.918100115964438e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 340
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 69.75,
      "epoch": 0.016476613838422884,
      "grad_norm": 10.880109096342489,
      "learning_rate": 4.917858523386162e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.18174266815185547,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 341
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 79.3125,
      "epoch": 0.016524932354078082,
      "grad_norm": 10.622523201508118,
      "learning_rate": 4.917616930807885e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 342
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 81.3125,
      "epoch": 0.016573250869733283,
      "grad_norm": 10.25497111567043,
      "learning_rate": 4.91737533822961e-07,
      "loss": 0.0,
      "reward": 1.40625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 343
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.75,
      "epoch": 0.01662156938538848,
      "grad_norm": 0.0,
      "learning_rate": 4.917133745651333e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 344
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.6875,
      "epoch": 0.01666988790104368,
      "grad_norm": 0.0,
      "learning_rate": 4.916892153073057e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 345
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 78.5625,
      "epoch": 0.01671820641669888,
      "grad_norm": 16.027205308356162,
      "learning_rate": 4.916650560494781e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.4401872754096985,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 346
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.78125,
      "epoch": 0.016766524932354077,
      "grad_norm": 0.0,
      "learning_rate": 4.916408967916505e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 347
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 80.75,
      "epoch": 0.01681484344800928,
      "grad_norm": 24.819643071089974,
      "learning_rate": 4.91616737533823e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.30935919284820557,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 348
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.34375,
      "epoch": 0.016863161963664476,
      "grad_norm": 10.236286540959787,
      "learning_rate": 4.915925782759954e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.2962459325790405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 349
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 85.28125,
      "epoch": 0.016911480479319674,
      "grad_norm": 21.891612546955955,
      "learning_rate": 4.915684190181678e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.39135709404945374,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 350
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 85.59375,
      "epoch": 0.016959798994974875,
      "grad_norm": 115.92294357886504,
      "learning_rate": 4.915442597603401e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 351
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.9375,
      "epoch": 0.017008117510630073,
      "grad_norm": 0.0,
      "learning_rate": 4.915201005025125e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 352
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.0625,
      "epoch": 0.017056436026285274,
      "grad_norm": 7.7125128115930925,
      "learning_rate": 4.914959412446849e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.22049658000469208,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 353
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 80.4375,
      "epoch": 0.01710475454194047,
      "grad_norm": 14.249244712770867,
      "learning_rate": 4.914717819868573e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.4592459797859192,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 354
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 86.875,
      "epoch": 0.01715307305759567,
      "grad_norm": 54.888435305384476,
      "learning_rate": 4.914476227290297e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 355
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.03125,
      "epoch": 0.01720139157325087,
      "grad_norm": 7.6340510729873,
      "learning_rate": 4.914234634712022e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 356
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 92.03125,
      "epoch": 0.017249710088906068,
      "grad_norm": 11.7215712065444,
      "learning_rate": 4.913993042133746e-07,
      "loss": -0.0,
      "reward": 1.5,
      "reward_std": 0.28104937076568604,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 357
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 87.28125,
      "epoch": 0.01729802860456127,
      "grad_norm": 14.316055738907272,
      "learning_rate": 4.91375144955547e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.31795984506607056,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 358
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.75,
      "epoch": 0.017346347120216467,
      "grad_norm": 6.652828522179249,
      "learning_rate": 4.913509856977193e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 359
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.625,
      "epoch": 0.017394665635871664,
      "grad_norm": 0.0,
      "learning_rate": 4.913268264398918e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 360
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.5,
      "epoch": 0.017442984151526866,
      "grad_norm": 0.0,
      "learning_rate": 4.913026671820641e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 361
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.5,
      "epoch": 0.017491302667182063,
      "grad_norm": 9.825558447066523,
      "learning_rate": 4.912785079242365e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 362
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.65625,
      "epoch": 0.017539621182837264,
      "grad_norm": 0.0,
      "learning_rate": 4.912543486664089e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 363
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.25,
      "epoch": 0.017587939698492462,
      "grad_norm": 6.143596270385686,
      "learning_rate": 4.912301894085814e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.23421898484230042,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 364
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 76.625,
      "epoch": 0.01763625821414766,
      "grad_norm": 11.310489414963142,
      "learning_rate": 4.912060301507538e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.32346823811531067,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 365
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 80.9375,
      "epoch": 0.01768457672980286,
      "grad_norm": 8.229594918415339,
      "learning_rate": 4.911818708929261e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 366
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.46875,
      "epoch": 0.01773289524545806,
      "grad_norm": 19.533129641900626,
      "learning_rate": 4.911577116350986e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.2799758315086365,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 367
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 68.46875,
      "epoch": 0.01778121376111326,
      "grad_norm": 5.544613491799219,
      "learning_rate": 4.911335523772709e-07,
      "loss": -0.0,
      "reward": 2.40625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 368
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 92.25,
      "epoch": 0.017829532276768457,
      "grad_norm": 14.927339746621724,
      "learning_rate": 4.911093931194433e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.399166464805603,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 369
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 76.65625,
      "epoch": 0.017877850792423655,
      "grad_norm": 24.214180330303336,
      "learning_rate": 4.910852338616157e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.4697399437427521,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 370
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 91.8125,
      "epoch": 0.017926169308078856,
      "grad_norm": 42.589408045005904,
      "learning_rate": 4.910610746037881e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.3230288028717041,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 371
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 75.4375,
      "epoch": 0.017974487823734054,
      "grad_norm": 22.042912460257035,
      "learning_rate": 4.910369153459606e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.3898441195487976,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 372
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 82.4375,
      "epoch": 0.018022806339389255,
      "grad_norm": 11.591195712493045,
      "learning_rate": 4.91012756088133e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.28104937076568604,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 373
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 89.6875,
      "epoch": 0.018071124855044453,
      "grad_norm": 8.43247630207679,
      "learning_rate": 4.909885968303054e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 374
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 91.25,
      "epoch": 0.01811944337069965,
      "grad_norm": 11.955116588371105,
      "learning_rate": 4.909644375724777e-07,
      "loss": -0.0,
      "reward": 1.59375,
      "reward_std": 0.34254783391952515,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 375
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.90625,
      "epoch": 0.01816776188635485,
      "grad_norm": 8.145479177681235,
      "learning_rate": 4.909402783146501e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 376
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 83.46875,
      "epoch": 0.01821608040201005,
      "grad_norm": 15.55332801634372,
      "learning_rate": 4.909161190568225e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.4019385576248169,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 377
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 94.21875,
      "epoch": 0.01826439891766525,
      "grad_norm": 11.113689407610597,
      "learning_rate": 4.908919597989949e-07,
      "loss": -0.0,
      "reward": 2.078125,
      "reward_std": 0.4136318862438202,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 378
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 90.65625,
      "epoch": 0.018312717433320448,
      "grad_norm": 13.356045240113064,
      "learning_rate": 4.908678005411674e-07,
      "loss": -0.0,
      "reward": 1.25,
      "reward_std": 0.45782607793807983,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 379
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 90.9375,
      "epoch": 0.018361035948975646,
      "grad_norm": 9.361273392001493,
      "learning_rate": 4.908436412833398e-07,
      "loss": -0.0,
      "reward": 1.453125,
      "reward_std": 0.37298911809921265,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 380
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 94.0625,
      "epoch": 0.018409354464630847,
      "grad_norm": 27.352071500500706,
      "learning_rate": 4.908194820255122e-07,
      "loss": -0.0,
      "reward": 1.359375,
      "reward_std": 0.29747018218040466,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 381
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 74.625,
      "epoch": 0.018457672980286045,
      "grad_norm": 19.770957385891332,
      "learning_rate": 4.907953227676845e-07,
      "loss": -0.0,
      "reward": 1.75,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 382
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 84.3125,
      "epoch": 0.018505991495941246,
      "grad_norm": 434.44732997537784,
      "learning_rate": 4.90771163509857e-07,
      "loss": -0.0,
      "reward": 1.078125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.0625,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 383
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.5,
      "epoch": 0.018554310011596443,
      "grad_norm": 0.0,
      "learning_rate": 4.907470042520293e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2561880052089691,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 384
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 89.25,
      "epoch": 0.01860262852725164,
      "grad_norm": 7.118263897500714,
      "learning_rate": 4.907228449942017e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.3319803476333618,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 385
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.40625,
      "epoch": 0.018650947042906842,
      "grad_norm": 0.0,
      "learning_rate": 4.906986857363741e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 386
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.09375,
      "epoch": 0.01869926555856204,
      "grad_norm": 8.195524042981212,
      "learning_rate": 4.906745264785466e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 387
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 87.84375,
      "epoch": 0.01874758407421724,
      "grad_norm": 11.458932464098542,
      "learning_rate": 4.90650367220719e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.4387563169002533,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 388
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 90.21875,
      "epoch": 0.01879590258987244,
      "grad_norm": 14.579241468433079,
      "learning_rate": 4.906262079628914e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 389
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 103.15625,
      "epoch": 0.018844221105527637,
      "grad_norm": 13.630851170333674,
      "learning_rate": 4.906020487050638e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.4592459797859192,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 390
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 88.5625,
      "epoch": 0.018892539621182838,
      "grad_norm": 10.233934422605776,
      "learning_rate": 4.905778894472362e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.39347878098487854,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 391
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 91.65625,
      "epoch": 0.018940858136838035,
      "grad_norm": 14.485826976752943,
      "learning_rate": 4.905537301894085e-07,
      "loss": 0.0,
      "reward": 1.265625,
      "reward_std": 0.48295050859451294,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 392
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 89.65625,
      "epoch": 0.018989176652493236,
      "grad_norm": 17.649824799966062,
      "learning_rate": 4.905295709315809e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.563550591468811,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 393
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.09375,
      "epoch": 0.019037495168148434,
      "grad_norm": 0.0,
      "learning_rate": 4.905054116737533e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 394
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.78125,
      "epoch": 0.019085813683803632,
      "grad_norm": 0.0,
      "learning_rate": 4.904812524159258e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 395
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 86.5625,
      "epoch": 0.019134132199458833,
      "grad_norm": 3.439427982328558,
      "learning_rate": 4.904570931580982e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 396
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.40625,
      "epoch": 0.01918245071511403,
      "grad_norm": 16.945922175863483,
      "learning_rate": 4.904329339002706e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.3298332095146179,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 397
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 91.0625,
      "epoch": 0.019230769230769232,
      "grad_norm": 8.77978275140754,
      "learning_rate": 4.90408774642443e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.3761745095252991,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 398
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 89.4375,
      "epoch": 0.01927908774642443,
      "grad_norm": 9.194897567679684,
      "learning_rate": 4.903846153846153e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 399
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.0,
      "epoch": 0.019327406262079627,
      "grad_norm": 0.0,
      "learning_rate": 4.903604561267878e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 400
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 84.4375,
      "epoch": 0.01937572477773483,
      "grad_norm": 11.842780692345325,
      "learning_rate": 4.903362968689601e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.43236586451530457,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 401
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 92.46875,
      "epoch": 0.019424043293390026,
      "grad_norm": 8.858385986811683,
      "learning_rate": 4.903121376111325e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.39135709404945374,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 402
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 92.09375,
      "epoch": 0.019472361809045227,
      "grad_norm": 13.017394606185686,
      "learning_rate": 4.90287978353305e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.36403754353523254,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 403
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 95.3125,
      "epoch": 0.019520680324700425,
      "grad_norm": 27.013567581872646,
      "learning_rate": 4.902638190954774e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.434487521648407,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 404
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 85.09375,
      "epoch": 0.019568998840355626,
      "grad_norm": 15.369814346160622,
      "learning_rate": 4.902396598376498e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 405
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.5625,
      "epoch": 0.019617317356010824,
      "grad_norm": 6.218792579077898,
      "learning_rate": 4.902155005798221e-07,
      "loss": -0.0,
      "reward": 2.390625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 406
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.21875,
      "epoch": 0.01966563587166602,
      "grad_norm": 6.873563373676089,
      "learning_rate": 4.901913413219946e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.24067938327789307,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 407
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 85.09375,
      "epoch": 0.019713954387321223,
      "grad_norm": 10.02987766647073,
      "learning_rate": 4.901671820641669e-07,
      "loss": 0.0,
      "reward": 1.234375,
      "reward_std": 0.15580293536186218,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 408
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 85.0625,
      "epoch": 0.01976227290297642,
      "grad_norm": 20.176976712474488,
      "learning_rate": 4.901430228063393e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.5013028383255005,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 409
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 80.75,
      "epoch": 0.01981059141863162,
      "grad_norm": 7.130153850330547,
      "learning_rate": 4.901188635485118e-07,
      "loss": -0.0,
      "reward": 2.21875,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 410
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 85.28125,
      "epoch": 0.01985890993428682,
      "grad_norm": 15.916912949164159,
      "learning_rate": 4.900947042906842e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.5313355922698975,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 411
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 91.5,
      "epoch": 0.019907228449942017,
      "grad_norm": 13.799271752033457,
      "learning_rate": 4.900705450328566e-07,
      "loss": -0.0,
      "reward": 1.90625,
      "reward_std": 0.29305070638656616,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 412
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.28125,
      "epoch": 0.019955546965597218,
      "grad_norm": 14.924750943890606,
      "learning_rate": 4.90046385775029e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.3697938621044159,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 413
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 84.375,
      "epoch": 0.020003865481252416,
      "grad_norm": 13.421274956116848,
      "learning_rate": 4.900222265172014e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.3676722049713135,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 414
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 82.3125,
      "epoch": 0.020052183996907617,
      "grad_norm": 89.25565971232406,
      "learning_rate": 4.899980672593737e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.4355512857437134,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 415
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.9375,
      "epoch": 0.020100502512562814,
      "grad_norm": 7.318020636414703,
      "learning_rate": 4.899739080015461e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 416
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.875,
      "epoch": 0.020148821028218012,
      "grad_norm": 0.0,
      "learning_rate": 4.899497487437185e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 417
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 81.15625,
      "epoch": 0.020197139543873213,
      "grad_norm": 25.50035647056387,
      "learning_rate": 4.89925589485891e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.285190224647522,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 418
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 70.375,
      "epoch": 0.02024545805952841,
      "grad_norm": 20.22630999353325,
      "learning_rate": 4.899014302280634e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.3277115821838379,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 419
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 68.75,
      "epoch": 0.020293776575183612,
      "grad_norm": 14.583931182680555,
      "learning_rate": 4.898772709702358e-07,
      "loss": -0.0,
      "reward": 2.3125,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 420
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 72.25,
      "epoch": 0.02034209509083881,
      "grad_norm": 17.591615967373517,
      "learning_rate": 4.898531117124082e-07,
      "loss": -0.0,
      "reward": 1.484375,
      "reward_std": 0.15580293536186218,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 421
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.84375,
      "epoch": 0.020390413606494007,
      "grad_norm": 15.671478893750194,
      "learning_rate": 4.898289524545806e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 422
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 75.59375,
      "epoch": 0.02043873212214921,
      "grad_norm": 20.759312907365643,
      "learning_rate": 4.89804793196753e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 423
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.4375,
      "epoch": 0.020487050637804406,
      "grad_norm": 0.0,
      "learning_rate": 4.897806339389253e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 424
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.3125,
      "epoch": 0.020535369153459607,
      "grad_norm": 13.838148500936956,
      "learning_rate": 4.897564746810977e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 425
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 68.375,
      "epoch": 0.020583687669114805,
      "grad_norm": 5.034148922345902,
      "learning_rate": 4.897323154232702e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.23960642516613007,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 426
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 73.53125,
      "epoch": 0.020632006184770003,
      "grad_norm": 9.363826164909815,
      "learning_rate": 4.897081561654426e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 427
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 66.8125,
      "epoch": 0.020680324700425204,
      "grad_norm": 12.41835451730632,
      "learning_rate": 4.89683996907615e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.30856335163116455,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 428
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 66.875,
      "epoch": 0.0207286432160804,
      "grad_norm": 8.012470858130866,
      "learning_rate": 4.896598376497874e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 429
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 76.15625,
      "epoch": 0.020776961731735603,
      "grad_norm": 16.63618407136281,
      "learning_rate": 4.896356783919598e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.3503679633140564,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 430
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 78.625,
      "epoch": 0.0208252802473908,
      "grad_norm": 36.53290325836137,
      "learning_rate": 4.896115191341322e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 431
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.15625,
      "epoch": 0.020873598763045998,
      "grad_norm": 30.85649289739766,
      "learning_rate": 4.895873598763045e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 432
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.09375,
      "epoch": 0.0209219172787012,
      "grad_norm": 11.092669017233156,
      "learning_rate": 4.895632006184771e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 433
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.25,
      "epoch": 0.020970235794356397,
      "grad_norm": 0.0,
      "learning_rate": 4.895390413606494e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 434
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.5,
      "epoch": 0.021018554310011598,
      "grad_norm": 0.0,
      "learning_rate": 4.895148821028218e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 435
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 81.59375,
      "epoch": 0.021066872825666796,
      "grad_norm": 12.282628289790079,
      "learning_rate": 4.894907228449942e-07,
      "loss": -0.0,
      "reward": 1.453125,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 436
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.6875,
      "epoch": 0.021115191341321993,
      "grad_norm": 0.0,
      "learning_rate": 4.894665635871666e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 437
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 69.0625,
      "epoch": 0.021163509856977195,
      "grad_norm": 23.01488434840663,
      "learning_rate": 4.89442404329339e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.3608423173427582,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 438
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 80.625,
      "epoch": 0.021211828372632392,
      "grad_norm": 12.995886939182096,
      "learning_rate": 4.894182450715113e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 439
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 66.9375,
      "epoch": 0.021260146888287593,
      "grad_norm": 8.524235351146608,
      "learning_rate": 4.893940858136838e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 440
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 70.21875,
      "epoch": 0.02130846540394279,
      "grad_norm": 11.069491895327308,
      "learning_rate": 4.893699265558562e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 441
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.09375,
      "epoch": 0.02135678391959799,
      "grad_norm": 0.0,
      "learning_rate": 4.893457672980286e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 442
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 75.375,
      "epoch": 0.02140510243525319,
      "grad_norm": 16.9115435056491,
      "learning_rate": 4.89321608040201e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.3998437523841858,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 443
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 69.90625,
      "epoch": 0.021453420950908388,
      "grad_norm": 9.099024107177849,
      "learning_rate": 4.892974487823734e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 444
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.4375,
      "epoch": 0.02150173946656359,
      "grad_norm": 32.92913619091926,
      "learning_rate": 4.892732895245458e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.3061639666557312,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 445
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.03125,
      "epoch": 0.021550057982218786,
      "grad_norm": 0.0,
      "learning_rate": 4.892491302667181e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 446
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 70.40625,
      "epoch": 0.021598376497873984,
      "grad_norm": 8.850412964228424,
      "learning_rate": 4.892249710088906e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 447
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.75,
      "epoch": 0.021646695013529185,
      "grad_norm": 5.960943194083108,
      "learning_rate": 4.892008117510629e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 448
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.40625,
      "epoch": 0.021695013529184383,
      "grad_norm": 0.0,
      "learning_rate": 4.891766524932354e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 449
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 70.53125,
      "epoch": 0.021743332044839584,
      "grad_norm": 51.20168758238621,
      "learning_rate": 4.891524932354078e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 450
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 67.625,
      "epoch": 0.021791650560494782,
      "grad_norm": 9.729841611976386,
      "learning_rate": 4.891283339775802e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 451
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.6875,
      "epoch": 0.02183996907614998,
      "grad_norm": 0.0,
      "learning_rate": 4.891041747197526e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 452
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.4375,
      "epoch": 0.02188828759180518,
      "grad_norm": 9.264172198421486,
      "learning_rate": 4.89080015461925e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.24280104041099548,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 453
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 68.59375,
      "epoch": 0.021936606107460378,
      "grad_norm": 11.99059328944805,
      "learning_rate": 4.890558562040974e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.4114847779273987,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 454
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 64.78125,
      "epoch": 0.02198492462311558,
      "grad_norm": 13.162676552068767,
      "learning_rate": 4.890316969462697e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 455
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.90625,
      "epoch": 0.022033243138770777,
      "grad_norm": 8.976868250632428,
      "learning_rate": 4.890075376884421e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 456
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 82.21875,
      "epoch": 0.022081561654425975,
      "grad_norm": 21.181030949766477,
      "learning_rate": 4.889833784306146e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.41560810804367065,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 457
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 73.59375,
      "epoch": 0.022129880170081176,
      "grad_norm": 22.880371513154756,
      "learning_rate": 4.88959219172787e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 458
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 71.875,
      "epoch": 0.022178198685736374,
      "grad_norm": 8.20865061344611,
      "learning_rate": 4.889350599149594e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.37298911809921265,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 459
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 70.6875,
      "epoch": 0.022226517201391575,
      "grad_norm": 5.192636980257663,
      "learning_rate": 4.889109006571318e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 460
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.71875,
      "epoch": 0.022274835717046772,
      "grad_norm": 12.4782229212193,
      "learning_rate": 4.888867413993042e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.39135709404945374,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 461
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 78.84375,
      "epoch": 0.02232315423270197,
      "grad_norm": 7.705437244499536,
      "learning_rate": 4.888625821414766e-07,
      "loss": -0.0,
      "reward": 1.375,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 462
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.90625,
      "epoch": 0.02237147274835717,
      "grad_norm": 6.714480490018273,
      "learning_rate": 4.88838422883649e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 463
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 72.65625,
      "epoch": 0.02241979126401237,
      "grad_norm": 22.80325009642474,
      "learning_rate": 4.888142636258215e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.34305399656295776,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 464
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 72.9375,
      "epoch": 0.02246810977966757,
      "grad_norm": 0.0,
      "learning_rate": 4.887901043679938e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 465
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 80.78125,
      "epoch": 0.022516428295322768,
      "grad_norm": 9.622241360802917,
      "learning_rate": 4.887659451101662e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.36306124925613403,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 466
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 77.40625,
      "epoch": 0.022564746810977965,
      "grad_norm": 30.474042156271274,
      "learning_rate": 4.887417858523386e-07,
      "loss": -0.0,
      "reward": 1.578125,
      "reward_std": 0.4295162558555603,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 467
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 73.0625,
      "epoch": 0.022613065326633167,
      "grad_norm": 28.794101968702552,
      "learning_rate": 4.88717626594511e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.36278265714645386,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 468
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.8125,
      "epoch": 0.022661383842288364,
      "grad_norm": 0.0,
      "learning_rate": 4.886934673366834e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 469
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.1875,
      "epoch": 0.022709702357943565,
      "grad_norm": 0.0,
      "learning_rate": 4.886693080788558e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 470
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 75.34375,
      "epoch": 0.022758020873598763,
      "grad_norm": 8.713341762698654,
      "learning_rate": 4.886451488210282e-07,
      "loss": -0.0,
      "reward": 1.09375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.0625,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 471
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.1875,
      "epoch": 0.02280633938925396,
      "grad_norm": 0.0,
      "learning_rate": 4.886209895632005e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 472
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 67.5625,
      "epoch": 0.022854657904909162,
      "grad_norm": 14.05914691554478,
      "learning_rate": 4.885968303053731e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.22835706174373627,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 473
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.21875,
      "epoch": 0.02290297642056436,
      "grad_norm": 0.0,
      "learning_rate": 4.885726710475454e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 474
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 68.0,
      "epoch": 0.02295129493621956,
      "grad_norm": 0.0,
      "learning_rate": 4.885485117897178e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 475
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 77.625,
      "epoch": 0.02299961345187476,
      "grad_norm": 13.321842869421719,
      "learning_rate": 4.885243525318902e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.32878512144088745,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 476
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.15625,
      "epoch": 0.023047931967529956,
      "grad_norm": 0.0,
      "learning_rate": 4.885001932740626e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 477
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.625,
      "epoch": 0.023096250483185157,
      "grad_norm": 9.860316951625006,
      "learning_rate": 4.88476034016235e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 478
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 77.5625,
      "epoch": 0.023144568998840355,
      "grad_norm": 13.70045565173991,
      "learning_rate": 4.884518747584073e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.32666346430778503,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 479
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 75.25,
      "epoch": 0.023192887514495556,
      "grad_norm": 17.153223455957168,
      "learning_rate": 4.884277155005799e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.3676722049713135,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 480
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 76.40625,
      "epoch": 0.023241206030150754,
      "grad_norm": 13.621776871209752,
      "learning_rate": 4.884035562427522e-07,
      "loss": -0.0,
      "reward": 1.40625,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 481
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 60.3125,
      "epoch": 0.02328952454580595,
      "grad_norm": 10.869767949768068,
      "learning_rate": 4.883793969849246e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.31068500876426697,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 482
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.03125,
      "epoch": 0.023337843061461153,
      "grad_norm": 0.0,
      "learning_rate": 4.88355237727097e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 483
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.34375,
      "epoch": 0.02338616157711635,
      "grad_norm": 0.0,
      "learning_rate": 4.883310784692694e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 484
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 74.4375,
      "epoch": 0.02343448009277155,
      "grad_norm": 14.1366570941377,
      "learning_rate": 4.883069192114418e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.4808131456375122,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 485
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 70.34375,
      "epoch": 0.02348279860842675,
      "grad_norm": 7.616077200608903,
      "learning_rate": 4.882827599536141e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.15580293536186218,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 486
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.1875,
      "epoch": 0.023531117124081947,
      "grad_norm": 0.0,
      "learning_rate": 4.882586006957866e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 487
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 69.875,
      "epoch": 0.023579435639737148,
      "grad_norm": 7.0909547902536705,
      "learning_rate": 4.88234441437959e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 488
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 74.25,
      "epoch": 0.023627754155392346,
      "grad_norm": 6.130440102928585,
      "learning_rate": 4.882102821801314e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 489
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.375,
      "epoch": 0.023676072671047547,
      "grad_norm": 7.955557451709033,
      "learning_rate": 4.881861229223038e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.31663835048675537,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 490
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 69.65625,
      "epoch": 0.023724391186702744,
      "grad_norm": 7.066195392439719,
      "learning_rate": 4.881619636644762e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 491
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.03125,
      "epoch": 0.023772709702357942,
      "grad_norm": 14.328376827622144,
      "learning_rate": 4.881378044066486e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.2553878426551819,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 492
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 71.71875,
      "epoch": 0.023821028218013143,
      "grad_norm": 10.054483313112117,
      "learning_rate": 4.88113645148821e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 493
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.96875,
      "epoch": 0.02386934673366834,
      "grad_norm": 0.0,
      "learning_rate": 4.880894858909934e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 494
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 75.03125,
      "epoch": 0.023917665249323542,
      "grad_norm": 11.232413321494532,
      "learning_rate": 4.880653266331657e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.15580293536186218,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 495
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 67.9375,
      "epoch": 0.02396598376497874,
      "grad_norm": 12.612876020758904,
      "learning_rate": 4.880411673753382e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 496
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 82.65625,
      "epoch": 0.024014302280633937,
      "grad_norm": 28.349032102605644,
      "learning_rate": 4.880170081175106e-07,
      "loss": 0.0,
      "reward": 1.484375,
      "reward_std": 0.49453338980674744,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 497
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 70.46875,
      "epoch": 0.02406262079628914,
      "grad_norm": 13.817913480370045,
      "learning_rate": 4.87992848859683e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.3006648123264313,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 498
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 80.78125,
      "epoch": 0.024110939311944336,
      "grad_norm": 13.128150017387576,
      "learning_rate": 4.879686896018554e-07,
      "loss": -0.0,
      "reward": 1.453125,
      "reward_std": 0.3457430303096771,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 499
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.625,
      "epoch": 0.024159257827599537,
      "grad_norm": 0.0,
      "learning_rate": 4.879445303440278e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 500
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.71875,
      "epoch": 0.024207576343254735,
      "grad_norm": 0.0,
      "learning_rate": 4.879203710862002e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 501
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.84375,
      "epoch": 0.024255894858909933,
      "grad_norm": 0.0,
      "learning_rate": 4.878962118283726e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 502
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 66.59375,
      "epoch": 0.024304213374565134,
      "grad_norm": 9.814961596677476,
      "learning_rate": 4.878720525705451e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.37298908829689026,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 503
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.625,
      "epoch": 0.02435253189022033,
      "grad_norm": 10.649769357649667,
      "learning_rate": 4.878478933127175e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 504
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.625,
      "epoch": 0.024400850405875533,
      "grad_norm": 0.0,
      "learning_rate": 4.878237340548898e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 505
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 66.28125,
      "epoch": 0.02444916892153073,
      "grad_norm": 5.47706507441087,
      "learning_rate": 4.877995747970622e-07,
      "loss": -0.0,
      "reward": 2.359375,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 506
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.3125,
      "epoch": 0.024497487437185928,
      "grad_norm": 0.0,
      "learning_rate": 4.877754155392346e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 507
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 65.28125,
      "epoch": 0.02454580595284113,
      "grad_norm": 10.204198021470699,
      "learning_rate": 4.87751256281407e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 508
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.03125,
      "epoch": 0.024594124468496327,
      "grad_norm": 0.0,
      "learning_rate": 4.877270970235794e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 509
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 64.84375,
      "epoch": 0.024642442984151528,
      "grad_norm": 12.27448696875897,
      "learning_rate": 4.877029377657518e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 510
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 79.5625,
      "epoch": 0.024690761499806726,
      "grad_norm": 23.03521166496352,
      "learning_rate": 4.876787785079243e-07,
      "loss": -0.0,
      "reward": 1.40625,
      "reward_std": 0.5126694440841675,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 511
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.3125,
      "epoch": 0.024739080015461924,
      "grad_norm": 0.0,
      "learning_rate": 4.876546192500966e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 512
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 59.34375,
      "epoch": 0.024787398531117125,
      "grad_norm": 8.979111147658259,
      "learning_rate": 4.876304599922691e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.2873118817806244,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 513
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 63.46875,
      "epoch": 0.024835717046772322,
      "grad_norm": 5.3028667089158334,
      "learning_rate": 4.876063007344414e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 514
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.6875,
      "epoch": 0.024884035562427523,
      "grad_norm": 0.0,
      "learning_rate": 4.875821414766138e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 515
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 72.3125,
      "epoch": 0.02493235407808272,
      "grad_norm": 31.134820705323644,
      "learning_rate": 4.875579822187862e-07,
      "loss": 0.0,
      "reward": 1.546875,
      "reward_std": 0.5066353678703308,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 516
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.78125,
      "epoch": 0.02498067259373792,
      "grad_norm": 0.0,
      "learning_rate": 4.875338229609586e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 517
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 70.03125,
      "epoch": 0.02502899110939312,
      "grad_norm": 11.597517091837508,
      "learning_rate": 4.87509663703131e-07,
      "loss": -0.0,
      "reward": 2.21875,
      "reward_std": 0.3107999563217163,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 518
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.6875,
      "epoch": 0.025077309625048318,
      "grad_norm": 8.844749014311075,
      "learning_rate": 4.874855044453034e-07,
      "loss": -0.0,
      "reward": 2.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 519
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 66.90625,
      "epoch": 0.02512562814070352,
      "grad_norm": 8.468241240768558,
      "learning_rate": 4.874613451874759e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.20633062720298767,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 520
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 77.34375,
      "epoch": 0.025173946656358717,
      "grad_norm": 11.418311348565982,
      "learning_rate": 4.874371859296482e-07,
      "loss": -0.0,
      "reward": 2.09375,
      "reward_std": 0.39408785104751587,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 521
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 80.9375,
      "epoch": 0.025222265172013914,
      "grad_norm": 19.51890273467776,
      "learning_rate": 4.874130266718206e-07,
      "loss": -0.0,
      "reward": 2.21875,
      "reward_std": 0.46561095118522644,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 522
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 83.75,
      "epoch": 0.025270583687669115,
      "grad_norm": 23.42064154632523,
      "learning_rate": 4.87388867413993e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.5110037326812744,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 523
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.5,
      "epoch": 0.025318902203324313,
      "grad_norm": 15.53338931908874,
      "learning_rate": 4.873647081561654e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.2989126443862915,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 524
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 72.875,
      "epoch": 0.025367220718979514,
      "grad_norm": 17.428506778575116,
      "learning_rate": 4.873405488983378e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 525
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.4375,
      "epoch": 0.025415539234634712,
      "grad_norm": 0.0,
      "learning_rate": 4.873163896405101e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 526
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 74.59375,
      "epoch": 0.02546385775028991,
      "grad_norm": 17.316353018692585,
      "learning_rate": 4.872922303826827e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.3913571238517761,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 527
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 69.96875,
      "epoch": 0.02551217626594511,
      "grad_norm": 6.009529455928404,
      "learning_rate": 4.87268071124855e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 528
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 81.84375,
      "epoch": 0.02556049478160031,
      "grad_norm": 15.710083562680584,
      "learning_rate": 4.872439118670274e-07,
      "loss": -0.0,
      "reward": 1.484375,
      "reward_std": 0.3761745095252991,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 529
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.5625,
      "epoch": 0.02560881329725551,
      "grad_norm": 182.28192933920656,
      "learning_rate": 4.872197526091998e-07,
      "loss": 0.0,
      "reward": 2.1875,
      "reward_std": 0.28566449880599976,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 530
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.53125,
      "epoch": 0.025657131812910707,
      "grad_norm": 8.820942384017993,
      "learning_rate": 4.871955933513722e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.21528217196464539,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 531
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 71.4375,
      "epoch": 0.025705450328565905,
      "grad_norm": 20.78581895769837,
      "learning_rate": 4.871714340935446e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.3924052119255066,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 532
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 84.90625,
      "epoch": 0.025753768844221106,
      "grad_norm": 15.480363292686828,
      "learning_rate": 4.87147274835717e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.38452720642089844,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 533
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 80.5,
      "epoch": 0.025802087359876304,
      "grad_norm": 35.15365628314348,
      "learning_rate": 4.871231155778895e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.482940673828125,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 534
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 65.96875,
      "epoch": 0.025850405875531505,
      "grad_norm": 8.71723493388719,
      "learning_rate": 4.870989563200619e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2651395797729492,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 535
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.53125,
      "epoch": 0.025898724391186703,
      "grad_norm": 4.648114010891773,
      "learning_rate": 4.870747970622342e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 536
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.125,
      "epoch": 0.0259470429068419,
      "grad_norm": 19.17050476490381,
      "learning_rate": 4.870506378044066e-07,
      "loss": -0.0,
      "reward": 2.21875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 537
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.90625,
      "epoch": 0.0259953614224971,
      "grad_norm": 5.258547639546589,
      "learning_rate": 4.87026478546579e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 538
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 86.78125,
      "epoch": 0.0260436799381523,
      "grad_norm": 23.79795397653507,
      "learning_rate": 4.870023192887514e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.4411253333091736,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 539
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 78.5625,
      "epoch": 0.0260919984538075,
      "grad_norm": 15.848306931029354,
      "learning_rate": 4.869781600309238e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.41187620162963867,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 540
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.75,
      "epoch": 0.026140316969462698,
      "grad_norm": 7.8958877519243345,
      "learning_rate": 4.869540007730962e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 541
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 80.15625,
      "epoch": 0.026188635485117896,
      "grad_norm": 9.597526547182666,
      "learning_rate": 4.869298415152687e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 542
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 87.03125,
      "epoch": 0.026236954000773097,
      "grad_norm": 22.58840091518499,
      "learning_rate": 4.869056822574411e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.4419417083263397,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 543
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 84.09375,
      "epoch": 0.026285272516428294,
      "grad_norm": 14.96519422585085,
      "learning_rate": 4.868815229996135e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.5249877572059631,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 544
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.6875,
      "epoch": 0.026333591032083496,
      "grad_norm": 0.0,
      "learning_rate": 4.868573637417858e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 545
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 85.4375,
      "epoch": 0.026381909547738693,
      "grad_norm": 8.235009682133901,
      "learning_rate": 4.868332044839582e-07,
      "loss": 0.0,
      "reward": 1.40625,
      "reward_std": 0.21316051483154297,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 546
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.28125,
      "epoch": 0.02643022806339389,
      "grad_norm": 0.0,
      "learning_rate": 4.868090452261306e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 547
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 76.875,
      "epoch": 0.026478546579049092,
      "grad_norm": 8.705253118099312,
      "learning_rate": 4.86784885968303e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 548
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 65.25,
      "epoch": 0.02652686509470429,
      "grad_norm": 9.513764902428557,
      "learning_rate": 4.867607267104754e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 549
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 81.25,
      "epoch": 0.02657518361035949,
      "grad_norm": 17.079076561843625,
      "learning_rate": 4.867365674526479e-07,
      "loss": 0.0,
      "reward": 1.390625,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 550
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.46875,
      "epoch": 0.02662350212601469,
      "grad_norm": 0.0,
      "learning_rate": 4.867124081948203e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 551
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 86.5,
      "epoch": 0.026671820641669886,
      "grad_norm": 28.81409650736757,
      "learning_rate": 4.866882489369926e-07,
      "loss": -0.0,
      "reward": 1.90625,
      "reward_std": 0.5234746932983398,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 552
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 71.875,
      "epoch": 0.026720139157325087,
      "grad_norm": 14.75879058674214,
      "learning_rate": 4.866640896791651e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 553
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.90625,
      "epoch": 0.026768457672980285,
      "grad_norm": 0.0,
      "learning_rate": 4.866399304213374e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 554
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 86.0,
      "epoch": 0.026816776188635486,
      "grad_norm": 11.047819193283896,
      "learning_rate": 4.866157711635098e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 555
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 75.1875,
      "epoch": 0.026865094704290684,
      "grad_norm": 11.045756314836517,
      "learning_rate": 4.865916119056822e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.3977220952510834,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 556
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 75.5625,
      "epoch": 0.02691341321994588,
      "grad_norm": 26.962997461084925,
      "learning_rate": 4.865674526478546e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.3977220952510834,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 557
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.8125,
      "epoch": 0.026961731735601083,
      "grad_norm": 0.0,
      "learning_rate": 4.865432933900271e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 558
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.71875,
      "epoch": 0.02701005025125628,
      "grad_norm": 9.842308760907946,
      "learning_rate": 4.865191341321994e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.37086743116378784,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 559
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 86.46875,
      "epoch": 0.02705836876691148,
      "grad_norm": 6.574217658581166,
      "learning_rate": 4.864949748743719e-07,
      "loss": -0.0,
      "reward": 2.21875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 560
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.71875,
      "epoch": 0.02710668728256668,
      "grad_norm": 0.0,
      "learning_rate": 4.864708156165442e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 561
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 77.625,
      "epoch": 0.027155005798221877,
      "grad_norm": 12.26366611233329,
      "learning_rate": 4.864466563587166e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.45561128854751587,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 562
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.875,
      "epoch": 0.027203324313877078,
      "grad_norm": 14.748930614872933,
      "learning_rate": 4.86422497100889e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.3029785752296448,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 563
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 79.5625,
      "epoch": 0.027251642829532276,
      "grad_norm": 12.666866928732986,
      "learning_rate": 4.863983378430614e-07,
      "loss": -0.0,
      "reward": 2.078125,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 564
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 88.4375,
      "epoch": 0.027299961345187477,
      "grad_norm": 15.743321740635636,
      "learning_rate": 4.863741785852339e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.4376729428768158,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 565
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 79.09375,
      "epoch": 0.027348279860842675,
      "grad_norm": 10.195641359256372,
      "learning_rate": 4.863500193274063e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 566
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 82.25,
      "epoch": 0.027396598376497876,
      "grad_norm": 10.907727490474926,
      "learning_rate": 4.863258600695787e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.36162739992141724,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 567
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.15625,
      "epoch": 0.027444916892153073,
      "grad_norm": 0.0,
      "learning_rate": 4.86301700811751e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 568
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 79.15625,
      "epoch": 0.02749323540780827,
      "grad_norm": 15.413787127124014,
      "learning_rate": 4.862775415539234e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.3319803476333618,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 569
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.625,
      "epoch": 0.027541553923463472,
      "grad_norm": 0.0,
      "learning_rate": 4.862533822960958e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 570
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 75.71875,
      "epoch": 0.02758987243911867,
      "grad_norm": 10.5823978952262,
      "learning_rate": 4.862292230382682e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.32666346430778503,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 571
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.84375,
      "epoch": 0.02763819095477387,
      "grad_norm": 9.491438730117283,
      "learning_rate": 4.862050637804406e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 572
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.15625,
      "epoch": 0.02768650947042907,
      "grad_norm": 17.98488644847277,
      "learning_rate": 4.861809045226131e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 573
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 95.28125,
      "epoch": 0.027734827986084266,
      "grad_norm": 7.143915930872392,
      "learning_rate": 4.861567452647855e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 574
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.625,
      "epoch": 0.027783146501739468,
      "grad_norm": 15.506202836877268,
      "learning_rate": 4.861325860069579e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.3225545287132263,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 575
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.09375,
      "epoch": 0.027831465017394665,
      "grad_norm": 0.0,
      "learning_rate": 4.861084267491302e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 576
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.34375,
      "epoch": 0.027879783533049866,
      "grad_norm": 0.0,
      "learning_rate": 4.860842674913027e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 577
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 85.15625,
      "epoch": 0.027928102048705064,
      "grad_norm": 23.29879977709132,
      "learning_rate": 4.86060108233475e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.5465803742408752,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 578
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.375,
      "epoch": 0.027976420564360262,
      "grad_norm": 0.0,
      "learning_rate": 4.860359489756474e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.2651395797729492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 579
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 79.03125,
      "epoch": 0.028024739080015463,
      "grad_norm": 57.75075147335275,
      "learning_rate": 4.860117897178198e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.3503679633140564,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 580
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 83.8125,
      "epoch": 0.02807305759567066,
      "grad_norm": 26.568794794523306,
      "learning_rate": 4.859876304599923e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.4408525228500366,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 581
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.75,
      "epoch": 0.028121376111325862,
      "grad_norm": 0.0,
      "learning_rate": 4.859634712021647e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 582
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.65625,
      "epoch": 0.02816969462698106,
      "grad_norm": 44.95834215412784,
      "learning_rate": 4.859393119443371e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 583
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 97.96875,
      "epoch": 0.028218013142636257,
      "grad_norm": 10.121012066341688,
      "learning_rate": 4.859151526865095e-07,
      "loss": 0.0,
      "reward": 1.484375,
      "reward_std": 0.4429898262023926,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 584
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.25,
      "epoch": 0.028266331658291458,
      "grad_norm": 10.131389694230561,
      "learning_rate": 4.858909934286818e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 585
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 94.40625,
      "epoch": 0.028314650173946656,
      "grad_norm": 58.80639393249367,
      "learning_rate": 4.858668341708542e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.5476284623146057,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 586
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 86.6875,
      "epoch": 0.028362968689601857,
      "grad_norm": 8.828339292418924,
      "learning_rate": 4.858426749130266e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 587
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 80.71875,
      "epoch": 0.028411287205257055,
      "grad_norm": 25.548684296070743,
      "learning_rate": 4.85818515655199e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.5871879458427429,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 588
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 84.0,
      "epoch": 0.028459605720912252,
      "grad_norm": 23.29936667898002,
      "learning_rate": 4.857943563973715e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.3998437523841858,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 589
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.46875,
      "epoch": 0.028507924236567454,
      "grad_norm": 3.9519306934567786,
      "learning_rate": 4.857701971395439e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 590
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 94.1875,
      "epoch": 0.02855624275222265,
      "grad_norm": 6.617977867247309,
      "learning_rate": 4.857460378817163e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 591
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 87.4375,
      "epoch": 0.028604561267877852,
      "grad_norm": 26.16308878709521,
      "learning_rate": 4.857218786238886e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 592
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 77.03125,
      "epoch": 0.02865287978353305,
      "grad_norm": 10.98802101091287,
      "learning_rate": 4.856977193660611e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.43447771668434143,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 593
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 82.4375,
      "epoch": 0.028701198299188248,
      "grad_norm": 8.435266499143792,
      "learning_rate": 4.856735601082334e-07,
      "loss": 0.0,
      "reward": 1.453125,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 594
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 88.5,
      "epoch": 0.02874951681484345,
      "grad_norm": 15.018818638443955,
      "learning_rate": 4.856494008504058e-07,
      "loss": -0.0,
      "reward": 1.328125,
      "reward_std": 0.5207444429397583,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 595
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.40625,
      "epoch": 0.028797835330498647,
      "grad_norm": 0.0,
      "learning_rate": 4.856252415925783e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 596
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.75,
      "epoch": 0.028846153846153848,
      "grad_norm": 12.29543341148788,
      "learning_rate": 4.856010823347507e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.3273632526397705,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 597
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 81.40625,
      "epoch": 0.028894472361809045,
      "grad_norm": 11.57057900874171,
      "learning_rate": 4.855769230769231e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.39774754643440247,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 598
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 83.15625,
      "epoch": 0.028942790877464243,
      "grad_norm": 13.638865445364557,
      "learning_rate": 4.855527638190954e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.31300365924835205,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 599
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.59375,
      "epoch": 0.028991109393119444,
      "grad_norm": 0.0,
      "learning_rate": 4.855286045612679e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 600
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 79.03125,
      "epoch": 0.029039427908774642,
      "grad_norm": 24.387182565631484,
      "learning_rate": 4.855044453034402e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.50025475025177,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 601
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 77.71875,
      "epoch": 0.029087746424429843,
      "grad_norm": 19.546669111113115,
      "learning_rate": 4.854802860456126e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.39347878098487854,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 602
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 77.3125,
      "epoch": 0.02913606494008504,
      "grad_norm": 8.914834101162231,
      "learning_rate": 4.85456126787785e-07,
      "loss": -0.0,
      "reward": 2.09375,
      "reward_std": 0.3230288028717041,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 603
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.59375,
      "epoch": 0.02918438345574024,
      "grad_norm": 0.0,
      "learning_rate": 4.854319675299575e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 604
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 72.625,
      "epoch": 0.02923270197139544,
      "grad_norm": 22.606481864718393,
      "learning_rate": 4.854078082721299e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.360832542181015,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 605
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 70.90625,
      "epoch": 0.029281020487050637,
      "grad_norm": 15.157736091062512,
      "learning_rate": 4.853836490143023e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.4465522766113281,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 606
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.125,
      "epoch": 0.02932933900270584,
      "grad_norm": 0.0,
      "learning_rate": 4.853594897564747e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 607
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 73.34375,
      "epoch": 0.029377657518361036,
      "grad_norm": 30.40187799840543,
      "learning_rate": 4.85335330498647e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.3995281159877777,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 608
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 70.65625,
      "epoch": 0.029425976034016234,
      "grad_norm": 17.88754685373943,
      "learning_rate": 4.853111712408194e-07,
      "loss": 0.0,
      "reward": 1.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.03125,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 609
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 69.5,
      "epoch": 0.029474294549671435,
      "grad_norm": 17.98301444406225,
      "learning_rate": 4.852870119829918e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 610
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 70.1875,
      "epoch": 0.029522613065326633,
      "grad_norm": 18.58905209012272,
      "learning_rate": 4.852628527251642e-07,
      "loss": -0.0,
      "reward": 1.578125,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 611
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.0,
      "epoch": 0.029570931580981834,
      "grad_norm": 0.0,
      "learning_rate": 4.852386934673367e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 612
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 63.78125,
      "epoch": 0.02961925009663703,
      "grad_norm": 5.909104866739217,
      "learning_rate": 4.852145342095091e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.26726123690605164,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 613
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 68.28125,
      "epoch": 0.02966756861229223,
      "grad_norm": 27.4862772215809,
      "learning_rate": 4.851903749516815e-07,
      "loss": -0.0,
      "reward": 1.578125,
      "reward_std": 0.4295162558555603,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 614
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 65.375,
      "epoch": 0.02971588712794743,
      "grad_norm": 13.354676195942373,
      "learning_rate": 4.851662156938539e-07,
      "loss": -0.0,
      "reward": 1.921875,
      "reward_std": 0.3051002025604248,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 615
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.53125,
      "epoch": 0.029764205643602628,
      "grad_norm": 12.726357408766225,
      "learning_rate": 4.851420564360262e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 616
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.78125,
      "epoch": 0.02981252415925783,
      "grad_norm": 0.0,
      "learning_rate": 4.851178971781987e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 617
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 68.40625,
      "epoch": 0.029860842674913027,
      "grad_norm": 11.953017837134174,
      "learning_rate": 4.85093737920371e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 618
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.4375,
      "epoch": 0.029909161190568224,
      "grad_norm": 0.0,
      "learning_rate": 4.850695786625434e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 619
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.125,
      "epoch": 0.029957479706223426,
      "grad_norm": 0.0,
      "learning_rate": 4.850454194047159e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 620
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.3125,
      "epoch": 0.030005798221878623,
      "grad_norm": 0.0,
      "learning_rate": 4.850212601468883e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 621
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.59375,
      "epoch": 0.030054116737533824,
      "grad_norm": 0.0,
      "learning_rate": 4.849971008890607e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 622
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.375,
      "epoch": 0.030102435253189022,
      "grad_norm": 0.0,
      "learning_rate": 4.849729416312331e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 623
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 62.1875,
      "epoch": 0.03015075376884422,
      "grad_norm": 7.186592311831673,
      "learning_rate": 4.849487823734055e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 624
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.9375,
      "epoch": 0.03019907228449942,
      "grad_norm": 0.0,
      "learning_rate": 4.849246231155778e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 625
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 77.6875,
      "epoch": 0.03024739080015462,
      "grad_norm": 11.538361579283595,
      "learning_rate": 4.849004638577502e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.37405285239219666,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 626
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.8125,
      "epoch": 0.03029570931580982,
      "grad_norm": 0.0,
      "learning_rate": 4.848763045999227e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 627
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 62.46875,
      "epoch": 0.030344027831465017,
      "grad_norm": 5.643264362122223,
      "learning_rate": 4.848521453420951e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 628
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 63.59375,
      "epoch": 0.030392346347120215,
      "grad_norm": 9.459660570730495,
      "learning_rate": 4.848279860842675e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 629
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.28125,
      "epoch": 0.030440664862775416,
      "grad_norm": 0.0,
      "learning_rate": 4.848038268264399e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 630
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 66.34375,
      "epoch": 0.030488983378430614,
      "grad_norm": 7.430737971991645,
      "learning_rate": 4.847796675686123e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 631
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 74.15625,
      "epoch": 0.030537301894085815,
      "grad_norm": 16.41308718469688,
      "learning_rate": 4.847555083107846e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 632
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 63.125,
      "epoch": 0.030585620409741013,
      "grad_norm": 15.607706678723433,
      "learning_rate": 4.847313490529571e-07,
      "loss": -0.0,
      "reward": 2.40625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 633
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.84375,
      "epoch": 0.03063393892539621,
      "grad_norm": 16.756879437565757,
      "learning_rate": 4.847071897951294e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.2982703447341919,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 634
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.6875,
      "epoch": 0.03068225744105141,
      "grad_norm": 0.0,
      "learning_rate": 4.846830305373019e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 635
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.90625,
      "epoch": 0.03073057595670661,
      "grad_norm": 0.0,
      "learning_rate": 4.846588712794743e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 636
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.0,
      "epoch": 0.03077889447236181,
      "grad_norm": 0.0,
      "learning_rate": 4.846347120216467e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 637
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.78125,
      "epoch": 0.030827212988017008,
      "grad_norm": 0.0,
      "learning_rate": 4.846105527638191e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 638
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.40625,
      "epoch": 0.030875531503672206,
      "grad_norm": 0.0,
      "learning_rate": 4.845863935059914e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 639
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 72.4375,
      "epoch": 0.030923850019327407,
      "grad_norm": 0.0,
      "learning_rate": 4.845622342481639e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 640
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 78.5,
      "epoch": 0.030972168534982605,
      "grad_norm": 9.001764839221995,
      "learning_rate": 4.845380749903362e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 641
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 69.8125,
      "epoch": 0.031020487050637806,
      "grad_norm": 25.825212619281874,
      "learning_rate": 4.845139157325086e-07,
      "loss": 0.0,
      "reward": 1.4375,
      "reward_std": 0.24359199404716492,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 642
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.78125,
      "epoch": 0.031068805566293003,
      "grad_norm": 0.0,
      "learning_rate": 4.844897564746811e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 643
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.59375,
      "epoch": 0.0311171240819482,
      "grad_norm": 0.0,
      "learning_rate": 4.844655972168535e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 644
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 72.625,
      "epoch": 0.031165442597603402,
      "grad_norm": 9.370095185243363,
      "learning_rate": 4.844414379590259e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.30935919284820557,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 645
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 69.15625,
      "epoch": 0.0312137611132586,
      "grad_norm": 41.23615584608041,
      "learning_rate": 4.844172787011983e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.22152717411518097,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 646
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 76.9375,
      "epoch": 0.0312620796289138,
      "grad_norm": 44.30554988782511,
      "learning_rate": 4.843931194433707e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.43236586451530457,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 647
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.3125,
      "epoch": 0.031310398144569,
      "grad_norm": 0.0,
      "learning_rate": 4.843689601855431e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 648
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 67.25,
      "epoch": 0.031358716660224197,
      "grad_norm": 31.591499827293326,
      "learning_rate": 4.843448009277154e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 649
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 67.75,
      "epoch": 0.031407035175879394,
      "grad_norm": 8.84804562986537,
      "learning_rate": 4.843206416698878e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 650
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 80.15625,
      "epoch": 0.0314553536915346,
      "grad_norm": 4.8970159682098995,
      "learning_rate": 4.842964824120603e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 651
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.1875,
      "epoch": 0.031503672207189796,
      "grad_norm": 0.0,
      "learning_rate": 4.842723231542327e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 652
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 80.6875,
      "epoch": 0.031551990722844994,
      "grad_norm": 17.96494965294427,
      "learning_rate": 4.842481638964051e-07,
      "loss": -0.0,
      "reward": 1.484375,
      "reward_std": 0.4592459797859192,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 653
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.59375,
      "epoch": 0.03160030923850019,
      "grad_norm": 7.87880414547985,
      "learning_rate": 4.842240046385775e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 654
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.40625,
      "epoch": 0.03164862775415539,
      "grad_norm": 0.0,
      "learning_rate": 4.841998453807499e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 655
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.71875,
      "epoch": 0.031696946269810594,
      "grad_norm": 0.0,
      "learning_rate": 4.841756861229222e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 656
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.8125,
      "epoch": 0.03174526478546579,
      "grad_norm": 0.0,
      "learning_rate": 4.841515268650947e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 657
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 62.625,
      "epoch": 0.03179358330112099,
      "grad_norm": 12.33371051158612,
      "learning_rate": 4.84127367607267e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 658
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 74.875,
      "epoch": 0.03184190181677619,
      "grad_norm": 32.58718299272937,
      "learning_rate": 4.841032083494395e-07,
      "loss": -0.0,
      "reward": 1.75,
      "reward_std": 0.38816189765930176,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 659
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 63.125,
      "epoch": 0.031890220332431385,
      "grad_norm": 6.429997326563001,
      "learning_rate": 4.840790490916119e-07,
      "loss": -0.0,
      "reward": 2.40625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 660
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 69.40625,
      "epoch": 0.03193853884808659,
      "grad_norm": 16.54318704366806,
      "learning_rate": 4.840548898337843e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 661
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 67.6875,
      "epoch": 0.03198685736374179,
      "grad_norm": 10.802788440236785,
      "learning_rate": 4.840307305759567e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 662
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 75.09375,
      "epoch": 0.032035175879396985,
      "grad_norm": 30.219287293633776,
      "learning_rate": 4.840065713181291e-07,
      "loss": -0.0,
      "reward": 2.109375,
      "reward_std": 0.3051002025604248,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 663
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.59375,
      "epoch": 0.03208349439505218,
      "grad_norm": 8.620698139132884,
      "learning_rate": 4.839824120603015e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 664
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 68.875,
      "epoch": 0.03213181291070738,
      "grad_norm": 11.996497553356471,
      "learning_rate": 4.839582528024738e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 665
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 69.8125,
      "epoch": 0.032180131426362585,
      "grad_norm": 0.0,
      "learning_rate": 4.839340935446463e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 666
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 82.9375,
      "epoch": 0.03222844994201778,
      "grad_norm": 11.315451271701056,
      "learning_rate": 4.839099342868187e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 667
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.875,
      "epoch": 0.03227676845767298,
      "grad_norm": 13.014718478698589,
      "learning_rate": 4.838857750289911e-07,
      "loss": -0.0,
      "reward": 2.4375,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 668
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.71875,
      "epoch": 0.03232508697332818,
      "grad_norm": 19.17690137288859,
      "learning_rate": 4.838616157711635e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.2993341088294983,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 669
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 73.21875,
      "epoch": 0.032373405488983376,
      "grad_norm": 8.81082055317522,
      "learning_rate": 4.838374565133359e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 670
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.65625,
      "epoch": 0.03242172400463858,
      "grad_norm": 0.0,
      "learning_rate": 4.838132972555083e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 671
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.84375,
      "epoch": 0.03247004252029378,
      "grad_norm": 0.0,
      "learning_rate": 4.837891379976806e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 672
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.375,
      "epoch": 0.032518361035948976,
      "grad_norm": 0.0,
      "learning_rate": 4.837649787398531e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 673
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 71.5625,
      "epoch": 0.03256667955160417,
      "grad_norm": 0.0,
      "learning_rate": 4.837408194820255e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 674
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 67.9375,
      "epoch": 0.03261499806725937,
      "grad_norm": 4.777210249843169,
      "learning_rate": 4.837166602241979e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 675
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.0,
      "epoch": 0.032663316582914576,
      "grad_norm": 0.0,
      "learning_rate": 4.836925009663703e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 676
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.09375,
      "epoch": 0.03271163509856977,
      "grad_norm": 8.230800291937914,
      "learning_rate": 4.836683417085427e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 677
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 79.09375,
      "epoch": 0.03275995361422497,
      "grad_norm": 27.190450409960217,
      "learning_rate": 4.836441824507151e-07,
      "loss": 0.0,
      "reward": 1.453125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 678
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 71.71875,
      "epoch": 0.03280827212988017,
      "grad_norm": 54.391136497460245,
      "learning_rate": 4.836200231928875e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.3051002025604248,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 679
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.28125,
      "epoch": 0.032856590645535366,
      "grad_norm": 0.0,
      "learning_rate": 4.835958639350599e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 680
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.625,
      "epoch": 0.03290490916119057,
      "grad_norm": 12.155327906500414,
      "learning_rate": 4.835717046772322e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.23827512562274933,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 681
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.65625,
      "epoch": 0.03295322767684577,
      "grad_norm": 12.018770823210593,
      "learning_rate": 4.835475454194047e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.3072218894958496,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 682
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 71.625,
      "epoch": 0.033001546192500966,
      "grad_norm": 33.50900348277447,
      "learning_rate": 4.835233861615771e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.3662523031234741,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 683
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.71875,
      "epoch": 0.033049864708156164,
      "grad_norm": 0.0,
      "learning_rate": 4.834992269037495e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 684
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 85.375,
      "epoch": 0.03309818322381136,
      "grad_norm": 12.701721132799488,
      "learning_rate": 4.834750676459219e-07,
      "loss": 0.0,
      "reward": 1.53125,
      "reward_std": 0.33725470304489136,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 685
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 80.1875,
      "epoch": 0.033146501739466566,
      "grad_norm": 28.35743236750874,
      "learning_rate": 4.834509083880943e-07,
      "loss": -0.0,
      "reward": 2.265625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 686
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 83.78125,
      "epoch": 0.033194820255121764,
      "grad_norm": 11.10088468859443,
      "learning_rate": 4.834267491302667e-07,
      "loss": -0.0,
      "reward": 2.109375,
      "reward_std": 0.37298911809921265,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 687
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.40625,
      "epoch": 0.03324313877077696,
      "grad_norm": 0.0,
      "learning_rate": 4.834025898724391e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 688
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 93.75,
      "epoch": 0.03329145728643216,
      "grad_norm": 15.005074476885738,
      "learning_rate": 4.833784306146114e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.434487521648407,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 689
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.09375,
      "epoch": 0.03333977580208736,
      "grad_norm": 8.011955821521116,
      "learning_rate": 4.83354271356784e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 690
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.5625,
      "epoch": 0.03338809431774256,
      "grad_norm": 0.0,
      "learning_rate": 4.833301120989563e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 691
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.0,
      "epoch": 0.03343641283339776,
      "grad_norm": 27.645893375825747,
      "learning_rate": 4.833059528411287e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.22049658000469208,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 692
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 69.09375,
      "epoch": 0.03348473134905296,
      "grad_norm": 9.567998819592749,
      "learning_rate": 4.832817935833011e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.25947636365890503,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 693
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.1875,
      "epoch": 0.033533049864708155,
      "grad_norm": 0.0,
      "learning_rate": 4.832576343254735e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 694
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.0625,
      "epoch": 0.03358136838036335,
      "grad_norm": 0.0,
      "learning_rate": 4.832334750676459e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 695
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 83.96875,
      "epoch": 0.03362968689601856,
      "grad_norm": 17.194680559072097,
      "learning_rate": 4.832093158098182e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.4555858373641968,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 696
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.21875,
      "epoch": 0.033678005411673755,
      "grad_norm": 9.199564210910864,
      "learning_rate": 4.831851565519908e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 697
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 89.5625,
      "epoch": 0.03372632392732895,
      "grad_norm": 24.961803329211154,
      "learning_rate": 4.831609972941631e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.4171832799911499,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 698
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.71875,
      "epoch": 0.03377464244298415,
      "grad_norm": 0.0,
      "learning_rate": 4.831368380363355e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 699
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.0,
      "epoch": 0.03382296095863935,
      "grad_norm": 9.976470515230933,
      "learning_rate": 4.831126787785079e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 700
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.125,
      "epoch": 0.03387127947429455,
      "grad_norm": 5.767873605215611,
      "learning_rate": 4.830885195206803e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 701
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 93.78125,
      "epoch": 0.03391959798994975,
      "grad_norm": 7.678129716819046,
      "learning_rate": 4.830643602628527e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.23827512562274933,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 702
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 100.5,
      "epoch": 0.03396791650560495,
      "grad_norm": 24.15603108826663,
      "learning_rate": 4.830402010050251e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 703
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 89.59375,
      "epoch": 0.034016235021260145,
      "grad_norm": 48.56315296243857,
      "learning_rate": 4.830160417471975e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.420942485332489,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 704
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 102.875,
      "epoch": 0.03406455353691534,
      "grad_norm": 10.266408170166347,
      "learning_rate": 4.829918824893699e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.3006556034088135,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 705
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.875,
      "epoch": 0.03411287205257055,
      "grad_norm": 16.188665533169686,
      "learning_rate": 4.829677232315423e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.2651395797729492,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 706
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 104.0625,
      "epoch": 0.034161190568225745,
      "grad_norm": 14.50642985218045,
      "learning_rate": 4.829435639737147e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.32666346430778503,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 707
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.25,
      "epoch": 0.03420950908388094,
      "grad_norm": 0.0,
      "learning_rate": 4.829194047158871e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 708
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 99.125,
      "epoch": 0.03425782759953614,
      "grad_norm": 14.695149025532306,
      "learning_rate": 4.828952454580595e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.3029785752296448,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 709
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 68.96875,
      "epoch": 0.03430614611519134,
      "grad_norm": 12.643878566521005,
      "learning_rate": 4.828710862002319e-07,
      "loss": -0.0,
      "reward": 1.375,
      "reward_std": 0.21316051483154297,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 710
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 85.3125,
      "epoch": 0.03435446463084654,
      "grad_norm": 3.8508668661010854,
      "learning_rate": 4.828469269424043e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 711
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.1875,
      "epoch": 0.03440278314650174,
      "grad_norm": 0.0,
      "learning_rate": 4.828227676845766e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 712
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.625,
      "epoch": 0.03445110166215694,
      "grad_norm": 0.0,
      "learning_rate": 4.827986084267492e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 713
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.9375,
      "epoch": 0.034499420177812136,
      "grad_norm": 0.0,
      "learning_rate": 4.827744491689215e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 714
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 92.4375,
      "epoch": 0.034547738693467334,
      "grad_norm": 7.424575246437304,
      "learning_rate": 4.827502899110939e-07,
      "loss": -0.0,
      "reward": 2.390625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 715
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 99.125,
      "epoch": 0.03459605720912254,
      "grad_norm": 87.44035257674682,
      "learning_rate": 4.827261306532663e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.4786914885044098,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 716
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 90.40625,
      "epoch": 0.034644375724777736,
      "grad_norm": 14.12605951940578,
      "learning_rate": 4.827019713954387e-07,
      "loss": 0.0,
      "reward": 1.1875,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.15625,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 717
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 103.5,
      "epoch": 0.034692694240432934,
      "grad_norm": 11.039276828154502,
      "learning_rate": 4.826778121376111e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 718
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 83.5,
      "epoch": 0.03474101275608813,
      "grad_norm": 5.364642402444726,
      "learning_rate": 4.826536528797835e-07,
      "loss": -0.0,
      "reward": 2.375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 719
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 93.9375,
      "epoch": 0.03478933127174333,
      "grad_norm": 46.13206352413814,
      "learning_rate": 4.82629493621956e-07,
      "loss": -0.0,
      "reward": 2.3125,
      "reward_std": 0.4387563169002533,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 720
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.34375,
      "epoch": 0.034837649787398534,
      "grad_norm": 0.0,
      "learning_rate": 4.826053343641284e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 721
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 80.8125,
      "epoch": 0.03488596830305373,
      "grad_norm": 12.188736656431658,
      "learning_rate": 4.825811751063007e-07,
      "loss": -0.0,
      "reward": 2.328125,
      "reward_std": 0.24831002950668335,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 722
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 90.34375,
      "epoch": 0.03493428681870893,
      "grad_norm": 8.474149932225279,
      "learning_rate": 4.825570158484731e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 723
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.40625,
      "epoch": 0.03498260533436413,
      "grad_norm": 0.0,
      "learning_rate": 4.825328565906455e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 724
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.25,
      "epoch": 0.035030923850019324,
      "grad_norm": 7.506768111291514,
      "learning_rate": 4.825086973328179e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.25326618552207947,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 725
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.96875,
      "epoch": 0.03507924236567453,
      "grad_norm": 0.0,
      "learning_rate": 4.824845380749903e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 726
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 97.90625,
      "epoch": 0.03512756088132973,
      "grad_norm": 189.81011222885738,
      "learning_rate": 4.824603788171627e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.5214560031890869,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 727
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.6875,
      "epoch": 0.035175879396984924,
      "grad_norm": 102.81817094670848,
      "learning_rate": 4.824362195593352e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.3319548964500427,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 728
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 80.5625,
      "epoch": 0.03522419791264012,
      "grad_norm": 12.45307194720028,
      "learning_rate": 4.824120603015075e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.29305070638656616,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 729
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 97.71875,
      "epoch": 0.03527251642829532,
      "grad_norm": 9.390118393615186,
      "learning_rate": 4.8238790104368e-07,
      "loss": 0.0,
      "reward": 1.09375,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.0625,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 730
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.875,
      "epoch": 0.035320834943950524,
      "grad_norm": 0.0,
      "learning_rate": 4.823637417858523e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 731
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.4375,
      "epoch": 0.03536915345960572,
      "grad_norm": 6.020499169322649,
      "learning_rate": 4.823395825280247e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 732
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.8125,
      "epoch": 0.03541747197526092,
      "grad_norm": 0.0,
      "learning_rate": 4.823154232701971e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 733
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 82.46875,
      "epoch": 0.03546579049091612,
      "grad_norm": 8.554332083501984,
      "learning_rate": 4.822912640123695e-07,
      "loss": -0.0,
      "reward": 1.5625,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 734
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 76.28125,
      "epoch": 0.035514109006571315,
      "grad_norm": 14.533620958881228,
      "learning_rate": 4.822671047545419e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.30509042739868164,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 735
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 73.28125,
      "epoch": 0.03556242752222652,
      "grad_norm": 12.656647869271634,
      "learning_rate": 4.822429454967143e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 736
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.9375,
      "epoch": 0.03561074603788172,
      "grad_norm": 0.0,
      "learning_rate": 4.822187862388868e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 737
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.65625,
      "epoch": 0.035659064553536915,
      "grad_norm": 10.203883022365265,
      "learning_rate": 4.821946269810591e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.3029785752296448,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 738
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 82.625,
      "epoch": 0.03570738306919211,
      "grad_norm": 10.971401575727358,
      "learning_rate": 4.821704677232315e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.43768274784088135,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 739
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.1875,
      "epoch": 0.03575570158484731,
      "grad_norm": 0.0,
      "learning_rate": 4.821463084654039e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 740
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.03125,
      "epoch": 0.035804020100502515,
      "grad_norm": 0.0,
      "learning_rate": 4.821221492075763e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 741
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 83.1875,
      "epoch": 0.03585233861615771,
      "grad_norm": 11.874524067294828,
      "learning_rate": 4.820979899497487e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 742
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.4375,
      "epoch": 0.03590065713181291,
      "grad_norm": 0.0,
      "learning_rate": 4.820738306919211e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 743
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.21875,
      "epoch": 0.03594897564746811,
      "grad_norm": 16.050132059422644,
      "learning_rate": 4.820496714340936e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.3735883831977844,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 744
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 74.53125,
      "epoch": 0.035997294163123306,
      "grad_norm": 21.91979511646889,
      "learning_rate": 4.820255121762659e-07,
      "loss": -0.0,
      "reward": 2.140625,
      "reward_std": 0.5013028383255005,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 745
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.71875,
      "epoch": 0.03604561267877851,
      "grad_norm": 0.0,
      "learning_rate": 4.820013529184383e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 0.96875,
      "step": 746
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.15625,
      "epoch": 0.03609393119443371,
      "grad_norm": 0.0,
      "learning_rate": 4.819771936606107e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 747
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 80.4375,
      "epoch": 0.036142249710088906,
      "grad_norm": 11.54462612390227,
      "learning_rate": 4.819530344027831e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 748
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.34375,
      "epoch": 0.0361905682257441,
      "grad_norm": 0.0,
      "learning_rate": 4.819288751449555e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.3119301199913025,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 0.90625,
      "step": 749
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.3125,
      "epoch": 0.0362388867413993,
      "grad_norm": 0.0,
      "learning_rate": 4.819047158871279e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 750
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.84375,
      "epoch": 0.036287205257054506,
      "grad_norm": 0.0,
      "learning_rate": 4.818805566293004e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 751
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.625,
      "epoch": 0.0363355237727097,
      "grad_norm": 5.839419274966806,
      "learning_rate": 4.818563973714727e-07,
      "loss": -0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 752
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.78125,
      "epoch": 0.0363838422883649,
      "grad_norm": 6.266661733753141,
      "learning_rate": 4.818322381136452e-07,
      "loss": -0.0,
      "reward": 2.078125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 753
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.46875,
      "epoch": 0.0364321608040201,
      "grad_norm": 0.0,
      "learning_rate": 4.818080788558175e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 754
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.375,
      "epoch": 0.036480479319675296,
      "grad_norm": 0.0,
      "learning_rate": 4.817839195979899e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 755
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.09375,
      "epoch": 0.0365287978353305,
      "grad_norm": 0.0,
      "learning_rate": 4.817597603401623e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 0.9375,
      "step": 756
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 93.3125,
      "epoch": 0.0365771163509857,
      "grad_norm": 15.835378471384333,
      "learning_rate": 4.817356010823347e-07,
      "loss": 0.0,
      "reward": 0.984375,
      "reward_std": 0.24831002950668335,
      "rewards/accuracy_reward": 0.03125,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 0.9375,
      "step": 757
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 82.84375,
      "epoch": 0.036625434866640896,
      "grad_norm": 7.968956903464979,
      "learning_rate": 4.817114418245071e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.3325796127319336,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 0.9375,
      "step": 758
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.65625,
      "epoch": 0.036673753382296094,
      "grad_norm": 0.0,
      "learning_rate": 4.816872825666796e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 0.96875,
      "step": 759
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.3125,
      "epoch": 0.03672207189795129,
      "grad_norm": 0.0,
      "learning_rate": 4.81663123308852e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 760
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 74.21875,
      "epoch": 0.036770390413606496,
      "grad_norm": 7.422776406648508,
      "learning_rate": 4.816389640510244e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 761
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 81.71875,
      "epoch": 0.036818708929261694,
      "grad_norm": 0.0,
      "learning_rate": 4.816148047931967e-07,
      "loss": 0.0,
      "reward": 0.96875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 0.96875,
      "step": 762
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.3125,
      "epoch": 0.03686702744491689,
      "grad_norm": 0.0,
      "learning_rate": 4.815906455353691e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 0.96875,
      "step": 763
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.5625,
      "epoch": 0.03691534596057209,
      "grad_norm": 0.0,
      "learning_rate": 4.815664862775415e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 764
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 82.4375,
      "epoch": 0.03696366447622729,
      "grad_norm": 7.604424968467871,
      "learning_rate": 4.815423270197139e-07,
      "loss": 0.0,
      "reward": 1.5,
      "reward_std": 0.3362240791320801,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 0.9375,
      "step": 765
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.875,
      "epoch": 0.03701198299188249,
      "grad_norm": 0.0,
      "learning_rate": 4.815181677618863e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 0.96875,
      "step": 766
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.1875,
      "epoch": 0.03706030150753769,
      "grad_norm": 0.0,
      "learning_rate": 4.814940085040588e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 767
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.375,
      "epoch": 0.03710862002319289,
      "grad_norm": 0.0,
      "learning_rate": 4.814698492462312e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 768
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 70.09375,
      "epoch": 0.037156938538848085,
      "grad_norm": 11.236052706360347,
      "learning_rate": 4.814456899884035e-07,
      "loss": -0.0,
      "reward": 2.3125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 769
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.28125,
      "epoch": 0.03720525705450328,
      "grad_norm": 0.0,
      "learning_rate": 4.81421530730576e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 770
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.9375,
      "epoch": 0.03725357557015849,
      "grad_norm": 0.0,
      "learning_rate": 4.813973714727483e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 0.9375,
      "step": 771
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.0625,
      "epoch": 0.037301894085813685,
      "grad_norm": 7.936719617202704,
      "learning_rate": 4.813732122149207e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 772
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 80.3125,
      "epoch": 0.03735021260146888,
      "grad_norm": 14.311284662216584,
      "learning_rate": 4.813490529570931e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 773
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.53125,
      "epoch": 0.03739853111712408,
      "grad_norm": 18.489445466439868,
      "learning_rate": 4.813248936992655e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 774
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.78125,
      "epoch": 0.03744684963277928,
      "grad_norm": 0.0,
      "learning_rate": 4.81300734441438e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 775
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 88.53125,
      "epoch": 0.03749516814843448,
      "grad_norm": 11.263950366899145,
      "learning_rate": 4.812765751836103e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 776
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 83.125,
      "epoch": 0.03754348666408968,
      "grad_norm": 23.979997739267283,
      "learning_rate": 4.812524159257828e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.40774765610694885,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 777
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 76.84375,
      "epoch": 0.03759180517974488,
      "grad_norm": 13.54159396109102,
      "learning_rate": 4.812282566679551e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.28754231333732605,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 778
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 74.46875,
      "epoch": 0.037640123695400075,
      "grad_norm": 9.158176314475016,
      "learning_rate": 4.812040974101275e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 779
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.40625,
      "epoch": 0.03768844221105527,
      "grad_norm": 0.0,
      "learning_rate": 4.811799381522999e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 780
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.15625,
      "epoch": 0.03773676072671048,
      "grad_norm": 0.0,
      "learning_rate": 4.811557788944723e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 781
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.0,
      "epoch": 0.037785079242365675,
      "grad_norm": 0.0,
      "learning_rate": 4.811316196366448e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 782
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.0,
      "epoch": 0.03783339775802087,
      "grad_norm": 5.718393884399284,
      "learning_rate": 4.811074603788172e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 783
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.125,
      "epoch": 0.03788171627367607,
      "grad_norm": 0.0,
      "learning_rate": 4.810833011209896e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 784
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.90625,
      "epoch": 0.03793003478933127,
      "grad_norm": 0.0,
      "learning_rate": 4.810591418631619e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 785
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.5,
      "epoch": 0.03797835330498647,
      "grad_norm": 0.0,
      "learning_rate": 4.810349826053343e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 786
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.03125,
      "epoch": 0.03802667182064167,
      "grad_norm": 0.0,
      "learning_rate": 4.810108233475067e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 787
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.875,
      "epoch": 0.03807499033629687,
      "grad_norm": 29.882995340475595,
      "learning_rate": 4.809866640896791e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 788
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.0,
      "epoch": 0.038123308851952066,
      "grad_norm": 0.0,
      "learning_rate": 4.809625048318515e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 789
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.875,
      "epoch": 0.038171627367607264,
      "grad_norm": 0.0,
      "learning_rate": 4.80938345574024e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 790
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.65625,
      "epoch": 0.03821994588326247,
      "grad_norm": 0.0,
      "learning_rate": 4.809141863161964e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 791
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 89.5,
      "epoch": 0.038268264398917666,
      "grad_norm": 9.173959765669888,
      "learning_rate": 4.808900270583688e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 792
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.71875,
      "epoch": 0.038316582914572864,
      "grad_norm": 0.0,
      "learning_rate": 4.808658678005412e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 793
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 100.25,
      "epoch": 0.03836490143022806,
      "grad_norm": 14.345421913984959,
      "learning_rate": 4.808417085427135e-07,
      "loss": 0.0,
      "reward": 1.09375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.0625,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 794
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.59375,
      "epoch": 0.03841321994588326,
      "grad_norm": 0.0,
      "learning_rate": 4.808175492848859e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 795
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.625,
      "epoch": 0.038461538461538464,
      "grad_norm": 4.612879620401458,
      "learning_rate": 4.807933900270583e-07,
      "loss": -0.0,
      "reward": 2.40625,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 796
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.5,
      "epoch": 0.03850985697719366,
      "grad_norm": 0.0,
      "learning_rate": 4.807692307692307e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 797
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 87.34375,
      "epoch": 0.03855817549284886,
      "grad_norm": 17.498057646849485,
      "learning_rate": 4.807450715114032e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 798
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 85.40625,
      "epoch": 0.03860649400850406,
      "grad_norm": 10.336931818759867,
      "learning_rate": 4.807209122535756e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 799
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 93.90625,
      "epoch": 0.038654812524159254,
      "grad_norm": 6.646972726998763,
      "learning_rate": 4.80696752995748e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 800
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 96.5,
      "epoch": 0.03870313103981446,
      "grad_norm": 17.61425650037914,
      "learning_rate": 4.806725937379204e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.5193343162536621,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 801
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.03125,
      "epoch": 0.03875144955546966,
      "grad_norm": 0.0,
      "learning_rate": 4.806484344800927e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 802
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 94.0625,
      "epoch": 0.038799768071124854,
      "grad_norm": 7.055790591101708,
      "learning_rate": 4.806242752222652e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 803
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.125,
      "epoch": 0.03884808658678005,
      "grad_norm": 11.24677003447741,
      "learning_rate": 4.806001159644375e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 804
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 95.8125,
      "epoch": 0.03889640510243525,
      "grad_norm": 9.128232520653365,
      "learning_rate": 4.805759567066099e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 805
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.59375,
      "epoch": 0.038944723618090454,
      "grad_norm": 0.0,
      "learning_rate": 4.805517974487824e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 806
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.0,
      "epoch": 0.03899304213374565,
      "grad_norm": 30.076657650209267,
      "learning_rate": 4.805276381909548e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 807
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.03125,
      "epoch": 0.03904136064940085,
      "grad_norm": 0.0,
      "learning_rate": 4.805034789331272e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 808
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.8125,
      "epoch": 0.03908967916505605,
      "grad_norm": 0.0,
      "learning_rate": 4.804793196752995e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 809
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.8125,
      "epoch": 0.03913799768071125,
      "grad_norm": 0.0,
      "learning_rate": 4.80455160417472e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 810
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.46875,
      "epoch": 0.03918631619636645,
      "grad_norm": 6.381092665281912,
      "learning_rate": 4.804310011596443e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 811
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 94.90625,
      "epoch": 0.03923463471202165,
      "grad_norm": 61.58264768883825,
      "learning_rate": 4.804068419018167e-07,
      "loss": -0.0,
      "reward": 1.921875,
      "reward_std": 0.30935919284820557,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 812
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 87.5,
      "epoch": 0.039282953227676845,
      "grad_norm": 8.587625690164817,
      "learning_rate": 4.803826826439891e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.22049658000469208,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 813
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.6875,
      "epoch": 0.03933127174333204,
      "grad_norm": 0.0,
      "learning_rate": 4.803585233861616e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 814
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.59375,
      "epoch": 0.03937959025898725,
      "grad_norm": 0.0,
      "learning_rate": 4.80334364128334e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 815
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 98.59375,
      "epoch": 0.039427908774642445,
      "grad_norm": 0.0,
      "learning_rate": 4.803102048705063e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 816
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 95.34375,
      "epoch": 0.03947622729029764,
      "grad_norm": 11.82637554383909,
      "learning_rate": 4.802860456126788e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.3708576560020447,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 817
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 101.59375,
      "epoch": 0.03952454580595284,
      "grad_norm": 6.803235324425333,
      "learning_rate": 4.802618863548511e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 818
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 96.53125,
      "epoch": 0.03957286432160804,
      "grad_norm": 5.574719815942869,
      "learning_rate": 4.802377270970235e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 819
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 89.71875,
      "epoch": 0.03962118283726324,
      "grad_norm": 10.201563753453446,
      "learning_rate": 4.802135678391959e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.2893187999725342,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 820
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.4375,
      "epoch": 0.03966950135291844,
      "grad_norm": 0.0,
      "learning_rate": 4.801894085813684e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 821
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 102.34375,
      "epoch": 0.03971781986857364,
      "grad_norm": 37.62087016409019,
      "learning_rate": 4.801652493235408e-07,
      "loss": 0.0,
      "reward": 1.5,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 822
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.78125,
      "epoch": 0.039766138384228836,
      "grad_norm": 0.0,
      "learning_rate": 4.801410900657132e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 823
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.65625,
      "epoch": 0.03981445689988403,
      "grad_norm": 0.0,
      "learning_rate": 4.801169308078856e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 824
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 104.46875,
      "epoch": 0.03986277541553924,
      "grad_norm": 6.373375091310423,
      "learning_rate": 4.800927715500579e-07,
      "loss": -0.0,
      "reward": 1.09375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.0625,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 825
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 98.125,
      "epoch": 0.039911093931194436,
      "grad_norm": 12.82739181555276,
      "learning_rate": 4.800686122922303e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.5038894414901733,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 826
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.625,
      "epoch": 0.03995941244684963,
      "grad_norm": 0.0,
      "learning_rate": 4.800444530344027e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 827
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 96.59375,
      "epoch": 0.04000773096250483,
      "grad_norm": 11.211856798313276,
      "learning_rate": 4.800202937765751e-07,
      "loss": 0.0,
      "reward": 1.546875,
      "reward_std": 0.25069957971572876,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 828
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 95.84375,
      "epoch": 0.04005604947816003,
      "grad_norm": 7.509299503105637,
      "learning_rate": 4.799961345187476e-07,
      "loss": -0.0,
      "reward": 1.515625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 829
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 100.625,
      "epoch": 0.04010436799381523,
      "grad_norm": 17.217735934969998,
      "learning_rate": 4.7997197526092e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.4592205286026001,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 830
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.71875,
      "epoch": 0.04015268650947043,
      "grad_norm": 0.0,
      "learning_rate": 4.799478160030924e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 831
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.1875,
      "epoch": 0.04020100502512563,
      "grad_norm": 0.0,
      "learning_rate": 4.799236567452648e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 832
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 95.875,
      "epoch": 0.040249323540780826,
      "grad_norm": 5.784762280024729,
      "learning_rate": 4.798994974874372e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 833
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 98.09375,
      "epoch": 0.040297642056436024,
      "grad_norm": 7.3172688475997605,
      "learning_rate": 4.798753382296096e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.3051002323627472,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 834
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.53125,
      "epoch": 0.04034596057209123,
      "grad_norm": 0.0,
      "learning_rate": 4.798511789717819e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 835
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.09375,
      "epoch": 0.040394279087746426,
      "grad_norm": 0.0,
      "learning_rate": 4.798270197139543e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 836
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.125,
      "epoch": 0.040442597603401624,
      "grad_norm": 0.0,
      "learning_rate": 4.798028604561268e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 837
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.5,
      "epoch": 0.04049091611905682,
      "grad_norm": 0.0,
      "learning_rate": 4.797787011982992e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 838
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 88.6875,
      "epoch": 0.04053923463471202,
      "grad_norm": 12.28368905312229,
      "learning_rate": 4.797545419404716e-07,
      "loss": -0.0,
      "reward": 2.140625,
      "reward_std": 0.44946879148483276,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 839
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 94.625,
      "epoch": 0.040587553150367224,
      "grad_norm": 6.061909596895182,
      "learning_rate": 4.79730382682644e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 840
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 108.0,
      "epoch": 0.04063587166602242,
      "grad_norm": 10.225807510244545,
      "learning_rate": 4.797062234248164e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.28566449880599976,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 841
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 101.9375,
      "epoch": 0.04068419018167762,
      "grad_norm": 12.646208496271484,
      "learning_rate": 4.796820641669887e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.2608962655067444,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 842
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 109.96875,
      "epoch": 0.04073250869733282,
      "grad_norm": 5.967776127193752,
      "learning_rate": 4.796579049091612e-07,
      "loss": 0.0,
      "reward": 1.140625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.09375,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 843
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 101.65625,
      "epoch": 0.040780827212988015,
      "grad_norm": 17.019349268346954,
      "learning_rate": 4.796337456513335e-07,
      "loss": -0.0,
      "reward": 1.6875,
      "reward_std": 0.3823235034942627,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 844
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 102.3125,
      "epoch": 0.04082914572864322,
      "grad_norm": 8.685474622472322,
      "learning_rate": 4.79609586393506e-07,
      "loss": -0.0,
      "reward": 1.6875,
      "reward_std": 0.27102428674697876,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 845
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 107.90625,
      "epoch": 0.04087746424429842,
      "grad_norm": 28.95345990477218,
      "learning_rate": 4.795854271356784e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.6296420097351074,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 846
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.625,
      "epoch": 0.040925782759953615,
      "grad_norm": 0.0,
      "learning_rate": 4.795612678778508e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 847
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 109.84375,
      "epoch": 0.04097410127560881,
      "grad_norm": 8.102644632628698,
      "learning_rate": 4.795371086200232e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.3234682083129883,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 848
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.59375,
      "epoch": 0.04102241979126401,
      "grad_norm": 0.0,
      "learning_rate": 4.795129493621955e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 849
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 99.59375,
      "epoch": 0.041070738306919215,
      "grad_norm": 8.37080195260839,
      "learning_rate": 4.79488790104368e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 850
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.71875,
      "epoch": 0.04111905682257441,
      "grad_norm": 5.4921267841486126,
      "learning_rate": 4.794646308465403e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 851
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.125,
      "epoch": 0.04116737533822961,
      "grad_norm": 0.0,
      "learning_rate": 4.794404715887128e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 852
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.6875,
      "epoch": 0.04121569385388481,
      "grad_norm": 0.0,
      "learning_rate": 4.794163123308852e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 853
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 95.4375,
      "epoch": 0.041264012369540005,
      "grad_norm": 11.990479790567758,
      "learning_rate": 4.793921530730576e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 854
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 108.03125,
      "epoch": 0.04131233088519521,
      "grad_norm": 12.581987640141126,
      "learning_rate": 4.7936799381523e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.3234682083129883,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 855
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 102.53125,
      "epoch": 0.04136064940085041,
      "grad_norm": 14.134321841362977,
      "learning_rate": 4.793438345574023e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.285190224647522,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 856
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 97.84375,
      "epoch": 0.041408967916505605,
      "grad_norm": 8.03648633888074,
      "learning_rate": 4.793196752995748e-07,
      "loss": -0.0,
      "reward": 2.140625,
      "reward_std": 0.3277115821838379,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 857
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 100.78125,
      "epoch": 0.0414572864321608,
      "grad_norm": 93.73663587653057,
      "learning_rate": 4.792955160417471e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 858
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 104.28125,
      "epoch": 0.041505604947816,
      "grad_norm": 6.955821230682496,
      "learning_rate": 4.792713567839195e-07,
      "loss": -0.0,
      "reward": 1.21875,
      "reward_std": 0.18600594997406006,
      "rewards/accuracy_reward": 0.15625,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 859
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 95.8125,
      "epoch": 0.041553923463471205,
      "grad_norm": 7.474120953995301,
      "learning_rate": 4.79247197526092e-07,
      "loss": -0.0,
      "reward": 2.3125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 860
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.09375,
      "epoch": 0.0416022419791264,
      "grad_norm": 0.0,
      "learning_rate": 4.792230382682644e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 861
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 104.125,
      "epoch": 0.0416505604947816,
      "grad_norm": 7.726201911523174,
      "learning_rate": 4.791988790104368e-07,
      "loss": -0.0,
      "reward": 2.3125,
      "reward_std": 0.27939093112945557,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 862
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.125,
      "epoch": 0.0416988790104368,
      "grad_norm": 0.0,
      "learning_rate": 4.791747197526092e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 863
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.8125,
      "epoch": 0.041747197526091996,
      "grad_norm": 5.377623145592117,
      "learning_rate": 4.791505604947816e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 864
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 106.4375,
      "epoch": 0.0417955160417472,
      "grad_norm": 45.48453093909144,
      "learning_rate": 4.791264012369539e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.4343488812446594,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 865
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.6875,
      "epoch": 0.0418438345574024,
      "grad_norm": 0.0,
      "learning_rate": 4.791022419791263e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 866
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 89.78125,
      "epoch": 0.041892153073057596,
      "grad_norm": 17.81004154539386,
      "learning_rate": 4.790780827212987e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.4429898262023926,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 867
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 99.84375,
      "epoch": 0.041940471588712794,
      "grad_norm": 192.51571311647606,
      "learning_rate": 4.790539234634712e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.3913571238517761,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 868
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 98.46875,
      "epoch": 0.04198879010436799,
      "grad_norm": 11.229086672285138,
      "learning_rate": 4.790297642056436e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 869
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.6875,
      "epoch": 0.042037108620023196,
      "grad_norm": 0.0,
      "learning_rate": 4.79005604947816e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 870
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 106.3125,
      "epoch": 0.042085427135678394,
      "grad_norm": 6.19017552147854,
      "learning_rate": 4.789814456899884e-07,
      "loss": 0.0,
      "reward": 1.28125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 871
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 104.125,
      "epoch": 0.04213374565133359,
      "grad_norm": 21.534418834506475,
      "learning_rate": 4.789572864321608e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.41878676414489746,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 872
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.15625,
      "epoch": 0.04218206416698879,
      "grad_norm": 0.0,
      "learning_rate": 4.789331271743332e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 873
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 106.375,
      "epoch": 0.04223038268264399,
      "grad_norm": 8.991943013666415,
      "learning_rate": 4.789089679165056e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 874
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 99.78125,
      "epoch": 0.04227870119829919,
      "grad_norm": 18.249922059745007,
      "learning_rate": 4.78884808658678e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.2529362142086029,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 875
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.34375,
      "epoch": 0.04232701971395439,
      "grad_norm": 0.0,
      "learning_rate": 4.788606494008504e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 876
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 102.8125,
      "epoch": 0.04237533822960959,
      "grad_norm": 0.0,
      "learning_rate": 4.788364901430228e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 877
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 104.78125,
      "epoch": 0.042423656745264784,
      "grad_norm": 30.260047948791392,
      "learning_rate": 4.788123308851952e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.48448851704597473,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 878
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.09375,
      "epoch": 0.04247197526091998,
      "grad_norm": 0.0,
      "learning_rate": 4.787881716273676e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 879
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.625,
      "epoch": 0.04252029377657519,
      "grad_norm": 5.95005910016059,
      "learning_rate": 4.7876401236954e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.32416999340057373,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 880
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 102.65625,
      "epoch": 0.042568612292230384,
      "grad_norm": 6.964006534989229,
      "learning_rate": 4.787398531117124e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 881
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.0,
      "epoch": 0.04261693080788558,
      "grad_norm": 8.720546482706732,
      "learning_rate": 4.787156938538847e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.26621314883232117,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 882
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 95.59375,
      "epoch": 0.04266524932354078,
      "grad_norm": 7.2953457427014525,
      "learning_rate": 4.786915345960573e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 883
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.53125,
      "epoch": 0.04271356783919598,
      "grad_norm": 0.0,
      "learning_rate": 4.786673753382296e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 884
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 107.09375,
      "epoch": 0.04276188635485118,
      "grad_norm": 8.674821740873439,
      "learning_rate": 4.78643216080402e-07,
      "loss": -0.0,
      "reward": 1.921875,
      "reward_std": 0.188242569565773,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 885
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.0625,
      "epoch": 0.04281020487050638,
      "grad_norm": 0.0,
      "learning_rate": 4.786190568225744e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 886
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 111.65625,
      "epoch": 0.04285852338616158,
      "grad_norm": 0.0,
      "learning_rate": 4.785948975647468e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 887
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.125,
      "epoch": 0.042906841901816775,
      "grad_norm": 0.0,
      "learning_rate": 4.785707383069192e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 888
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 96.875,
      "epoch": 0.04295516041747197,
      "grad_norm": 24.922993001547518,
      "learning_rate": 4.785465790490915e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.4401872754096985,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 889
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 110.15625,
      "epoch": 0.04300347893312718,
      "grad_norm": 4.409583582407591,
      "learning_rate": 4.78522419791264e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 890
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 102.84375,
      "epoch": 0.043051797448782375,
      "grad_norm": 4.883488787368306,
      "learning_rate": 4.784982605334364e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.12387890368700027,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 891
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.71875,
      "epoch": 0.04310011596443757,
      "grad_norm": 0.0,
      "learning_rate": 4.784741012756088e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 892
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.40625,
      "epoch": 0.04314843448009277,
      "grad_norm": 8.697131042900011,
      "learning_rate": 4.784499420177812e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 893
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 103.34375,
      "epoch": 0.04319675299574797,
      "grad_norm": 6.417157448623974,
      "learning_rate": 4.784257827599536e-07,
      "loss": -0.0,
      "reward": 1.421875,
      "reward_std": 0.20633062720298767,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 894
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 108.5,
      "epoch": 0.04324507151140317,
      "grad_norm": 13.690241403666711,
      "learning_rate": 4.78401623502126e-07,
      "loss": -0.0,
      "reward": 1.234375,
      "reward_std": 0.3676722049713135,
      "rewards/accuracy_reward": 0.15625,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 895
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 106.96875,
      "epoch": 0.04329339002705837,
      "grad_norm": 31.998181238718693,
      "learning_rate": 4.783774642442983e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.3831920325756073,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 896
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 111.8125,
      "epoch": 0.04334170854271357,
      "grad_norm": 8.570547082450757,
      "learning_rate": 4.783533049864708e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 897
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 113.34375,
      "epoch": 0.043390027058368766,
      "grad_norm": 6.254496956574186,
      "learning_rate": 4.783291457286431e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 898
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.9375,
      "epoch": 0.043438345574023964,
      "grad_norm": 0.0,
      "learning_rate": 4.783049864708156e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 899
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 94.0,
      "epoch": 0.04348666408967917,
      "grad_norm": 11.778631773430671,
      "learning_rate": 4.78280827212988e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.21316051483154297,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 900
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.3125,
      "epoch": 0.043534982605334366,
      "grad_norm": 0.0,
      "learning_rate": 4.782566679551604e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 901
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 108.0625,
      "epoch": 0.043583301120989563,
      "grad_norm": 62.863730481502785,
      "learning_rate": 4.782325086973328e-07,
      "loss": 0.0,
      "reward": 1.453125,
      "reward_std": 0.3618338406085968,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 902
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 100.75,
      "epoch": 0.04363161963664476,
      "grad_norm": 6.227494153758416,
      "learning_rate": 4.782083494395052e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 903
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 103.875,
      "epoch": 0.04367993815229996,
      "grad_norm": 11.494116569264834,
      "learning_rate": 4.781841901816776e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.3740372061729431,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 904
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.46875,
      "epoch": 0.043728256667955163,
      "grad_norm": 0.0,
      "learning_rate": 4.7816003092385e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 905
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 104.5,
      "epoch": 0.04377657518361036,
      "grad_norm": 12.436447190523102,
      "learning_rate": 4.781358716660223e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.33090677857398987,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 906
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 102.40625,
      "epoch": 0.04382489369926556,
      "grad_norm": 14.160108069065108,
      "learning_rate": 4.781117124081948e-07,
      "loss": -0.0,
      "reward": 1.359375,
      "reward_std": 0.27785414457321167,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 907
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.90625,
      "epoch": 0.043873212214920757,
      "grad_norm": 0.0,
      "learning_rate": 4.780875531503672e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 908
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 94.28125,
      "epoch": 0.043921530730575954,
      "grad_norm": 7.337228901705353,
      "learning_rate": 4.780633938925396e-07,
      "loss": -0.0,
      "reward": 2.34375,
      "reward_std": 0.31300365924835205,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 909
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.65625,
      "epoch": 0.04396984924623116,
      "grad_norm": 0.0,
      "learning_rate": 4.78039234634712e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 910
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.4375,
      "epoch": 0.044018167761886356,
      "grad_norm": 0.0,
      "learning_rate": 4.780150753768844e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 911
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 99.90625,
      "epoch": 0.044066486277541554,
      "grad_norm": 21.529036184807655,
      "learning_rate": 4.779909161190568e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.44338250160217285,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 912
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 94.84375,
      "epoch": 0.04411480479319675,
      "grad_norm": 10.849444631550668,
      "learning_rate": 4.779667568612292e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.3051002323627472,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 913
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 92.5,
      "epoch": 0.04416312330885195,
      "grad_norm": 7.103800509452839,
      "learning_rate": 4.779425976034017e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 914
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 101.3125,
      "epoch": 0.044211441824507154,
      "grad_norm": 7.3704855511504865,
      "learning_rate": 4.77918438345574e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 915
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.09375,
      "epoch": 0.04425976034016235,
      "grad_norm": 0.0,
      "learning_rate": 4.778942790877464e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 916
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 87.96875,
      "epoch": 0.04430807885581755,
      "grad_norm": 9.737624321810026,
      "learning_rate": 4.778701198299188e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.3945523202419281,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 917
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 98.84375,
      "epoch": 0.04435639737147275,
      "grad_norm": 29.0368672500991,
      "learning_rate": 4.778459605720912e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.4411507844924927,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 918
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 103.09375,
      "epoch": 0.044404715887127945,
      "grad_norm": 13.313123473968824,
      "learning_rate": 4.778218013142636e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 919
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.65625,
      "epoch": 0.04445303440278315,
      "grad_norm": 0.0,
      "learning_rate": 4.77797642056436e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 920
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.0,
      "epoch": 0.04450135291843835,
      "grad_norm": 0.0,
      "learning_rate": 4.777734827986084e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 921
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.28125,
      "epoch": 0.044549671434093545,
      "grad_norm": 33.22533959580156,
      "learning_rate": 4.777493235407808e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.26409149169921875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 922
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.15625,
      "epoch": 0.04459798994974874,
      "grad_norm": 0.0,
      "learning_rate": 4.777251642829533e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 923
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 89.3125,
      "epoch": 0.04464630846540394,
      "grad_norm": 10.359780719851013,
      "learning_rate": 4.777010050251256e-07,
      "loss": -0.0,
      "reward": 1.5625,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 924
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 88.96875,
      "epoch": 0.044694626981059145,
      "grad_norm": 15.432617160610157,
      "learning_rate": 4.77676845767298e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.4592205286026001,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 925
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.09375,
      "epoch": 0.04474294549671434,
      "grad_norm": 0.0,
      "learning_rate": 4.776526865094704e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 926
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.375,
      "epoch": 0.04479126401236954,
      "grad_norm": 9.670055490671993,
      "learning_rate": 4.776285272516428e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.21316051483154297,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 927
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 98.8125,
      "epoch": 0.04483958252802474,
      "grad_norm": 6.802600285286002,
      "learning_rate": 4.776043679938152e-07,
      "loss": 0.0,
      "reward": 1.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.03125,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 928
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.21875,
      "epoch": 0.044887901043679936,
      "grad_norm": 0.0,
      "learning_rate": 4.775802087359875e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 929
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.0625,
      "epoch": 0.04493621955933514,
      "grad_norm": 0.0,
      "learning_rate": 4.775560494781601e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 930
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 89.71875,
      "epoch": 0.04498453807499034,
      "grad_norm": 8.716730734773591,
      "learning_rate": 4.775318902203324e-07,
      "loss": 0.0,
      "reward": 1.21875,
      "reward_std": 0.18600594997406006,
      "rewards/accuracy_reward": 0.15625,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 931
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.5,
      "epoch": 0.045032856590645536,
      "grad_norm": 0.0,
      "learning_rate": 4.775077309625048e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 932
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.65625,
      "epoch": 0.04508117510630073,
      "grad_norm": 0.0,
      "learning_rate": 4.774835717046772e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 933
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 87.15625,
      "epoch": 0.04512949362195593,
      "grad_norm": 17.60719656533872,
      "learning_rate": 4.774594124468496e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.5681338310241699,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 934
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 78.90625,
      "epoch": 0.045177812137611136,
      "grad_norm": 18.717411445475896,
      "learning_rate": 4.77435253189022e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 935
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.34375,
      "epoch": 0.04522613065326633,
      "grad_norm": 95.75505162665814,
      "learning_rate": 4.774110939311944e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 936
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.40625,
      "epoch": 0.04527444916892153,
      "grad_norm": 6.1278629718194875,
      "learning_rate": 4.773869346733669e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.20348434150218964,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 937
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 80.53125,
      "epoch": 0.04532276768457673,
      "grad_norm": 9.055913407167154,
      "learning_rate": 4.773627754155392e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 938
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 85.0625,
      "epoch": 0.045371086200231926,
      "grad_norm": 13.697212202208142,
      "learning_rate": 4.773386161577116e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 939
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.84375,
      "epoch": 0.04541940471588713,
      "grad_norm": 0.0,
      "learning_rate": 4.77314456899884e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 940
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.96875,
      "epoch": 0.04546772323154233,
      "grad_norm": 0.0,
      "learning_rate": 4.772902976420564e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 941
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.03125,
      "epoch": 0.045516041747197526,
      "grad_norm": 0.0,
      "learning_rate": 4.772661383842288e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 942
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.71875,
      "epoch": 0.045564360262852724,
      "grad_norm": 0.0,
      "learning_rate": 4.772419791264012e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 943
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 88.21875,
      "epoch": 0.04561267877850792,
      "grad_norm": 21.182503943162757,
      "learning_rate": 4.772178198685736e-07,
      "loss": -0.0,
      "reward": 1.75,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 944
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.21875,
      "epoch": 0.045660997294163126,
      "grad_norm": 0.0,
      "learning_rate": 4.771936606107461e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 945
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.0625,
      "epoch": 0.045709315809818324,
      "grad_norm": 4.917186715068007,
      "learning_rate": 4.771695013529184e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 946
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 87.0,
      "epoch": 0.04575763432547352,
      "grad_norm": 18.98164721430703,
      "learning_rate": 4.771453420950909e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.3360668420791626,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 947
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.3125,
      "epoch": 0.04580595284112872,
      "grad_norm": 0.0,
      "learning_rate": 4.771211828372632e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 948
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.96875,
      "epoch": 0.04585427135678392,
      "grad_norm": 0.0,
      "learning_rate": 4.770970235794356e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 949
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.84375,
      "epoch": 0.04590258987243912,
      "grad_norm": 6.869657120006914,
      "learning_rate": 4.77072864321608e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 950
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.78125,
      "epoch": 0.04595090838809432,
      "grad_norm": 0.0,
      "learning_rate": 4.770487050637804e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 951
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 90.25,
      "epoch": 0.04599922690374952,
      "grad_norm": 11.527342287313806,
      "learning_rate": 4.770245458059528e-07,
      "loss": 0.0,
      "reward": 1.28125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 952
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 80.09375,
      "epoch": 0.046047545419404715,
      "grad_norm": 27.774043992270173,
      "learning_rate": 4.770003865481253e-07,
      "loss": -0.0,
      "reward": 2.328125,
      "reward_std": 0.2662131190299988,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 953
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.34375,
      "epoch": 0.04609586393505991,
      "grad_norm": 0.0,
      "learning_rate": 4.769762272902977e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 954
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 97.125,
      "epoch": 0.04614418245071512,
      "grad_norm": 13.512967421989819,
      "learning_rate": 4.7695206803247e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.39860397577285767,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 955
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.875,
      "epoch": 0.046192500966370315,
      "grad_norm": 208.7534697007393,
      "learning_rate": 4.769279087746424e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.21366669237613678,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 956
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 90.75,
      "epoch": 0.04624081948202551,
      "grad_norm": 21.657549856680205,
      "learning_rate": 4.769037495168148e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.5673428773880005,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 957
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.78125,
      "epoch": 0.04628913799768071,
      "grad_norm": 0.0,
      "learning_rate": 4.768795902589872e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 958
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 92.625,
      "epoch": 0.04633745651333591,
      "grad_norm": 12.859589355547163,
      "learning_rate": 4.768554310011596e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.5092827081680298,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 959
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.59375,
      "epoch": 0.04638577502899111,
      "grad_norm": 0.0,
      "learning_rate": 4.7683127174333205e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 960
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.1875,
      "epoch": 0.04643409354464631,
      "grad_norm": 23.97016001998279,
      "learning_rate": 4.768071124855044e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.24280104041099548,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 961
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 96.4375,
      "epoch": 0.04648241206030151,
      "grad_norm": 14.787778915897661,
      "learning_rate": 4.767829532276768e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 962
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.34375,
      "epoch": 0.046530730575956705,
      "grad_norm": 0.0,
      "learning_rate": 4.7675879396984924e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 963
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 88.21875,
      "epoch": 0.0465790490916119,
      "grad_norm": 20.559626045782746,
      "learning_rate": 4.7673463471202165e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.36093956232070923,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 964
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.71875,
      "epoch": 0.04662736760726711,
      "grad_norm": 0.0,
      "learning_rate": 4.76710475454194e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 965
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.125,
      "epoch": 0.046675686122922305,
      "grad_norm": 9.057256907836425,
      "learning_rate": 4.766863161963664e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 966
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 90.71875,
      "epoch": 0.0467240046385775,
      "grad_norm": 9.477861926788112,
      "learning_rate": 4.7666215693853884e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.3998437523841858,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 967
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.71875,
      "epoch": 0.0467723231542327,
      "grad_norm": 0.0,
      "learning_rate": 4.766379976807112e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 968
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 80.46875,
      "epoch": 0.0468206416698879,
      "grad_norm": 14.99128285882874,
      "learning_rate": 4.766138384228836e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.35303613543510437,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 969
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.71875,
      "epoch": 0.0468689601855431,
      "grad_norm": 0.0,
      "learning_rate": 4.76589679165056e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 970
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.75,
      "epoch": 0.0469172787011983,
      "grad_norm": 0.0,
      "learning_rate": 4.7656551990722844e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 971
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 95.4375,
      "epoch": 0.0469655972168535,
      "grad_norm": 14.971375732423247,
      "learning_rate": 4.765413606494008e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.3230288028717041,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 972
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.8125,
      "epoch": 0.047013915732508696,
      "grad_norm": 6.498317185120281,
      "learning_rate": 4.7651720139157327e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 973
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.90625,
      "epoch": 0.047062234248163894,
      "grad_norm": 0.0,
      "learning_rate": 4.764930421337456e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 974
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.03125,
      "epoch": 0.0471105527638191,
      "grad_norm": 0.0,
      "learning_rate": 4.7646888287591804e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 975
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 87.34375,
      "epoch": 0.047158871279474296,
      "grad_norm": 14.481040081846455,
      "learning_rate": 4.764447236180904e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 976
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.46875,
      "epoch": 0.047207189795129494,
      "grad_norm": 4.13753914264847,
      "learning_rate": 4.7642056436026287e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 977
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 86.875,
      "epoch": 0.04725550831078469,
      "grad_norm": 15.822012531804274,
      "learning_rate": 4.7639640510243523e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.4311256408691406,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 978
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.34375,
      "epoch": 0.04730382682643989,
      "grad_norm": 11.841201539870733,
      "learning_rate": 4.763722458446076e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.31663835048675537,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 979
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 88.53125,
      "epoch": 0.047352145342095094,
      "grad_norm": 16.245732901551335,
      "learning_rate": 4.7634808658678006e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.37086743116378784,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 980
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 89.96875,
      "epoch": 0.04740046385775029,
      "grad_norm": 8.888904669311435,
      "learning_rate": 4.763239273289524e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.21316051483154297,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 981
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.5,
      "epoch": 0.04744878237340549,
      "grad_norm": 13.330086879058538,
      "learning_rate": 4.7629976807112483e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 982
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 93.46875,
      "epoch": 0.04749710088906069,
      "grad_norm": 10.243077630490173,
      "learning_rate": 4.7627560881329724e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 983
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.59375,
      "epoch": 0.047545419404715884,
      "grad_norm": 0.0,
      "learning_rate": 4.7625144955546966e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 984
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.1875,
      "epoch": 0.04759373792037109,
      "grad_norm": 10.850350573116865,
      "learning_rate": 4.76227290297642e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 985
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 84.625,
      "epoch": 0.04764205643602629,
      "grad_norm": 12.58778479493184,
      "learning_rate": 4.7620313103981443e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.3913571238517761,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 986
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.15625,
      "epoch": 0.047690374951681484,
      "grad_norm": 0.0,
      "learning_rate": 4.7617897178198684e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 987
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.0,
      "epoch": 0.04773869346733668,
      "grad_norm": 3.9380821987933077,
      "learning_rate": 4.7615481252415926e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.2993341088294983,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 988
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.96875,
      "epoch": 0.04778701198299188,
      "grad_norm": 12.934966203921656,
      "learning_rate": 4.761306532663316e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.3187600076198578,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 989
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 92.09375,
      "epoch": 0.047835330498647084,
      "grad_norm": 10.885953696055651,
      "learning_rate": 4.7610649400850403e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.378745436668396,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 990
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.09375,
      "epoch": 0.04788364901430228,
      "grad_norm": 0.0,
      "learning_rate": 4.7608233475067645e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 991
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 98.40625,
      "epoch": 0.04793196752995748,
      "grad_norm": 18.483912610373284,
      "learning_rate": 4.760581754928488e-07,
      "loss": -0.0,
      "reward": 1.15625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.125,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 992
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.1875,
      "epoch": 0.04798028604561268,
      "grad_norm": 0.0,
      "learning_rate": 4.7603401623502127e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 993
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.9375,
      "epoch": 0.048028604561267875,
      "grad_norm": 13.362494663117106,
      "learning_rate": 4.7600985697719363e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.3255898952484131,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 994
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 93.9375,
      "epoch": 0.04807692307692308,
      "grad_norm": 15.963837873926355,
      "learning_rate": 4.7598569771936605e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.35307908058166504,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 995
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.96875,
      "epoch": 0.04812524159257828,
      "grad_norm": 0.0,
      "learning_rate": 4.759615384615384e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 996
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 90.0625,
      "epoch": 0.048173560108233475,
      "grad_norm": 23.164800815631054,
      "learning_rate": 4.759373792037109e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.3945523202419281,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 997
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 107.3125,
      "epoch": 0.04822187862388867,
      "grad_norm": 6.590468989334676,
      "learning_rate": 4.7591321994588324e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 998
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.46875,
      "epoch": 0.04827019713954387,
      "grad_norm": 0.0,
      "learning_rate": 4.758890606880556e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 999
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.53125,
      "epoch": 0.048318515655199075,
      "grad_norm": 4.481694006429593,
      "learning_rate": 4.7586490143022806e-07,
      "loss": -0.0,
      "reward": 2.40625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 1000
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.1875,
      "epoch": 0.04836683417085427,
      "grad_norm": 0.0,
      "learning_rate": 4.758407421724004e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1001
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.40625,
      "epoch": 0.04841515268650947,
      "grad_norm": 8.504185310492169,
      "learning_rate": 4.7581658291457284e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1002
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 98.21875,
      "epoch": 0.04846347120216467,
      "grad_norm": 10.596599441013291,
      "learning_rate": 4.7579242365674525e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.4537375569343567,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1003
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 89.03125,
      "epoch": 0.048511789717819866,
      "grad_norm": 46.49679422372622,
      "learning_rate": 4.7576826439891766e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.4535670876502991,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1004
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.65625,
      "epoch": 0.04856010823347507,
      "grad_norm": 0.0,
      "learning_rate": 4.7574410514109e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1005
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.84375,
      "epoch": 0.04860842674913027,
      "grad_norm": 0.0,
      "learning_rate": 4.7571994588326244e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1006
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.1875,
      "epoch": 0.048656745264785466,
      "grad_norm": 0.0,
      "learning_rate": 4.7569578662543485e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1007
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.34375,
      "epoch": 0.04870506378044066,
      "grad_norm": 15.256357716952682,
      "learning_rate": 4.7567162736760727e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1008
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.40625,
      "epoch": 0.04875338229609586,
      "grad_norm": 0.0,
      "learning_rate": 4.7564746810977963e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1009
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 98.125,
      "epoch": 0.048801700811751066,
      "grad_norm": 11.68397305957899,
      "learning_rate": 4.7562330885195204e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.34624922275543213,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1010
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.375,
      "epoch": 0.04885001932740626,
      "grad_norm": 0.0,
      "learning_rate": 4.7559914959412445e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1011
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.40625,
      "epoch": 0.04889833784306146,
      "grad_norm": 0.0,
      "learning_rate": 4.755749903362968e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1012
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.15625,
      "epoch": 0.04894665635871666,
      "grad_norm": 4.20012383079448,
      "learning_rate": 4.755508310784693e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.260896235704422,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1013
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.875,
      "epoch": 0.048994974874371856,
      "grad_norm": 0.0,
      "learning_rate": 4.7552667182064164e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1014
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.0625,
      "epoch": 0.04904329339002706,
      "grad_norm": 0.0,
      "learning_rate": 4.7550251256281406e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1015
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 88.75,
      "epoch": 0.04909161190568226,
      "grad_norm": 8.254582393890862,
      "learning_rate": 4.754783533049864e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1016
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.09375,
      "epoch": 0.049139930421337456,
      "grad_norm": 0.0,
      "learning_rate": 4.754541940471589e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1017
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 95.125,
      "epoch": 0.049188248936992654,
      "grad_norm": 5.032091526492167,
      "learning_rate": 4.7543003478933124e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.31512534618377686,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1018
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.15625,
      "epoch": 0.04923656745264785,
      "grad_norm": 0.0,
      "learning_rate": 4.754058755315036e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1019
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.96875,
      "epoch": 0.049284885968303056,
      "grad_norm": 0.0,
      "learning_rate": 4.7538171627367607e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1020
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.0,
      "epoch": 0.049333204483958254,
      "grad_norm": 5.676129250249802,
      "learning_rate": 4.7535755701584843e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1021
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 90.21875,
      "epoch": 0.04938152299961345,
      "grad_norm": 15.302071945213521,
      "learning_rate": 4.7533339775802085e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.3230288028717041,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1022
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 91.09375,
      "epoch": 0.04942984151526865,
      "grad_norm": 7.947828055633014,
      "learning_rate": 4.7530923850019326e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1023
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.84375,
      "epoch": 0.04947816003092385,
      "grad_norm": 4.979743008310121,
      "learning_rate": 4.7528507924236567e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1024
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 94.0625,
      "epoch": 0.04952647854657905,
      "grad_norm": 17.361162589575752,
      "learning_rate": 4.7526091998453803e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.37405285239219666,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1025
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 91.375,
      "epoch": 0.04957479706223425,
      "grad_norm": 9.076032548093096,
      "learning_rate": 4.7523676072671045e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1026
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.9375,
      "epoch": 0.04962311557788945,
      "grad_norm": 0.0,
      "learning_rate": 4.7521260146888286e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1027
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 96.09375,
      "epoch": 0.049671434093544645,
      "grad_norm": 4.300397576667233,
      "learning_rate": 4.751884422110553e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1028
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.375,
      "epoch": 0.04971975260919984,
      "grad_norm": 0.0,
      "learning_rate": 4.7516428295322763e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1029
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 95.65625,
      "epoch": 0.04976807112485505,
      "grad_norm": 6.010551593411472,
      "learning_rate": 4.7514012369540005e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1030
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.84375,
      "epoch": 0.049816389640510245,
      "grad_norm": 0.0,
      "learning_rate": 4.7511596443757246e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1031
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.09375,
      "epoch": 0.04986470815616544,
      "grad_norm": 0.0,
      "learning_rate": 4.750918051797448e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1032
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.5,
      "epoch": 0.04991302667182064,
      "grad_norm": 0.0,
      "learning_rate": 4.750676459219173e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1033
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.03125,
      "epoch": 0.04996134518747584,
      "grad_norm": 0.0,
      "learning_rate": 4.7504348666408965e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1034
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 100.53125,
      "epoch": 0.05000966370313104,
      "grad_norm": 9.495890597661782,
      "learning_rate": 4.7501932740626206e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.37086743116378784,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1035
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 89.03125,
      "epoch": 0.05005798221878624,
      "grad_norm": 9.55437318102908,
      "learning_rate": 4.749951681484344e-07,
      "loss": -0.0,
      "reward": 1.671875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1036
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 95.8125,
      "epoch": 0.05010630073444144,
      "grad_norm": 17.534335012108922,
      "learning_rate": 4.749710088906069e-07,
      "loss": -0.0,
      "reward": 1.5625,
      "reward_std": 0.22049658000469208,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1037
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 93.96875,
      "epoch": 0.050154619250096635,
      "grad_norm": 6.916830251062474,
      "learning_rate": 4.7494684963277925e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.32878512144088745,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1038
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.65625,
      "epoch": 0.05020293776575183,
      "grad_norm": 9.061771328638113,
      "learning_rate": 4.7492269037495167e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1039
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.96875,
      "epoch": 0.05025125628140704,
      "grad_norm": 0.0,
      "learning_rate": 4.748985311171241e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1040
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.53125,
      "epoch": 0.050299574797062235,
      "grad_norm": 0.0,
      "learning_rate": 4.7487437185929644e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 1041
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 97.625,
      "epoch": 0.05034789331271743,
      "grad_norm": 12.302103726203384,
      "learning_rate": 4.7485021260146885e-07,
      "loss": -0.0,
      "reward": 1.359375,
      "reward_std": 0.27785414457321167,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 1042
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.1875,
      "epoch": 0.05039621182837263,
      "grad_norm": 0.0,
      "learning_rate": 4.7482605334364127e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1043
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.21875,
      "epoch": 0.05044453034402783,
      "grad_norm": 6.448578673215318,
      "learning_rate": 4.748018940858137e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1044
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 93.09375,
      "epoch": 0.05049284885968303,
      "grad_norm": 9.699868148229948,
      "learning_rate": 4.7477773482798604e-07,
      "loss": 0.0,
      "reward": 1.5,
      "reward_std": 0.26726123690605164,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1045
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 84.59375,
      "epoch": 0.05054116737533823,
      "grad_norm": 6.799773704802736,
      "learning_rate": 4.7475357557015845e-07,
      "loss": -0.0,
      "reward": 1.40625,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 1046
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.78125,
      "epoch": 0.05058948589099343,
      "grad_norm": 0.0,
      "learning_rate": 4.7472941631233087e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1047
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 82.3125,
      "epoch": 0.050637804406648626,
      "grad_norm": 19.621567669896258,
      "learning_rate": 4.747052570545033e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.32732683420181274,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1048
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.15625,
      "epoch": 0.050686122922303824,
      "grad_norm": 5.522668214300306,
      "learning_rate": 4.7468109779667564e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1049
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.28125,
      "epoch": 0.05073444143795903,
      "grad_norm": 0.0,
      "learning_rate": 4.746569385388481e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 1050
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 98.34375,
      "epoch": 0.050782759953614226,
      "grad_norm": 22.06428722543093,
      "learning_rate": 4.7463277928102047e-07,
      "loss": 0.0,
      "reward": 1.546875,
      "reward_std": 0.3998437523841858,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1051
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 80.4375,
      "epoch": 0.050831078469269424,
      "grad_norm": 19.254998793385276,
      "learning_rate": 4.7460862002319283e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1052
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.96875,
      "epoch": 0.05087939698492462,
      "grad_norm": 8.63917317588581,
      "learning_rate": 4.745844607653653e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1053
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.40625,
      "epoch": 0.05092771550057982,
      "grad_norm": 68.8034465114792,
      "learning_rate": 4.7456030150753766e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1054
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.03125,
      "epoch": 0.050976034016235024,
      "grad_norm": 0.0,
      "learning_rate": 4.7453614224971007e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1055
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 89.875,
      "epoch": 0.05102435253189022,
      "grad_norm": 15.63097525182494,
      "learning_rate": 4.7451198299188243e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.5670700669288635,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1056
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.15625,
      "epoch": 0.05107267104754542,
      "grad_norm": 0.0,
      "learning_rate": 4.744878237340549e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1057
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.125,
      "epoch": 0.05112098956320062,
      "grad_norm": 0.0,
      "learning_rate": 4.7446366447622726e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1058
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 76.6875,
      "epoch": 0.051169308078855814,
      "grad_norm": 5.8483717024077135,
      "learning_rate": 4.7443950521839967e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1059
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 92.03125,
      "epoch": 0.05121762659451102,
      "grad_norm": 10.618105522510014,
      "learning_rate": 4.744153459605721e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1060
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.9375,
      "epoch": 0.05126594511016622,
      "grad_norm": 0.0,
      "learning_rate": 4.7439118670274445e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1061
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.40625,
      "epoch": 0.051314263625821414,
      "grad_norm": 0.0,
      "learning_rate": 4.7436702744491686e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1062
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.625,
      "epoch": 0.05136258214147661,
      "grad_norm": 0.0,
      "learning_rate": 4.743428681870893e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2630178928375244,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1063
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.90625,
      "epoch": 0.05141090065713181,
      "grad_norm": 0.0,
      "learning_rate": 4.743187089292617e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1064
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 92.5625,
      "epoch": 0.051459219172787014,
      "grad_norm": 24.06732842937342,
      "learning_rate": 4.7429454967143405e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.32327672839164734,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1065
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.46875,
      "epoch": 0.05150753768844221,
      "grad_norm": 0.0,
      "learning_rate": 4.7427039041360646e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1066
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 85.75,
      "epoch": 0.05155585620409741,
      "grad_norm": 7.566994972318919,
      "learning_rate": 4.742462311557789e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1067
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.78125,
      "epoch": 0.05160417471975261,
      "grad_norm": 6.460645383159783,
      "learning_rate": 4.742220718979513e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1068
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 80.75,
      "epoch": 0.051652493235407805,
      "grad_norm": 15.267282459515924,
      "learning_rate": 4.7419791264012365e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1069
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 88.5625,
      "epoch": 0.05170081175106301,
      "grad_norm": 28.354863154585107,
      "learning_rate": 4.741737533822961e-07,
      "loss": -0.0,
      "reward": 1.46875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1070
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 80.8125,
      "epoch": 0.05174913026671821,
      "grad_norm": 13.338851960107108,
      "learning_rate": 4.741495941244685e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1071
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.3125,
      "epoch": 0.051797448782373405,
      "grad_norm": 0.0,
      "learning_rate": 4.7412543486664084e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1072
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.34375,
      "epoch": 0.0518457672980286,
      "grad_norm": 0.0,
      "learning_rate": 4.741012756088133e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1073
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 90.46875,
      "epoch": 0.0518940858136838,
      "grad_norm": 7.953203357445116,
      "learning_rate": 4.7407711635098567e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1074
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 86.71875,
      "epoch": 0.051942404329339005,
      "grad_norm": 7.282892944550635,
      "learning_rate": 4.740529570931581e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1075
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 94.15625,
      "epoch": 0.0519907228449942,
      "grad_norm": 37.740870794890355,
      "learning_rate": 4.7402879783533044e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.6025131940841675,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1076
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.125,
      "epoch": 0.0520390413606494,
      "grad_norm": 0.0,
      "learning_rate": 4.740046385775029e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1077
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.5625,
      "epoch": 0.0520873598763046,
      "grad_norm": 0.0,
      "learning_rate": 4.7398047931967527e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1078
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 77.9375,
      "epoch": 0.052135678391959796,
      "grad_norm": 12.809753989450105,
      "learning_rate": 4.739563200618477e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.34624922275543213,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1079
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 84.3125,
      "epoch": 0.052183996907615,
      "grad_norm": 40.18125555206969,
      "learning_rate": 4.739321608040201e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.28838545083999634,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1080
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.53125,
      "epoch": 0.0522323154232702,
      "grad_norm": 6.2320390186747865,
      "learning_rate": 4.739080015461925e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.22049658000469208,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1081
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 67.25,
      "epoch": 0.052280633938925396,
      "grad_norm": 27.0887050710217,
      "learning_rate": 4.7388384228836487e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1082
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 78.4375,
      "epoch": 0.05232895245458059,
      "grad_norm": 39.659402424592315,
      "learning_rate": 4.738596830305373e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.4889746904373169,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1083
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.8125,
      "epoch": 0.05237727097023579,
      "grad_norm": 0.0,
      "learning_rate": 4.738355237727097e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1084
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 75.59375,
      "epoch": 0.052425589485890996,
      "grad_norm": 14.948604359916114,
      "learning_rate": 4.7381136451488206e-07,
      "loss": -0.0,
      "reward": 1.578125,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1085
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.28125,
      "epoch": 0.05247390800154619,
      "grad_norm": 9.17642533020604,
      "learning_rate": 4.7378720525705447e-07,
      "loss": -0.0,
      "reward": 2.109375,
      "reward_std": 0.2731459438800812,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1086
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.46875,
      "epoch": 0.05252222651720139,
      "grad_norm": 0.0,
      "learning_rate": 4.737630459992269e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1087
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.5,
      "epoch": 0.05257054503285659,
      "grad_norm": 9.643793195712247,
      "learning_rate": 4.737388867413993e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1088
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.3125,
      "epoch": 0.052618863548511786,
      "grad_norm": 14.927547364284088,
      "learning_rate": 4.7371472748357166e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.24067938327789307,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1089
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 78.5625,
      "epoch": 0.05266718206416699,
      "grad_norm": 16.08118018226924,
      "learning_rate": 4.736905682257441e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.22835706174373627,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1090
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.3125,
      "epoch": 0.05271550057982219,
      "grad_norm": 18.127351828619997,
      "learning_rate": 4.736664089679165e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1091
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.03125,
      "epoch": 0.052763819095477386,
      "grad_norm": 25.496730346784265,
      "learning_rate": 4.7364224971008885e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.3298332095146179,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1092
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 76.3125,
      "epoch": 0.052812137611132584,
      "grad_norm": 37.11620967987105,
      "learning_rate": 4.736180904522613e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1093
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.25,
      "epoch": 0.05286045612678778,
      "grad_norm": 7.338975530186529,
      "learning_rate": 4.7359393119443367e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.18600594997406006,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1094
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.625,
      "epoch": 0.052908774642442986,
      "grad_norm": 0.0,
      "learning_rate": 4.735697719366061e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1095
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 79.5,
      "epoch": 0.052957093158098184,
      "grad_norm": 21.339706551792393,
      "learning_rate": 4.7354561267877845e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.36306124925613403,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1096
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 89.375,
      "epoch": 0.05300541167375338,
      "grad_norm": 4.769752987185003,
      "learning_rate": 4.735214534209509e-07,
      "loss": 0.0,
      "reward": 1.40625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1097
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.71875,
      "epoch": 0.05305373018940858,
      "grad_norm": 0.0,
      "learning_rate": 4.734972941631233e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1098
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.3125,
      "epoch": 0.05310204870506378,
      "grad_norm": 0.0,
      "learning_rate": 4.734731349052957e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1099
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.75,
      "epoch": 0.05315036722071898,
      "grad_norm": 5.853178321793066,
      "learning_rate": 4.734489756474681e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1100
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 77.125,
      "epoch": 0.05319868573637418,
      "grad_norm": 8.124325015935199,
      "learning_rate": 4.734248163896405e-07,
      "loss": -0.0,
      "reward": 1.546875,
      "reward_std": 0.24464011192321777,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1101
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.03125,
      "epoch": 0.05324700425202938,
      "grad_norm": 0.0,
      "learning_rate": 4.734006571318129e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1102
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 83.9375,
      "epoch": 0.053295322767684575,
      "grad_norm": 8.084673715003296,
      "learning_rate": 4.733764978739853e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1103
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 84.0625,
      "epoch": 0.05334364128333977,
      "grad_norm": 3.926835035631467,
      "learning_rate": 4.733523386161577e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1104
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 76.84375,
      "epoch": 0.05339195979899498,
      "grad_norm": 7.352287888688086,
      "learning_rate": 4.7332817935833006e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1105
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 80.1875,
      "epoch": 0.053440278314650175,
      "grad_norm": 28.92301863527359,
      "learning_rate": 4.733040201005025e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.6082562208175659,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1106
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.0,
      "epoch": 0.05348859683030537,
      "grad_norm": 0.0,
      "learning_rate": 4.732798608426749e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1107
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 81.875,
      "epoch": 0.05353691534596057,
      "grad_norm": 10.477979365571176,
      "learning_rate": 4.732557015848473e-07,
      "loss": 0.0,
      "reward": 1.140625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.09375,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 1108
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 77.28125,
      "epoch": 0.05358523386161577,
      "grad_norm": 12.370366207127905,
      "learning_rate": 4.7323154232701967e-07,
      "loss": -0.0,
      "reward": 1.453125,
      "reward_std": 0.4398043751716614,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1109
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 86.71875,
      "epoch": 0.05363355237727097,
      "grad_norm": 18.34601901836026,
      "learning_rate": 4.7320738306919213e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.34988388419151306,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1110
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 85.9375,
      "epoch": 0.05368187089292617,
      "grad_norm": 4.420280093743492,
      "learning_rate": 4.731832238113645e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1111
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 85.5,
      "epoch": 0.05373018940858137,
      "grad_norm": 9.242421814790472,
      "learning_rate": 4.7315906455353685e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.4570988416671753,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1112
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.5,
      "epoch": 0.053778507924236565,
      "grad_norm": 12.60115526817454,
      "learning_rate": 4.731349052957093e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1113
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 69.5625,
      "epoch": 0.05382682643989176,
      "grad_norm": 11.124451498026307,
      "learning_rate": 4.731107460378817e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1114
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 85.03125,
      "epoch": 0.05387514495554697,
      "grad_norm": 23.48270505867365,
      "learning_rate": 4.730865867800541e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1115
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.71875,
      "epoch": 0.053923463471202165,
      "grad_norm": 6.876507971877479,
      "learning_rate": 4.7306242752222646e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1116
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.75,
      "epoch": 0.05397178198685736,
      "grad_norm": 0.0,
      "learning_rate": 4.730382682643989e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 1117
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.625,
      "epoch": 0.05402010050251256,
      "grad_norm": 42.94512485396039,
      "learning_rate": 4.730141090065713e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.38981863856315613,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1118
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.5,
      "epoch": 0.05406841901816776,
      "grad_norm": 0.0,
      "learning_rate": 4.729899497487437e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1119
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 81.34375,
      "epoch": 0.05411673753382296,
      "grad_norm": 8.925493663189561,
      "learning_rate": 4.729657904909161e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.36680150032043457,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1120
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 85.3125,
      "epoch": 0.05416505604947816,
      "grad_norm": 21.693759254412228,
      "learning_rate": 4.729416312330885e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.6328117847442627,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1121
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.875,
      "epoch": 0.05421337456513336,
      "grad_norm": 0.0,
      "learning_rate": 4.729174719752609e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1122
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 83.34375,
      "epoch": 0.054261693080788556,
      "grad_norm": 8.768535736448577,
      "learning_rate": 4.728933127174333e-07,
      "loss": -0.0,
      "reward": 1.515625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1123
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 86.71875,
      "epoch": 0.054310011596443754,
      "grad_norm": 13.59507339665833,
      "learning_rate": 4.728691534596057e-07,
      "loss": -0.0,
      "reward": 1.5625,
      "reward_std": 0.3471825420856476,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1124
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 86.09375,
      "epoch": 0.05435833011209896,
      "grad_norm": 108.3193824798332,
      "learning_rate": 4.7284499420177807e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.46244120597839355,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1125
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 67.875,
      "epoch": 0.054406648627754156,
      "grad_norm": 9.115068368866737,
      "learning_rate": 4.728208349439505e-07,
      "loss": -0.0,
      "reward": 2.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 1126
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 71.40625,
      "epoch": 0.054454967143409354,
      "grad_norm": 22.207442729131614,
      "learning_rate": 4.727966756861229e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.3255898952484131,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1127
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.8125,
      "epoch": 0.05450328565906455,
      "grad_norm": 0.0,
      "learning_rate": 4.727725164282953e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.260896235704422,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1128
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 78.8125,
      "epoch": 0.05455160417471975,
      "grad_norm": 13.321106466133458,
      "learning_rate": 4.727483571704677e-07,
      "loss": -0.0,
      "reward": 1.5625,
      "reward_std": 0.32666346430778503,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1129
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 94.0,
      "epoch": 0.054599922690374954,
      "grad_norm": 14.079664091544457,
      "learning_rate": 4.7272419791264014e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.36722296476364136,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1130
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 79.3125,
      "epoch": 0.05464824120603015,
      "grad_norm": 18.965978215547974,
      "learning_rate": 4.727000386548125e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.405046284198761,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1131
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 68.96875,
      "epoch": 0.05469655972168535,
      "grad_norm": 36.21199461648481,
      "learning_rate": 4.726758793969849e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.22152717411518097,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1132
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.46875,
      "epoch": 0.05474487823734055,
      "grad_norm": 9.27424198780814,
      "learning_rate": 4.7265172013915733e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.26726123690605164,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1133
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 77.96875,
      "epoch": 0.05479319675299575,
      "grad_norm": 9.148414156125455,
      "learning_rate": 4.726275608813297e-07,
      "loss": -0.0,
      "reward": 1.921875,
      "reward_std": 0.29747018218040466,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1134
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 91.8125,
      "epoch": 0.05484151526865095,
      "grad_norm": 23.706140441032787,
      "learning_rate": 4.726034016235021e-07,
      "loss": -0.0,
      "reward": 1.328125,
      "reward_std": 0.3945268988609314,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1135
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.0625,
      "epoch": 0.05488983378430615,
      "grad_norm": 18.153013449501636,
      "learning_rate": 4.7257924236567446e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.2982703447341919,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1136
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 76.78125,
      "epoch": 0.054938152299961344,
      "grad_norm": 42.41009273555504,
      "learning_rate": 4.7255508310784693e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.5886077880859375,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1137
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 76.5625,
      "epoch": 0.05498647081561654,
      "grad_norm": 7.61543411678296,
      "learning_rate": 4.725309238500193e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.21316051483154297,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1138
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 77.71875,
      "epoch": 0.05503478933127175,
      "grad_norm": 19.833570847278175,
      "learning_rate": 4.725067645921917e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.3913571238517761,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1139
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.0625,
      "epoch": 0.055083107846926944,
      "grad_norm": 15.189183035277726,
      "learning_rate": 4.724826053343641e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1140
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.8125,
      "epoch": 0.05513142636258214,
      "grad_norm": 0.0,
      "learning_rate": 4.7245844607653653e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1141
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 74.90625,
      "epoch": 0.05517974487823734,
      "grad_norm": 13.192439210860819,
      "learning_rate": 4.724342868187089e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.434487521648407,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1142
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 80.15625,
      "epoch": 0.05522806339389254,
      "grad_norm": 10.767536675388872,
      "learning_rate": 4.7241012756088136e-07,
      "loss": -0.0,
      "reward": 1.125,
      "reward_std": 0.28104937076568604,
      "rewards/accuracy_reward": 0.09375,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 1143
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.375,
      "epoch": 0.05527638190954774,
      "grad_norm": 9.63376263036863,
      "learning_rate": 4.723859683030537e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.3319803476333618,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1144
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.4375,
      "epoch": 0.05532470042520294,
      "grad_norm": 10.90469730248068,
      "learning_rate": 4.723618090452261e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1145
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 76.65625,
      "epoch": 0.05537301894085814,
      "grad_norm": 8.611345157782,
      "learning_rate": 4.723376497873985e-07,
      "loss": 0.0,
      "reward": 2.1875,
      "reward_std": 0.30509042739868164,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1146
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 76.625,
      "epoch": 0.055421337456513335,
      "grad_norm": 23.11895709101471,
      "learning_rate": 4.723134905295709e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.42641252279281616,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1147
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.53125,
      "epoch": 0.05546965597216853,
      "grad_norm": 0.0,
      "learning_rate": 4.722893312717433e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1148
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.71875,
      "epoch": 0.05551797448782374,
      "grad_norm": 0.0,
      "learning_rate": 4.722651720139157e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1149
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 86.6875,
      "epoch": 0.055566293003478935,
      "grad_norm": 20.662725079007824,
      "learning_rate": 4.7224101275608815e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.57025545835495,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1150
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 85.09375,
      "epoch": 0.05561461151913413,
      "grad_norm": 15.114496001072173,
      "learning_rate": 4.722168534982605e-07,
      "loss": -0.0,
      "reward": 1.46875,
      "reward_std": 0.34988388419151306,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1151
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 76.28125,
      "epoch": 0.05566293003478933,
      "grad_norm": 47.73857966967533,
      "learning_rate": 4.721926942404329e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.37538355588912964,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1152
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.40625,
      "epoch": 0.05571124855044453,
      "grad_norm": 8.670537329554797,
      "learning_rate": 4.7216853498260534e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1153
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.53125,
      "epoch": 0.05575956706609973,
      "grad_norm": 0.0,
      "learning_rate": 4.721443757247777e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1154
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.125,
      "epoch": 0.05580788558175493,
      "grad_norm": 46.19455328426678,
      "learning_rate": 4.721202164669501e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.15580293536186218,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1155
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.65625,
      "epoch": 0.05585620409741013,
      "grad_norm": 5.3396688659864,
      "learning_rate": 4.7209605720912247e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1156
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.5625,
      "epoch": 0.055904522613065326,
      "grad_norm": 18.069778964569135,
      "learning_rate": 4.7207189795129494e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1157
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.21875,
      "epoch": 0.055952841128720524,
      "grad_norm": 0.0,
      "learning_rate": 4.720477386934673e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1158
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.75,
      "epoch": 0.05600115964437573,
      "grad_norm": 8.645322600946272,
      "learning_rate": 4.720235794356397e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1159
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.125,
      "epoch": 0.056049478160030926,
      "grad_norm": 5.072021675314164,
      "learning_rate": 4.719994201778121e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1160
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 81.53125,
      "epoch": 0.056097796675686123,
      "grad_norm": 42.584267723255905,
      "learning_rate": 4.7197526091998454e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.33725470304489136,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1161
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 76.03125,
      "epoch": 0.05614611519134132,
      "grad_norm": 44.20382445342338,
      "learning_rate": 4.719511016621569e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.3074946999549866,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1162
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 82.4375,
      "epoch": 0.05619443370699652,
      "grad_norm": 17.10800659372526,
      "learning_rate": 4.7192694240432937e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.40044301748275757,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1163
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 80.15625,
      "epoch": 0.056242752222651723,
      "grad_norm": 23.68657738562088,
      "learning_rate": 4.7190278314650173e-07,
      "loss": -0.0,
      "reward": 1.359375,
      "reward_std": 0.5225040912628174,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1164
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 86.8125,
      "epoch": 0.05629107073830692,
      "grad_norm": 11.872405660911333,
      "learning_rate": 4.718786238886741e-07,
      "loss": -0.0,
      "reward": 1.484375,
      "reward_std": 0.36162739992141724,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1165
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 80.34375,
      "epoch": 0.05633938925396212,
      "grad_norm": 26.70955658754936,
      "learning_rate": 4.718544646308465e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.39347878098487854,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1166
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.53125,
      "epoch": 0.056387707769617317,
      "grad_norm": 6.369523092788195,
      "learning_rate": 4.718303053730189e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1167
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.90625,
      "epoch": 0.056436026285272514,
      "grad_norm": 12.071830781577548,
      "learning_rate": 4.7180614611519133e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.3255898952484131,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1168
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 75.1875,
      "epoch": 0.05648434480092772,
      "grad_norm": 24.443157839718456,
      "learning_rate": 4.717819868573637e-07,
      "loss": 0.0,
      "reward": 1.546875,
      "reward_std": 0.4150615930557251,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1169
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 78.3125,
      "epoch": 0.056532663316582916,
      "grad_norm": 8.160297816367512,
      "learning_rate": 4.7175782759953616e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.30935919284820557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1170
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.03125,
      "epoch": 0.056580981832238114,
      "grad_norm": 30.717573689862032,
      "learning_rate": 4.717336683417085e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1171
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.5625,
      "epoch": 0.05662930034789331,
      "grad_norm": 0.0,
      "learning_rate": 4.7170950908388093e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1172
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 85.09375,
      "epoch": 0.05667761886354851,
      "grad_norm": 8.48658286430776,
      "learning_rate": 4.7168534982605334e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1173
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 78.6875,
      "epoch": 0.056725937379203714,
      "grad_norm": 11.510668632720732,
      "learning_rate": 4.716611905682257e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.35986602306365967,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1174
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.9375,
      "epoch": 0.05677425589485891,
      "grad_norm": 4.335207391862395,
      "learning_rate": 4.716370313103981e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1175
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.625,
      "epoch": 0.05682257441051411,
      "grad_norm": 6.028838987974425,
      "learning_rate": 4.716128720525705e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.3029785752296448,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1176
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.09375,
      "epoch": 0.05687089292616931,
      "grad_norm": 6.49360482939485,
      "learning_rate": 4.7158871279474295e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1177
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 90.0,
      "epoch": 0.056919211441824505,
      "grad_norm": 11.858180017766788,
      "learning_rate": 4.715645535369153e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.3541097044944763,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1178
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 93.5625,
      "epoch": 0.05696752995747971,
      "grad_norm": 13.712943190841893,
      "learning_rate": 4.715403942790877e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.4082317352294922,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1179
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 86.28125,
      "epoch": 0.05701584847313491,
      "grad_norm": 22.870421155051446,
      "learning_rate": 4.7151623502126013e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.3676722049713135,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1180
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 89.4375,
      "epoch": 0.057064166988790105,
      "grad_norm": 13.46699284933994,
      "learning_rate": 4.7149207576343255e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.47193512320518494,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1181
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 82.90625,
      "epoch": 0.0571124855044453,
      "grad_norm": 12.920633206441972,
      "learning_rate": 4.714679165056049e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1182
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 87.65625,
      "epoch": 0.0571608040201005,
      "grad_norm": 23.626591366316305,
      "learning_rate": 4.714437572477774e-07,
      "loss": -0.0,
      "reward": 1.578125,
      "reward_std": 0.4150615930557251,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1183
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 80.40625,
      "epoch": 0.057209122535755705,
      "grad_norm": 13.815662004384972,
      "learning_rate": 4.7141959798994974e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.4355512857437134,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1184
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 83.625,
      "epoch": 0.0572574410514109,
      "grad_norm": 12.663099938792257,
      "learning_rate": 4.713954387321221e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.31213027238845825,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1185
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 92.46875,
      "epoch": 0.0573057595670661,
      "grad_norm": 12.736952166913905,
      "learning_rate": 4.713712794742945e-07,
      "loss": 0.0,
      "reward": 1.1875,
      "reward_std": 0.32666346430778503,
      "rewards/accuracy_reward": 0.125,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 1186
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.03125,
      "epoch": 0.0573540780827213,
      "grad_norm": 8.209328451137422,
      "learning_rate": 4.713471202164669e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.2608962655067444,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1187
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 89.9375,
      "epoch": 0.057402396598376496,
      "grad_norm": 4.429050267283152,
      "learning_rate": 4.7132296095863934e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1188
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 88.53125,
      "epoch": 0.0574507151140317,
      "grad_norm": 27.19923336577701,
      "learning_rate": 4.712988017008117e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.28838545083999634,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1189
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 92.0,
      "epoch": 0.0574990336296869,
      "grad_norm": 35.39051308149413,
      "learning_rate": 4.7127464244298416e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.40044304728507996,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1190
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 85.625,
      "epoch": 0.057547352145342096,
      "grad_norm": 5.937316006755641,
      "learning_rate": 4.712504831851565e-07,
      "loss": -0.0,
      "reward": 2.421875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 1191
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 101.15625,
      "epoch": 0.05759567066099729,
      "grad_norm": 19.460873594031263,
      "learning_rate": 4.7122632392732894e-07,
      "loss": -0.0,
      "reward": 1.59375,
      "reward_std": 0.4408525228500366,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1192
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 92.03125,
      "epoch": 0.05764398917665249,
      "grad_norm": 11.887538010204697,
      "learning_rate": 4.7120216466950135e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1193
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.78125,
      "epoch": 0.057692307692307696,
      "grad_norm": 5.886677825994558,
      "learning_rate": 4.7117800541167377e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1194
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.4375,
      "epoch": 0.05774062620796289,
      "grad_norm": 12.64516665015178,
      "learning_rate": 4.711538461538461e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1195
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 76.125,
      "epoch": 0.05778894472361809,
      "grad_norm": 8.379416830762517,
      "learning_rate": 4.711296868960185e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.323003351688385,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1196
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.375,
      "epoch": 0.05783726323927329,
      "grad_norm": 7.712581772037064,
      "learning_rate": 4.7110552763819095e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1197
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.53125,
      "epoch": 0.057885581754928486,
      "grad_norm": 0.0,
      "learning_rate": 4.710813683803633e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1198
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.1875,
      "epoch": 0.05793390027058369,
      "grad_norm": 0.0,
      "learning_rate": 4.7105720912253573e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1199
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.125,
      "epoch": 0.05798221878623889,
      "grad_norm": 0.0,
      "learning_rate": 4.7103304986470814e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1200
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.53125,
      "epoch": 0.058030537301894086,
      "grad_norm": 0.0,
      "learning_rate": 4.7100889060688056e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1201
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.625,
      "epoch": 0.058078855817549284,
      "grad_norm": 10.502549803309497,
      "learning_rate": 4.709847313490529e-07,
      "loss": -0.0,
      "reward": 2.109375,
      "reward_std": 0.3051002025604248,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1202
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.375,
      "epoch": 0.05812717433320448,
      "grad_norm": 0.0,
      "learning_rate": 4.709605720912254e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1203
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 93.25,
      "epoch": 0.058175492848859686,
      "grad_norm": 13.02859030316284,
      "learning_rate": 4.7093641283339774e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.37722262740135193,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1204
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 83.5625,
      "epoch": 0.058223811364514884,
      "grad_norm": 6.320087080259669,
      "learning_rate": 4.709122535755701e-07,
      "loss": -0.0,
      "reward": 2.109375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1205
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.03125,
      "epoch": 0.05827212988017008,
      "grad_norm": 0.0,
      "learning_rate": 4.708880943177425e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 1206
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.46875,
      "epoch": 0.05832044839582528,
      "grad_norm": 5.563271810042273,
      "learning_rate": 4.7086393505991493e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1207
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 79.5,
      "epoch": 0.05836876691148048,
      "grad_norm": 11.380109052028253,
      "learning_rate": 4.7083977580208734e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1208
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.09375,
      "epoch": 0.05841708542713568,
      "grad_norm": 0.0,
      "learning_rate": 4.708156165442597e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1209
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.78125,
      "epoch": 0.05846540394279088,
      "grad_norm": 10.502992178150476,
      "learning_rate": 4.7079145728643217e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.37298911809921265,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1210
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.03125,
      "epoch": 0.05851372245844608,
      "grad_norm": 0.0,
      "learning_rate": 4.7076729802860453e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 1211
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 77.875,
      "epoch": 0.058562040974101275,
      "grad_norm": 15.739383530212788,
      "learning_rate": 4.7074313877077695e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.441916286945343,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1212
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 80.1875,
      "epoch": 0.05861035948975647,
      "grad_norm": 32.0246735311447,
      "learning_rate": 4.7071897951294936e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.3913571238517761,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1213
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 74.21875,
      "epoch": 0.05865867800541168,
      "grad_norm": 9.08506715454203,
      "learning_rate": 4.706948202551218e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1214
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 65.03125,
      "epoch": 0.058706996521066875,
      "grad_norm": 14.92401914582699,
      "learning_rate": 4.7067066099729413e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1215
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 68.28125,
      "epoch": 0.05875531503672207,
      "grad_norm": 21.11415862308649,
      "learning_rate": 4.706465017394665e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.38452720642089844,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1216
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 69.90625,
      "epoch": 0.05880363355237727,
      "grad_norm": 15.042779483018043,
      "learning_rate": 4.7062234248163896e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.3051002025604248,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1217
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 77.5,
      "epoch": 0.05885195206803247,
      "grad_norm": 30.73011218671331,
      "learning_rate": 4.705981832238113e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.5523518919944763,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1218
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 83.96875,
      "epoch": 0.05890027058368767,
      "grad_norm": 17.50371495246961,
      "learning_rate": 4.7057402396598374e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.5713036060333252,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1219
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.875,
      "epoch": 0.05894858909934287,
      "grad_norm": 0.0,
      "learning_rate": 4.7054986470815615e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1220
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.0,
      "epoch": 0.05899690761499807,
      "grad_norm": 0.0,
      "learning_rate": 4.7052570545032856e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1221
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.8125,
      "epoch": 0.059045226130653265,
      "grad_norm": 0.0,
      "learning_rate": 4.705015461925009e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1222
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 77.875,
      "epoch": 0.05909354464630846,
      "grad_norm": 12.928049366864391,
      "learning_rate": 4.704773869346734e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.34786468744277954,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1223
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 74.5,
      "epoch": 0.05914186316196367,
      "grad_norm": 13.032459360831615,
      "learning_rate": 4.7045322767684575e-07,
      "loss": -0.0,
      "reward": 2.078125,
      "reward_std": 0.3945268988609314,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1224
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 78.71875,
      "epoch": 0.059190181677618865,
      "grad_norm": 16.74802703801324,
      "learning_rate": 4.704290684190181e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.558118462562561,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1225
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 69.125,
      "epoch": 0.05923850019327406,
      "grad_norm": 11.423204441176226,
      "learning_rate": 4.704049091611905e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1226
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 76.71875,
      "epoch": 0.05928681870892926,
      "grad_norm": 9.481528837081662,
      "learning_rate": 4.7038074990336294e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.32666346430778503,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1227
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 70.8125,
      "epoch": 0.05933513722458446,
      "grad_norm": 10.18635689349688,
      "learning_rate": 4.7035659064553535e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.37086743116378784,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1228
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 80.78125,
      "epoch": 0.05938345574023966,
      "grad_norm": 22.679667103950408,
      "learning_rate": 4.703324313877077e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.4751303493976593,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1229
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 67.0,
      "epoch": 0.05943177425589486,
      "grad_norm": 4.749603261897035,
      "learning_rate": 4.703082721298802e-07,
      "loss": -0.0,
      "reward": 2.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1230
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 64.59375,
      "epoch": 0.05948009277155006,
      "grad_norm": 8.492740098190643,
      "learning_rate": 4.7028411287205254e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1231
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 67.46875,
      "epoch": 0.059528411287205256,
      "grad_norm": 10.979395460939624,
      "learning_rate": 4.7025995361422495e-07,
      "loss": 0.0,
      "reward": 2.1875,
      "reward_std": 0.35772839188575745,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1232
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 68.375,
      "epoch": 0.059576729802860454,
      "grad_norm": 48.98170110440381,
      "learning_rate": 4.7023579435639737e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.3676624000072479,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1233
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 57.84375,
      "epoch": 0.05962504831851566,
      "grad_norm": 6.75902314454459,
      "learning_rate": 4.702116350985698e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1234
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 68.46875,
      "epoch": 0.059673366834170856,
      "grad_norm": 12.02172812235064,
      "learning_rate": 4.7018747584074214e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.3051002323627472,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1235
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 62.4375,
      "epoch": 0.059721685349826054,
      "grad_norm": 7.727376320314803,
      "learning_rate": 4.701633165829145e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.3006648123264313,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1236
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 68.53125,
      "epoch": 0.05977000386548125,
      "grad_norm": 23.773095522499215,
      "learning_rate": 4.7013915732508697e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.23827512562274933,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1237
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 70.59375,
      "epoch": 0.05981832238113645,
      "grad_norm": 19.575890052185304,
      "learning_rate": 4.7011499806725933e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.3006556034088135,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1238
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 70.375,
      "epoch": 0.059866640896791654,
      "grad_norm": 239.7985947402616,
      "learning_rate": 4.7009083880943174e-07,
      "loss": -0.0,
      "reward": 2.3125,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 1239
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 64.1875,
      "epoch": 0.05991495941244685,
      "grad_norm": 8.168944385806842,
      "learning_rate": 4.7006667955160416e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1240
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 70.53125,
      "epoch": 0.05996327792810205,
      "grad_norm": 0.0,
      "learning_rate": 4.7004252029377657e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 1241
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 67.78125,
      "epoch": 0.06001159644375725,
      "grad_norm": 16.258814811877205,
      "learning_rate": 4.7001836103594893e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.49766814708709717,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1242
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 68.375,
      "epoch": 0.060059914959412444,
      "grad_norm": 6.251611213583932,
      "learning_rate": 4.699942017781214e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1243
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.65625,
      "epoch": 0.06010823347506765,
      "grad_norm": 0.0,
      "learning_rate": 4.6997004252029376e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 1244
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 62.65625,
      "epoch": 0.06015655199072285,
      "grad_norm": 8.464794607551339,
      "learning_rate": 4.6994588326246617e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1245
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.625,
      "epoch": 0.060204870506378044,
      "grad_norm": 0.0,
      "learning_rate": 4.6992172400463853e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1246
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 67.0625,
      "epoch": 0.06025318902203324,
      "grad_norm": 6.205469683793644,
      "learning_rate": 4.6989756474681095e-07,
      "loss": -0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 1247
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 71.21875,
      "epoch": 0.06030150753768844,
      "grad_norm": 21.473364791601405,
      "learning_rate": 4.6987340548898336e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.34254783391952515,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1248
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 65.46875,
      "epoch": 0.060349826053343644,
      "grad_norm": 8.006839898716146,
      "learning_rate": 4.698492462311557e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1249
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 70.4375,
      "epoch": 0.06039814456899884,
      "grad_norm": 5.735148669102786,
      "learning_rate": 4.698250869733282e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.26621314883232117,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1250
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.0,
      "epoch": 0.06044646308465404,
      "grad_norm": 0.0,
      "learning_rate": 4.6980092771550055e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1251
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.5625,
      "epoch": 0.06049478160030924,
      "grad_norm": 0.0,
      "learning_rate": 4.6977676845767296e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1252
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 74.78125,
      "epoch": 0.060543100115964435,
      "grad_norm": 5.8336955619155955,
      "learning_rate": 4.697526091998454e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1253
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 72.875,
      "epoch": 0.06059141863161964,
      "grad_norm": 5.63706735917253,
      "learning_rate": 4.697284499420178e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1254
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 72.6875,
      "epoch": 0.06063973714727484,
      "grad_norm": 29.134727593388664,
      "learning_rate": 4.6970429068419015e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.45159047842025757,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1255
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 82.71875,
      "epoch": 0.060688055662930035,
      "grad_norm": 15.173764478535686,
      "learning_rate": 4.696801314263625e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.43236586451530457,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1256
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 73.78125,
      "epoch": 0.06073637417858523,
      "grad_norm": 15.37618076751819,
      "learning_rate": 4.69655972168535e-07,
      "loss": -0.0,
      "reward": 1.34375,
      "reward_std": 0.32239729166030884,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 1257
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.65625,
      "epoch": 0.06078469269424043,
      "grad_norm": 0.0,
      "learning_rate": 4.6963181291070734e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1258
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 71.5,
      "epoch": 0.060833011209895635,
      "grad_norm": 29.925006412067578,
      "learning_rate": 4.6960765365287975e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.5386105179786682,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1259
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.0,
      "epoch": 0.06088132972555083,
      "grad_norm": 15.83099732457386,
      "learning_rate": 4.6958349439505217e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.2951723635196686,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1260
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.625,
      "epoch": 0.06092964824120603,
      "grad_norm": 0.0,
      "learning_rate": 4.695593351372246e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1261
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.15625,
      "epoch": 0.06097796675686123,
      "grad_norm": 0.0,
      "learning_rate": 4.6953517587939694e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1262
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.78125,
      "epoch": 0.061026285272516426,
      "grad_norm": 0.0,
      "learning_rate": 4.695110166215694e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1263
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 71.40625,
      "epoch": 0.06107460378817163,
      "grad_norm": 7.838365842172631,
      "learning_rate": 4.6948685736374177e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.34624922275543213,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1264
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.34375,
      "epoch": 0.06112292230382683,
      "grad_norm": 0.0,
      "learning_rate": 4.694626981059142e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1265
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.15625,
      "epoch": 0.061171240819482026,
      "grad_norm": 0.0,
      "learning_rate": 4.6943853884808654e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1266
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 59.65625,
      "epoch": 0.06121955933513722,
      "grad_norm": 9.81410180338966,
      "learning_rate": 4.6941437959025895e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1267
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.375,
      "epoch": 0.06126787785079242,
      "grad_norm": 0.0,
      "learning_rate": 4.6939022033243137e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 1268
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 74.65625,
      "epoch": 0.061316196366447626,
      "grad_norm": 12.461950200225882,
      "learning_rate": 4.6936606107460373e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.3940780758857727,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1269
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 72.8125,
      "epoch": 0.06136451488210282,
      "grad_norm": 9.194624057192476,
      "learning_rate": 4.693419018167762e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1270
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 69.65625,
      "epoch": 0.06141283339775802,
      "grad_norm": 4.618315951657601,
      "learning_rate": 4.6931774255894856e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.16807307302951813,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1271
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.21875,
      "epoch": 0.06146115191341322,
      "grad_norm": 0.0,
      "learning_rate": 4.6929358330112097e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1272
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.375,
      "epoch": 0.061509470429068416,
      "grad_norm": 0.0,
      "learning_rate": 4.692694240432934e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1273
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 76.25,
      "epoch": 0.06155778894472362,
      "grad_norm": 11.222770930411013,
      "learning_rate": 4.692452647854658e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.42190128564834595,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1274
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 75.0625,
      "epoch": 0.06160610746037882,
      "grad_norm": 17.48962031608994,
      "learning_rate": 4.6922110552763816e-07,
      "loss": -0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 1275
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.0,
      "epoch": 0.061654425976034016,
      "grad_norm": 13.677443891355782,
      "learning_rate": 4.6919694626981057e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.3187600374221802,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1276
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 63.84375,
      "epoch": 0.061702744491689214,
      "grad_norm": 12.945382847927231,
      "learning_rate": 4.69172787011983e-07,
      "loss": -0.0,
      "reward": 1.734375,
      "reward_std": 0.2961486577987671,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1277
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 68.78125,
      "epoch": 0.06175106300734441,
      "grad_norm": 226.34779764364598,
      "learning_rate": 4.6914862775415535e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1278
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.25,
      "epoch": 0.061799381522999616,
      "grad_norm": 0.0,
      "learning_rate": 4.6912446849632776e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1279
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 89.34375,
      "epoch": 0.061847700038654814,
      "grad_norm": 10.349449768307624,
      "learning_rate": 4.6910030923850017e-07,
      "loss": -0.0,
      "reward": 1.75,
      "reward_std": 0.3642283082008362,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1280
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.0625,
      "epoch": 0.06189601855431001,
      "grad_norm": 0.0,
      "learning_rate": 4.690761499806726e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1281
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 65.78125,
      "epoch": 0.06194433706996521,
      "grad_norm": 15.661910621605543,
      "learning_rate": 4.6905199072284495e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.3913571238517761,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1282
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.0,
      "epoch": 0.06199265558562041,
      "grad_norm": 9.052000040877735,
      "learning_rate": 4.690278314650174e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1283
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.21875,
      "epoch": 0.06204097410127561,
      "grad_norm": 0.0,
      "learning_rate": 4.690036722071898e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1284
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 65.5,
      "epoch": 0.06208929261693081,
      "grad_norm": 15.79165331508932,
      "learning_rate": 4.689795129493622e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.44239097833633423,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1285
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 70.875,
      "epoch": 0.06213761113258601,
      "grad_norm": 9.597307355507658,
      "learning_rate": 4.6895535369153455e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1286
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.34375,
      "epoch": 0.062185929648241205,
      "grad_norm": 0.0,
      "learning_rate": 4.68931194433707e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1287
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 73.09375,
      "epoch": 0.0622342481638964,
      "grad_norm": 51.94830191551938,
      "learning_rate": 4.689070351758794e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1288
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.25,
      "epoch": 0.06228256667955161,
      "grad_norm": 0.0,
      "learning_rate": 4.6888287591805174e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1289
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 77.34375,
      "epoch": 0.062330885195206805,
      "grad_norm": 22.563784547227552,
      "learning_rate": 4.688587166602242e-07,
      "loss": -0.0,
      "reward": 2.09375,
      "reward_std": 0.4998054802417755,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1290
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.46875,
      "epoch": 0.062379203710862,
      "grad_norm": 0.0,
      "learning_rate": 4.6883455740239656e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1291
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 68.9375,
      "epoch": 0.0624275222265172,
      "grad_norm": 19.50324285633205,
      "learning_rate": 4.68810398144569e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.2782030999660492,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1292
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 58.9375,
      "epoch": 0.0624758407421724,
      "grad_norm": 10.510682127911618,
      "learning_rate": 4.687862388867414e-07,
      "loss": -0.0,
      "reward": 2.28125,
      "reward_std": 0.30430006980895996,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1293
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 78.8125,
      "epoch": 0.0625241592578276,
      "grad_norm": 22.774237040816512,
      "learning_rate": 4.687620796289138e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.39774754643440247,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1294
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 75.8125,
      "epoch": 0.06257247777348279,
      "grad_norm": 24.319280756410564,
      "learning_rate": 4.6873792037108617e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.5323013663291931,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1295
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 74.15625,
      "epoch": 0.062620796289138,
      "grad_norm": 23.701124031531073,
      "learning_rate": 4.687137611132586e-07,
      "loss": -0.0,
      "reward": 1.90625,
      "reward_std": 0.37086743116378784,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1296
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 71.4375,
      "epoch": 0.0626691148047932,
      "grad_norm": 5.430025946309102,
      "learning_rate": 4.68689601855431e-07,
      "loss": -0.0,
      "reward": 2.40625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 1297
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 74.125,
      "epoch": 0.06271743332044839,
      "grad_norm": 11.95291336456527,
      "learning_rate": 4.6866544259760335e-07,
      "loss": 0.0,
      "reward": 1.265625,
      "reward_std": 0.3185884654521942,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 1298
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.3125,
      "epoch": 0.0627657518361036,
      "grad_norm": 15.50987938296631,
      "learning_rate": 4.6864128333977577e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.2877861559391022,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1299
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 69.625,
      "epoch": 0.06281407035175879,
      "grad_norm": 23.912216867015793,
      "learning_rate": 4.686171240819482e-07,
      "loss": -0.0,
      "reward": 2.140625,
      "reward_std": 0.33386313915252686,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1300
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 67.5,
      "epoch": 0.06286238886741399,
      "grad_norm": 0.0,
      "learning_rate": 4.685929648241206e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1301
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.0,
      "epoch": 0.0629107073830692,
      "grad_norm": 0.0,
      "learning_rate": 4.6856880556629296e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 1302
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.4375,
      "epoch": 0.06295902589872439,
      "grad_norm": 0.0,
      "learning_rate": 4.685446463084654e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 1303
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.40625,
      "epoch": 0.06300734441437959,
      "grad_norm": 10.045034758943874,
      "learning_rate": 4.685204870506378e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1304
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 62.9375,
      "epoch": 0.06305566293003478,
      "grad_norm": 40.12880868743318,
      "learning_rate": 4.684963277928102e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.3650856614112854,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1305
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 64.125,
      "epoch": 0.06310398144568999,
      "grad_norm": 7.386655218342526,
      "learning_rate": 4.6847216853498256e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1306
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 86.28125,
      "epoch": 0.06315229996134519,
      "grad_norm": 13.15985186653581,
      "learning_rate": 4.68448009277155e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.48943066596984863,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1307
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.96875,
      "epoch": 0.06320061847700038,
      "grad_norm": 0.0,
      "learning_rate": 4.684238500193274e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1308
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 70.3125,
      "epoch": 0.06324893699265559,
      "grad_norm": 7.397151933365568,
      "learning_rate": 4.6839969076149974e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.308285653591156,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1309
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.09375,
      "epoch": 0.06329725550831078,
      "grad_norm": 10.485304000724224,
      "learning_rate": 4.683755315036722e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.2461063265800476,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1310
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 67.875,
      "epoch": 0.06334557402396598,
      "grad_norm": 38.17141027426805,
      "learning_rate": 4.6835137224584457e-07,
      "loss": -0.0,
      "reward": 1.484375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1311
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.53125,
      "epoch": 0.06339389253962119,
      "grad_norm": 0.0,
      "learning_rate": 4.68327212988017e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1312
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.28125,
      "epoch": 0.06344221105527638,
      "grad_norm": 0.0,
      "learning_rate": 4.683030537301894e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1313
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.125,
      "epoch": 0.06349052957093158,
      "grad_norm": 0.0,
      "learning_rate": 4.682788944723618e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 1314
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.53125,
      "epoch": 0.06353884808658677,
      "grad_norm": 0.0,
      "learning_rate": 4.682547352145342e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1315
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 73.15625,
      "epoch": 0.06358716660224198,
      "grad_norm": 12.711550709994793,
      "learning_rate": 4.682305759567066e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1316
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.5625,
      "epoch": 0.06363548511789718,
      "grad_norm": 0.0,
      "learning_rate": 4.68206416698879e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1317
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.1875,
      "epoch": 0.06368380363355237,
      "grad_norm": 0.0,
      "learning_rate": 4.6818225744105136e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1318
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 81.09375,
      "epoch": 0.06373212214920758,
      "grad_norm": 7.669460998692466,
      "learning_rate": 4.681580981832238e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.20785759389400482,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1319
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.8125,
      "epoch": 0.06378044066486277,
      "grad_norm": 0.0,
      "learning_rate": 4.681339389253962e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1320
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.9375,
      "epoch": 0.06382875918051797,
      "grad_norm": 0.0,
      "learning_rate": 4.681097796675686e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1321
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 74.84375,
      "epoch": 0.06387707769617318,
      "grad_norm": 9.661417989858922,
      "learning_rate": 4.6808562040974096e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.2505247890949249,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1322
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 68.03125,
      "epoch": 0.06392539621182837,
      "grad_norm": 7.27446364278802,
      "learning_rate": 4.6806146115191343e-07,
      "loss": -0.0,
      "reward": 2.328125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1323
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 70.03125,
      "epoch": 0.06397371472748357,
      "grad_norm": 11.019984830632444,
      "learning_rate": 4.680373018940858e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1324
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.28125,
      "epoch": 0.06402203324313877,
      "grad_norm": 0.0,
      "learning_rate": 4.680131426362582e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1325
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 73.3125,
      "epoch": 0.06407035175879397,
      "grad_norm": 10.669214605764287,
      "learning_rate": 4.6798898337843056e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1326
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.34375,
      "epoch": 0.06411867027444917,
      "grad_norm": 0.0,
      "learning_rate": 4.6796482412060303e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1327
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 74.15625,
      "epoch": 0.06416698879010437,
      "grad_norm": 10.37404786943398,
      "learning_rate": 4.679406648627754e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1328
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.9375,
      "epoch": 0.06421530730575957,
      "grad_norm": 0.0,
      "learning_rate": 4.6791650560494775e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1329
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.75,
      "epoch": 0.06426362582141476,
      "grad_norm": 0.0,
      "learning_rate": 4.678923463471202e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1330
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.75,
      "epoch": 0.06431194433706997,
      "grad_norm": 6.972034326884459,
      "learning_rate": 4.678681870892926e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1331
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 70.5,
      "epoch": 0.06436026285272517,
      "grad_norm": 11.593584014918925,
      "learning_rate": 4.67844027831465e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.30856335163116455,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1332
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 66.65625,
      "epoch": 0.06440858136838036,
      "grad_norm": 34.198002799860554,
      "learning_rate": 4.678198685736374e-07,
      "loss": -0.0,
      "reward": 2.078125,
      "reward_std": 0.3629639744758606,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1333
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 74.96875,
      "epoch": 0.06445689988403557,
      "grad_norm": 10.140406875600952,
      "learning_rate": 4.677957093158098e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1334
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.53125,
      "epoch": 0.06450521839969076,
      "grad_norm": 16.44505877508684,
      "learning_rate": 4.677715500579822e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1335
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 82.6875,
      "epoch": 0.06455353691534596,
      "grad_norm": 110.69972175021512,
      "learning_rate": 4.677473908001546e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.3471825420856476,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 1336
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 69.96875,
      "epoch": 0.06460185543100117,
      "grad_norm": 28.629027548658396,
      "learning_rate": 4.67723231542327e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.2435920089483261,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1337
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 72.375,
      "epoch": 0.06465017394665636,
      "grad_norm": 0.0,
      "learning_rate": 4.676990722844994e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1338
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.65625,
      "epoch": 0.06469849246231156,
      "grad_norm": 0.0,
      "learning_rate": 4.676749130266718e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1339
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 75.15625,
      "epoch": 0.06474681097796675,
      "grad_norm": 23.270327760010524,
      "learning_rate": 4.676507537688442e-07,
      "loss": -0.0,
      "reward": 1.453125,
      "reward_std": 0.3457430303096771,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1340
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 67.21875,
      "epoch": 0.06479512949362196,
      "grad_norm": 60.67017278673892,
      "learning_rate": 4.676265945110166e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1341
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.71875,
      "epoch": 0.06484344800927716,
      "grad_norm": 8.387124448412738,
      "learning_rate": 4.6760243525318897e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1342
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 72.90625,
      "epoch": 0.06489176652493235,
      "grad_norm": 13.186429842841623,
      "learning_rate": 4.6757827599536144e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.24251843988895416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1343
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 90.90625,
      "epoch": 0.06494008504058756,
      "grad_norm": 131.0385381864973,
      "learning_rate": 4.675541167375338e-07,
      "loss": 0.0,
      "reward": 1.078125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.0625,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 1344
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.8125,
      "epoch": 0.06498840355624275,
      "grad_norm": 23.172113114035163,
      "learning_rate": 4.675299574797062e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1345
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 84.78125,
      "epoch": 0.06503672207189795,
      "grad_norm": 9.527391266851348,
      "learning_rate": 4.6750579822187857e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1346
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.25,
      "epoch": 0.06508504058755316,
      "grad_norm": 0.0,
      "learning_rate": 4.6748163896405104e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1347
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.46875,
      "epoch": 0.06513335910320835,
      "grad_norm": 0.0,
      "learning_rate": 4.674574797062234e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1348
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.03125,
      "epoch": 0.06518167761886355,
      "grad_norm": 0.0,
      "learning_rate": 4.6743332044839576e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1349
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.03125,
      "epoch": 0.06522999613451874,
      "grad_norm": 0.0,
      "learning_rate": 4.6740916119056823e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 1350
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.65625,
      "epoch": 0.06527831465017395,
      "grad_norm": 0.0,
      "learning_rate": 4.673850019327406e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1351
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 73.9375,
      "epoch": 0.06532663316582915,
      "grad_norm": 5.642699312263824,
      "learning_rate": 4.67360842674913e-07,
      "loss": 0.0,
      "reward": 1.390625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1352
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.25,
      "epoch": 0.06537495168148434,
      "grad_norm": 0.0,
      "learning_rate": 4.673366834170854e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1353
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.46875,
      "epoch": 0.06542327019713955,
      "grad_norm": 0.0,
      "learning_rate": 4.6731252415925783e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1354
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.125,
      "epoch": 0.06547158871279474,
      "grad_norm": 6.180291895128197,
      "learning_rate": 4.672883649014302e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.18857255578041077,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1355
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.1875,
      "epoch": 0.06551990722844994,
      "grad_norm": 0.0,
      "learning_rate": 4.672642056436026e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1356
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.0625,
      "epoch": 0.06556822574410515,
      "grad_norm": 0.0,
      "learning_rate": 4.67240046385775e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 1357
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.46875,
      "epoch": 0.06561654425976034,
      "grad_norm": 142.34386257311624,
      "learning_rate": 4.6721588712794743e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23827512562274933,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 1358
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.125,
      "epoch": 0.06566486277541554,
      "grad_norm": 18.522556012624445,
      "learning_rate": 4.671917278701198e-07,
      "loss": -0.0,
      "reward": 1.90625,
      "reward_std": 0.29305070638656616,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1359
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 81.40625,
      "epoch": 0.06571318129107073,
      "grad_norm": 9.897967995278696,
      "learning_rate": 4.671675686122922e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1360
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 72.34375,
      "epoch": 0.06576149980672594,
      "grad_norm": 22.637296192241063,
      "learning_rate": 4.671434093544646e-07,
      "loss": -0.0,
      "reward": 1.453125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1361
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.4375,
      "epoch": 0.06580981832238114,
      "grad_norm": 0.0,
      "learning_rate": 4.67119250096637e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 1362
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 76.4375,
      "epoch": 0.06585813683803633,
      "grad_norm": 5.337456010764276,
      "learning_rate": 4.6709509083880945e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1363
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.125,
      "epoch": 0.06590645535369154,
      "grad_norm": 0.0,
      "learning_rate": 4.670709315809818e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 1364
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.46875,
      "epoch": 0.06595477386934673,
      "grad_norm": 0.0,
      "learning_rate": 4.670467723231542e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1365
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.09375,
      "epoch": 0.06600309238500193,
      "grad_norm": 9.181116412126075,
      "learning_rate": 4.670226130653266e-07,
      "loss": -0.0,
      "reward": 1.734375,
      "reward_std": 0.24831002950668335,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1366
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 68.96875,
      "epoch": 0.06605141090065714,
      "grad_norm": 51.766004173986595,
      "learning_rate": 4.6699845380749905e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.3211550712585449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1367
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.625,
      "epoch": 0.06609972941631233,
      "grad_norm": 0.0,
      "learning_rate": 4.669742945496714e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1368
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.84375,
      "epoch": 0.06614804793196753,
      "grad_norm": 0.0,
      "learning_rate": 4.6695013529184377e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1369
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 70.0625,
      "epoch": 0.06619636644762272,
      "grad_norm": 9.3436458248785,
      "learning_rate": 4.6692597603401623e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1370
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.875,
      "epoch": 0.06624468496327793,
      "grad_norm": 0.0,
      "learning_rate": 4.669018167761886e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 1371
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 69.5625,
      "epoch": 0.06629300347893313,
      "grad_norm": 7.499816908130431,
      "learning_rate": 4.66877657518361e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1372
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.0625,
      "epoch": 0.06634132199458832,
      "grad_norm": 4.863487371675879,
      "learning_rate": 4.668534982605334e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.24831002950668335,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1373
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.15625,
      "epoch": 0.06638964051024353,
      "grad_norm": 0.0,
      "learning_rate": 4.6682933900270584e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1374
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.125,
      "epoch": 0.06643795902589872,
      "grad_norm": 12.789267017470927,
      "learning_rate": 4.668051797448782e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.27883464097976685,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1375
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 65.71875,
      "epoch": 0.06648627754155392,
      "grad_norm": 30.835289401068703,
      "learning_rate": 4.667810204870506e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.3761745095252991,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1376
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 87.9375,
      "epoch": 0.06653459605720913,
      "grad_norm": 18.535425864659263,
      "learning_rate": 4.66756861229223e-07,
      "loss": -0.0,
      "reward": 1.5,
      "reward_std": 0.6193310022354126,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 1377
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.3125,
      "epoch": 0.06658291457286432,
      "grad_norm": 28.699515815161494,
      "learning_rate": 4.6673270197139544e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.31983357667922974,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1378
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 65.59375,
      "epoch": 0.06663123308851952,
      "grad_norm": 7.327879775395915,
      "learning_rate": 4.667085427135678e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.31983357667922974,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1379
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 73.4375,
      "epoch": 0.06667955160417471,
      "grad_norm": 30.538126413618663,
      "learning_rate": 4.666843834557402e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.4692808985710144,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1380
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 73.375,
      "epoch": 0.06672787011982992,
      "grad_norm": 13.547556511433791,
      "learning_rate": 4.666602241979126e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.3987956643104553,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1381
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.4375,
      "epoch": 0.06677618863548512,
      "grad_norm": 0.0,
      "learning_rate": 4.66636064940085e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.2561880350112915,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1382
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.0,
      "epoch": 0.06682450715114031,
      "grad_norm": 13.645170767216305,
      "learning_rate": 4.6661190568225745e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.4797552525997162,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1383
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 86.03125,
      "epoch": 0.06687282566679552,
      "grad_norm": 36.08016918447566,
      "learning_rate": 4.665877464244298e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.5243093967437744,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1384
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 96.40625,
      "epoch": 0.06692114418245071,
      "grad_norm": 40.194286665808825,
      "learning_rate": 4.6656358716660223e-07,
      "loss": 0.0,
      "reward": 1.546875,
      "reward_std": 0.6026557087898254,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1385
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 80.25,
      "epoch": 0.06696946269810591,
      "grad_norm": 9.204502119056798,
      "learning_rate": 4.665394279087746e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.2715304493904114,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 1386
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.46875,
      "epoch": 0.06701778121376112,
      "grad_norm": 0.0,
      "learning_rate": 4.6651526865094705e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1387
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.90625,
      "epoch": 0.06706609972941631,
      "grad_norm": 0.0,
      "learning_rate": 4.664911093931194e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1388
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 78.125,
      "epoch": 0.06711441824507151,
      "grad_norm": 13.795575637573558,
      "learning_rate": 4.6646695013529183e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 1389
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 76.8125,
      "epoch": 0.0671627367607267,
      "grad_norm": 23.86196125901962,
      "learning_rate": 4.6644279087746424e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.3074946999549866,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1390
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 89.46875,
      "epoch": 0.06721105527638191,
      "grad_norm": 16.179665637886284,
      "learning_rate": 4.664186316196366e-07,
      "loss": -0.0,
      "reward": 1.421875,
      "reward_std": 0.4517943859100342,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 1391
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 76.28125,
      "epoch": 0.06725937379203711,
      "grad_norm": 11.495714501228965,
      "learning_rate": 4.66394472361809e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.33571791648864746,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1392
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 88.875,
      "epoch": 0.0673076923076923,
      "grad_norm": 87.8398537108363,
      "learning_rate": 4.6637031310398143e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.48572152853012085,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1393
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.34375,
      "epoch": 0.06735601082334751,
      "grad_norm": 55.157275876431704,
      "learning_rate": 4.6634615384615384e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.3708576261997223,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1394
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.40625,
      "epoch": 0.0674043293390027,
      "grad_norm": 6.222930153047393,
      "learning_rate": 4.663219945883262e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1395
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.625,
      "epoch": 0.0674526478546579,
      "grad_norm": 7.329107869653753,
      "learning_rate": 4.662978353304986e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.2783603370189667,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1396
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.15625,
      "epoch": 0.06750096637031311,
      "grad_norm": 0.0,
      "learning_rate": 4.6627367607267103e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1397
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.46875,
      "epoch": 0.0675492848859683,
      "grad_norm": 0.0,
      "learning_rate": 4.6624951681484345e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 1398
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.78125,
      "epoch": 0.0675976034016235,
      "grad_norm": 0.0,
      "learning_rate": 4.662253575570158e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1399
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 93.0625,
      "epoch": 0.0676459219172787,
      "grad_norm": 30.51172125321189,
      "learning_rate": 4.6620119829918827e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.5339856147766113,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1400
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 76.96875,
      "epoch": 0.0676942404329339,
      "grad_norm": 14.672434784609553,
      "learning_rate": 4.6617703904136063e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.4703400433063507,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1401
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.625,
      "epoch": 0.0677425589485891,
      "grad_norm": 0.0,
      "learning_rate": 4.66152879783533e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1402
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 70.5,
      "epoch": 0.0677908774642443,
      "grad_norm": 6.561757827099992,
      "learning_rate": 4.6612872052570546e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.30509042739868164,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1403
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 79.75,
      "epoch": 0.0678391959798995,
      "grad_norm": 21.280717435715722,
      "learning_rate": 4.661045612678778e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.4629974961280823,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1404
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 78.65625,
      "epoch": 0.06788751449555469,
      "grad_norm": 16.459594720563214,
      "learning_rate": 4.6608040201005024e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.4770200252532959,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1405
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 74.75,
      "epoch": 0.0679358330112099,
      "grad_norm": 13.986276268924811,
      "learning_rate": 4.660562427522226e-07,
      "loss": 0.0,
      "reward": 1.40625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1406
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 76.46875,
      "epoch": 0.0679841515268651,
      "grad_norm": 9.289517461393995,
      "learning_rate": 4.6603208349439506e-07,
      "loss": 0.0,
      "reward": 1.4375,
      "reward_std": 0.19266104698181152,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1407
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 80.0,
      "epoch": 0.06803247004252029,
      "grad_norm": 5.763787463394781,
      "learning_rate": 4.660079242365674e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1408
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 72.0625,
      "epoch": 0.0680807885581755,
      "grad_norm": 0.0,
      "learning_rate": 4.6598376497873984e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 1409
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 75.59375,
      "epoch": 0.06812910707383069,
      "grad_norm": 27.206045351698304,
      "learning_rate": 4.6595960572091225e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.42873120307922363,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1410
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.75,
      "epoch": 0.06817742558948589,
      "grad_norm": 12.415554742845963,
      "learning_rate": 4.659354464630846e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.31983357667922974,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1411
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.65625,
      "epoch": 0.0682257441051411,
      "grad_norm": 8.357824367653924,
      "learning_rate": 4.65911287205257e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.3619876801967621,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1412
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 62.03125,
      "epoch": 0.06827406262079629,
      "grad_norm": 50.40193013447613,
      "learning_rate": 4.6588712794742944e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.3908922076225281,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1413
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.0625,
      "epoch": 0.06832238113645149,
      "grad_norm": 0.0,
      "learning_rate": 4.6586296868960185e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1414
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 74.15625,
      "epoch": 0.06837069965210668,
      "grad_norm": 27.95175889507559,
      "learning_rate": 4.658388094317742e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.49680590629577637,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1415
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 83.71875,
      "epoch": 0.06841901816776189,
      "grad_norm": 96.44979125688249,
      "learning_rate": 4.6581465017394663e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.517636775970459,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1416
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 84.125,
      "epoch": 0.06846733668341709,
      "grad_norm": 10.408593241109706,
      "learning_rate": 4.6579049091611904e-07,
      "loss": -0.0,
      "reward": 1.125,
      "reward_std": 0.28104937076568604,
      "rewards/accuracy_reward": 0.09375,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 1417
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.59375,
      "epoch": 0.06851565519907228,
      "grad_norm": 9.884658299228297,
      "learning_rate": 4.6576633165829145e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.32223784923553467,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1418
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.34375,
      "epoch": 0.06856397371472749,
      "grad_norm": 0.0,
      "learning_rate": 4.657421724004638e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1419
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 82.03125,
      "epoch": 0.06861229223038268,
      "grad_norm": 15.632297163850291,
      "learning_rate": 4.657180131426363e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.3527624309062958,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1420
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 80.46875,
      "epoch": 0.06866061074603788,
      "grad_norm": 16.219619947175246,
      "learning_rate": 4.6569385388480864e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.41745609045028687,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1421
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 94.0625,
      "epoch": 0.06870892926169309,
      "grad_norm": 20.48376517654622,
      "learning_rate": 4.65669694626981e-07,
      "loss": 0.0,
      "reward": 1.484375,
      "reward_std": 0.3252435624599457,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1422
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 72.03125,
      "epoch": 0.06875724777734828,
      "grad_norm": 8.982507845564056,
      "learning_rate": 4.6564553536915347e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.22152717411518097,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1423
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.59375,
      "epoch": 0.06880556629300348,
      "grad_norm": 0.0,
      "learning_rate": 4.6562137611132583e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1424
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 78.75,
      "epoch": 0.06885388480865867,
      "grad_norm": 14.703537315931168,
      "learning_rate": 4.6559721685349824e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.36213356256484985,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1425
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 71.8125,
      "epoch": 0.06890220332431388,
      "grad_norm": 22.84685414762783,
      "learning_rate": 4.655730575956706e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.27883464097976685,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1426
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 65.59375,
      "epoch": 0.06895052183996908,
      "grad_norm": 9.19166850340154,
      "learning_rate": 4.6554889833784307e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.316745400428772,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1427
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 80.09375,
      "epoch": 0.06899884035562427,
      "grad_norm": 6.988073402919115,
      "learning_rate": 4.6552473908001543e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.308285653591156,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1428
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.65625,
      "epoch": 0.06904715887127948,
      "grad_norm": 6.222140340015941,
      "learning_rate": 4.6550057982218784e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.249358132481575,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1429
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 77.75,
      "epoch": 0.06909547738693467,
      "grad_norm": 15.911984842763726,
      "learning_rate": 4.6547642056436026e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.4078102707862854,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1430
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 79.875,
      "epoch": 0.06914379590258987,
      "grad_norm": 13.758437557638983,
      "learning_rate": 4.654522613065326e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.3503679633140564,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1431
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.125,
      "epoch": 0.06919211441824508,
      "grad_norm": 0.0,
      "learning_rate": 4.6542810204870503e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 1432
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 65.9375,
      "epoch": 0.06924043293390027,
      "grad_norm": 17.027306418683608,
      "learning_rate": 4.6540394279087745e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1433
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 75.9375,
      "epoch": 0.06928875144955547,
      "grad_norm": 15.254877138085254,
      "learning_rate": 4.6537978353304986e-07,
      "loss": -0.0,
      "reward": 1.921875,
      "reward_std": 0.4483071267604828,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1434
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 79.0625,
      "epoch": 0.06933706996521066,
      "grad_norm": 114.00694637812623,
      "learning_rate": 4.653556242752222e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.4152792692184448,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1435
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.34375,
      "epoch": 0.06938538848086587,
      "grad_norm": 0.0,
      "learning_rate": 4.6533146501739463e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1436
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 75.65625,
      "epoch": 0.06943370699652107,
      "grad_norm": 10.475767553929215,
      "learning_rate": 4.6530730575956705e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.33090677857398987,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1437
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 87.125,
      "epoch": 0.06948202551217626,
      "grad_norm": 59.35369451289745,
      "learning_rate": 4.6528314650173946e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.3179922103881836,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1438
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 72.6875,
      "epoch": 0.06953034402783147,
      "grad_norm": 11.389043763343466,
      "learning_rate": 4.652589872439118e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.40618184208869934,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1439
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 75.03125,
      "epoch": 0.06957866254348666,
      "grad_norm": 6.919788124919219,
      "learning_rate": 4.652348279860843e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1440
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.125,
      "epoch": 0.06962698105914186,
      "grad_norm": 0.0,
      "learning_rate": 4.6521066872825665e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1441
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.625,
      "epoch": 0.06967529957479707,
      "grad_norm": 0.0,
      "learning_rate": 4.65186509470429e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1442
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.4375,
      "epoch": 0.06972361809045226,
      "grad_norm": 0.0,
      "learning_rate": 4.651623502126015e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1443
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 74.15625,
      "epoch": 0.06977193660610746,
      "grad_norm": 9.838140732766377,
      "learning_rate": 4.6513819095477384e-07,
      "loss": 0.0,
      "reward": 1.53125,
      "reward_std": 0.24067938327789307,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1444
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 75.125,
      "epoch": 0.06982025512176265,
      "grad_norm": 27.880597466666497,
      "learning_rate": 4.6511403169694625e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.4861104488372803,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1445
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 68.125,
      "epoch": 0.06986857363741786,
      "grad_norm": 12.166879707271717,
      "learning_rate": 4.650898724391186e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.21366669237613678,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1446
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.4375,
      "epoch": 0.06991689215307306,
      "grad_norm": 0.0,
      "learning_rate": 4.650657131812911e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1447
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 80.75,
      "epoch": 0.06996521066872825,
      "grad_norm": 13.60239290342872,
      "learning_rate": 4.6504155392346344e-07,
      "loss": -0.0,
      "reward": 1.40625,
      "reward_std": 0.3503679633140564,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 0.9375,
      "step": 1448
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.1875,
      "epoch": 0.07001352918438346,
      "grad_norm": 0.0,
      "learning_rate": 4.6501739466563585e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1449
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.0625,
      "epoch": 0.07006184770003865,
      "grad_norm": 0.0,
      "learning_rate": 4.6499323540780827e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1450
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 77.0625,
      "epoch": 0.07011016621569385,
      "grad_norm": 5.03843190774618,
      "learning_rate": 4.649690761499807e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.28449735045433044,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 0.96875,
      "step": 1451
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 82.5,
      "epoch": 0.07015848473134906,
      "grad_norm": 23.99415233342749,
      "learning_rate": 4.6494491689215304e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.401734858751297,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 0.96875,
      "step": 1452
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.34375,
      "epoch": 0.07020680324700425,
      "grad_norm": 0.0,
      "learning_rate": 4.6492075763432545e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1453
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.71875,
      "epoch": 0.07025512176265945,
      "grad_norm": 0.0,
      "learning_rate": 4.6489659837649787e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 0.96875,
      "step": 1454
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 63.03125,
      "epoch": 0.07030344027831464,
      "grad_norm": 11.09615014949243,
      "learning_rate": 4.6487243911867023e-07,
      "loss": -0.0,
      "reward": 2.140625,
      "reward_std": 0.2641943693161011,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1455
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.09375,
      "epoch": 0.07035175879396985,
      "grad_norm": 0.0,
      "learning_rate": 4.6484827986084264e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1456
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 83.8125,
      "epoch": 0.07040007730962505,
      "grad_norm": 60.68941577592146,
      "learning_rate": 4.6482412060301506e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1457
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.53125,
      "epoch": 0.07044839582528024,
      "grad_norm": 0.0,
      "learning_rate": 4.6479996134518747e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1458
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 65.6875,
      "epoch": 0.07049671434093545,
      "grad_norm": 33.4121280269481,
      "learning_rate": 4.6477580208735983e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2657213509082794,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1459
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 66.5,
      "epoch": 0.07054503285659064,
      "grad_norm": 0.0,
      "learning_rate": 4.647516428295323e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.27883461117744446,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 0.9375,
      "step": 1460
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 73.625,
      "epoch": 0.07059335137224584,
      "grad_norm": 10.061826758693682,
      "learning_rate": 4.6472748357170466e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1461
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 75.125,
      "epoch": 0.07064166988790105,
      "grad_norm": 16.84263704340818,
      "learning_rate": 4.64703324313877e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.37682393193244934,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1462
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.125,
      "epoch": 0.07068998840355624,
      "grad_norm": 0.0,
      "learning_rate": 4.646791650560495e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1463
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 73.3125,
      "epoch": 0.07073830691921144,
      "grad_norm": 11.898665172111114,
      "learning_rate": 4.6465500579822185e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.29388895630836487,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 0.96875,
      "step": 1464
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.21875,
      "epoch": 0.07078662543486663,
      "grad_norm": 0.0,
      "learning_rate": 4.6463084654039426e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1465
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 65.25,
      "epoch": 0.07083494395052184,
      "grad_norm": 9.945960154427699,
      "learning_rate": 4.646066872825666e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.30315595865249634,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 0.96875,
      "step": 1466
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.625,
      "epoch": 0.07088326246617704,
      "grad_norm": 0.0,
      "learning_rate": 4.645825280247391e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1467
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.96875,
      "epoch": 0.07093158098183223,
      "grad_norm": 0.0,
      "learning_rate": 4.6455836876691145e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1468
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.0625,
      "epoch": 0.07097989949748744,
      "grad_norm": 14.562365467541992,
      "learning_rate": 4.6453420950908386e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.22152717411518097,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1469
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.65625,
      "epoch": 0.07102821801314263,
      "grad_norm": 6.660562182148717,
      "learning_rate": 4.645100502512563e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.21371951699256897,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 0.96875,
      "step": 1470
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.875,
      "epoch": 0.07107653652879783,
      "grad_norm": 0.0,
      "learning_rate": 4.644858909934287e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1471
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.09375,
      "epoch": 0.07112485504445304,
      "grad_norm": 7.019615108499829,
      "learning_rate": 4.6446173173560105e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 0.90625,
      "step": 1472
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 77.9375,
      "epoch": 0.07117317356010823,
      "grad_norm": 19.38075419496301,
      "learning_rate": 4.6443757247777346e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.4697399437427521,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1473
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 71.6875,
      "epoch": 0.07122149207576343,
      "grad_norm": 11.950654576421606,
      "learning_rate": 4.644134132199459e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.3399612307548523,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 0.96875,
      "step": 1474
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.78125,
      "epoch": 0.07126981059141863,
      "grad_norm": 6.758828801227606,
      "learning_rate": 4.6438925396211824e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1475
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.0,
      "epoch": 0.07131812910707383,
      "grad_norm": 0.0,
      "learning_rate": 4.6436509470429065e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1476
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.34375,
      "epoch": 0.07136644762272903,
      "grad_norm": 0.0,
      "learning_rate": 4.6434093544646306e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1477
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.90625,
      "epoch": 0.07141476613838423,
      "grad_norm": 0.0,
      "learning_rate": 4.643167761886355e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1478
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 66.25,
      "epoch": 0.07146308465403943,
      "grad_norm": 10.637181364108988,
      "learning_rate": 4.6429261693080784e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2970154285430908,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1479
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.65625,
      "epoch": 0.07151140316969462,
      "grad_norm": 0.0,
      "learning_rate": 4.642684576729803e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1480
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.03125,
      "epoch": 0.07155972168534983,
      "grad_norm": 0.0,
      "learning_rate": 4.6424429841515267e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1481
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.5,
      "epoch": 0.07160804020100503,
      "grad_norm": 0.0,
      "learning_rate": 4.64220139157325e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1482
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.6875,
      "epoch": 0.07165635871666022,
      "grad_norm": 0.0,
      "learning_rate": 4.641959798994975e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1483
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 78.09375,
      "epoch": 0.07170467723231543,
      "grad_norm": 22.95175691432007,
      "learning_rate": 4.6417182064166985e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.6241660118103027,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1484
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.9375,
      "epoch": 0.07175299574797062,
      "grad_norm": 0.0,
      "learning_rate": 4.6414766138384227e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1485
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.46875,
      "epoch": 0.07180131426362582,
      "grad_norm": 0.0,
      "learning_rate": 4.6412350212601463e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1486
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.375,
      "epoch": 0.07184963277928103,
      "grad_norm": 0.0,
      "learning_rate": 4.640993428681871e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1487
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.5625,
      "epoch": 0.07189795129493622,
      "grad_norm": 6.996885651089024,
      "learning_rate": 4.6407518361035945e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1488
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.875,
      "epoch": 0.07194626981059142,
      "grad_norm": 0.0,
      "learning_rate": 4.6405102435253187e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1489
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 74.0,
      "epoch": 0.07199458832624661,
      "grad_norm": 16.75274367288855,
      "learning_rate": 4.640268650947043e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.38816189765930176,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1490
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 74.71875,
      "epoch": 0.07204290684190182,
      "grad_norm": 11.927502577661793,
      "learning_rate": 4.640027058368767e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1491
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 65.53125,
      "epoch": 0.07209122535755702,
      "grad_norm": 9.635789977731376,
      "learning_rate": 4.6397854657904906e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.32666343450546265,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1492
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 74.8125,
      "epoch": 0.07213954387321221,
      "grad_norm": 13.208097639551152,
      "learning_rate": 4.6395438732122147e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.3845272362232208,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1493
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.34375,
      "epoch": 0.07218786238886742,
      "grad_norm": 0.0,
      "learning_rate": 4.639302280633939e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1494
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 73.375,
      "epoch": 0.0722361809045226,
      "grad_norm": 36.98502424314825,
      "learning_rate": 4.6390606880556624e-07,
      "loss": 0.0,
      "reward": 1.09375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.0625,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 1495
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.8125,
      "epoch": 0.07228449942017781,
      "grad_norm": 0.0,
      "learning_rate": 4.6388190954773866e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1496
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.15625,
      "epoch": 0.07233281793583302,
      "grad_norm": 0.0,
      "learning_rate": 4.6385775028991107e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1497
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 66.34375,
      "epoch": 0.0723811364514882,
      "grad_norm": 9.942362837165886,
      "learning_rate": 4.638335910320835e-07,
      "loss": -0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1498
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.5,
      "epoch": 0.07242945496714341,
      "grad_norm": 0.0,
      "learning_rate": 4.6380943177425585e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1499
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.875,
      "epoch": 0.0724777734827986,
      "grad_norm": 0.0,
      "learning_rate": 4.637852725164283e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1500
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 82.0625,
      "epoch": 0.0725260919984538,
      "grad_norm": 17.52257791075178,
      "learning_rate": 4.6376111325860067e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.4242810606956482,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1501
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 67.0625,
      "epoch": 0.07257441051410901,
      "grad_norm": 43.00911903681182,
      "learning_rate": 4.637369540007731e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1502
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 77.8125,
      "epoch": 0.0726227290297642,
      "grad_norm": 14.470212876758504,
      "learning_rate": 4.637127947429455e-07,
      "loss": -0.0,
      "reward": 1.75,
      "reward_std": 0.43768274784088135,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1503
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.53125,
      "epoch": 0.0726710475454194,
      "grad_norm": 0.0,
      "learning_rate": 4.6368863548511786e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1504
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 80.46875,
      "epoch": 0.0727193660610746,
      "grad_norm": 16.64493062583707,
      "learning_rate": 4.636644762272903e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.397722065448761,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1505
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 78.96875,
      "epoch": 0.0727676845767298,
      "grad_norm": 37.034465926963584,
      "learning_rate": 4.6364031696946264e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.4834761321544647,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1506
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 74.9375,
      "epoch": 0.072816003092385,
      "grad_norm": 13.109763089966728,
      "learning_rate": 4.636161577116351e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.3845272362232208,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1507
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.3125,
      "epoch": 0.0728643216080402,
      "grad_norm": 9.527064647706863,
      "learning_rate": 4.6359199845380746e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.3255898952484131,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1508
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 65.15625,
      "epoch": 0.0729126401236954,
      "grad_norm": 9.353842052228156,
      "learning_rate": 4.635678391959799e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1509
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 85.8125,
      "epoch": 0.07296095863935059,
      "grad_norm": 10.599790626683564,
      "learning_rate": 4.635436799381523e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1510
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.25,
      "epoch": 0.0730092771550058,
      "grad_norm": 0.0,
      "learning_rate": 4.635195206803247e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1511
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.9375,
      "epoch": 0.073057595670661,
      "grad_norm": 0.0,
      "learning_rate": 4.6349536142249706e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1512
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 62.8125,
      "epoch": 0.07310591418631619,
      "grad_norm": 6.277452586929866,
      "learning_rate": 4.6347120216466953e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.2435920089483261,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1513
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.84375,
      "epoch": 0.0731542327019714,
      "grad_norm": 0.0,
      "learning_rate": 4.634470429068419e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1514
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.5,
      "epoch": 0.07320255121762659,
      "grad_norm": 8.66155756729194,
      "learning_rate": 4.6342288364901425e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1515
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.6875,
      "epoch": 0.07325086973328179,
      "grad_norm": 0.0,
      "learning_rate": 4.6339872439118667e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 1516
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 72.46875,
      "epoch": 0.073299188248937,
      "grad_norm": 26.43604315555516,
      "learning_rate": 4.633745651333591e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1517
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.9375,
      "epoch": 0.07334750676459219,
      "grad_norm": 0.0,
      "learning_rate": 4.633504058755315e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1518
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 74.15625,
      "epoch": 0.07339582528024739,
      "grad_norm": 28.182342722247643,
      "learning_rate": 4.6332624661770385e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1519
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.1875,
      "epoch": 0.07344414379590258,
      "grad_norm": 0.0,
      "learning_rate": 4.633020873598763e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1520
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.65625,
      "epoch": 0.07349246231155779,
      "grad_norm": 14.556694803413084,
      "learning_rate": 4.632779281020487e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1521
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.53125,
      "epoch": 0.07354078082721299,
      "grad_norm": 5.452324662735303,
      "learning_rate": 4.632537688442211e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1522
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 79.84375,
      "epoch": 0.07358909934286818,
      "grad_norm": 9.539472270308535,
      "learning_rate": 4.632296095863935e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.31432515382766724,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 1523
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.59375,
      "epoch": 0.07363741785852339,
      "grad_norm": 15.917347428255672,
      "learning_rate": 4.6320545032856587e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.20633062720298767,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1524
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 81.09375,
      "epoch": 0.07368573637417858,
      "grad_norm": 25.480539194130877,
      "learning_rate": 4.631812910707383e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.5375107526779175,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1525
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 65.4375,
      "epoch": 0.07373405488983378,
      "grad_norm": 14.941732186148727,
      "learning_rate": 4.6315713181291064e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1526
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 89.5625,
      "epoch": 0.07378237340548899,
      "grad_norm": 13.751097752380657,
      "learning_rate": 4.631329725550831e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.34988388419151306,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1527
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 83.5625,
      "epoch": 0.07383069192114418,
      "grad_norm": 16.795873093287575,
      "learning_rate": 4.6310881329725547e-07,
      "loss": -0.0,
      "reward": 1.140625,
      "reward_std": 0.15580293536186218,
      "rewards/accuracy_reward": 0.125,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 1528
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 71.375,
      "epoch": 0.07387901043679938,
      "grad_norm": 23.52585150712569,
      "learning_rate": 4.630846540394279e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1529
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.5,
      "epoch": 0.07392732895245457,
      "grad_norm": 0.0,
      "learning_rate": 4.630604947816003e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1530
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 66.09375,
      "epoch": 0.07397564746810978,
      "grad_norm": 12.62610308397582,
      "learning_rate": 4.630363355237727e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2302001267671585,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 1531
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 70.125,
      "epoch": 0.07402396598376498,
      "grad_norm": 76.74271874790134,
      "learning_rate": 4.6301217626594507e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1532
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 63.0,
      "epoch": 0.07407228449942017,
      "grad_norm": 42.64178709658194,
      "learning_rate": 4.6298801700811754e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.2561880350112915,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1533
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 68.75,
      "epoch": 0.07412060301507538,
      "grad_norm": 13.25360636906237,
      "learning_rate": 4.629638577502899e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.2780436873435974,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1534
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 77.1875,
      "epoch": 0.07416892153073057,
      "grad_norm": 14.885169615111227,
      "learning_rate": 4.6293969849246226e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.4408525228500366,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1535
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.75,
      "epoch": 0.07421724004638577,
      "grad_norm": 0.0,
      "learning_rate": 4.629155392346347e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1536
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.625,
      "epoch": 0.07426555856204098,
      "grad_norm": 20.719793110789222,
      "learning_rate": 4.628913799768071e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.30210787057876587,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1537
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 63.71875,
      "epoch": 0.07431387707769617,
      "grad_norm": 9.550171949487614,
      "learning_rate": 4.628672207189795e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1538
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.15625,
      "epoch": 0.07436219559335137,
      "grad_norm": 5.005267170866048,
      "learning_rate": 4.6284306146115186e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.24251843988895416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1539
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.90625,
      "epoch": 0.07441051410900656,
      "grad_norm": 0.0,
      "learning_rate": 4.6281890220332433e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1540
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.03125,
      "epoch": 0.07445883262466177,
      "grad_norm": 0.0,
      "learning_rate": 4.627947429454967e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1541
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.25,
      "epoch": 0.07450715114031697,
      "grad_norm": 5.101625922830942,
      "learning_rate": 4.627705836876691e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1542
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 82.9375,
      "epoch": 0.07455546965597216,
      "grad_norm": 6.3727452694910625,
      "learning_rate": 4.627464244298415e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1543
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.28125,
      "epoch": 0.07460378817162737,
      "grad_norm": 0.0,
      "learning_rate": 4.6272226517201393e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1544
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 76.53125,
      "epoch": 0.07465210668728256,
      "grad_norm": 26.73387516405165,
      "learning_rate": 4.626981059141863e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1545
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 84.03125,
      "epoch": 0.07470042520293776,
      "grad_norm": 19.628278014465717,
      "learning_rate": 4.6267394665635865e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1546
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.4375,
      "epoch": 0.07474874371859297,
      "grad_norm": 0.0,
      "learning_rate": 4.626497873985311e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 1547
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.03125,
      "epoch": 0.07479706223424816,
      "grad_norm": 0.0,
      "learning_rate": 4.626256281407035e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1548
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.84375,
      "epoch": 0.07484538074990336,
      "grad_norm": 7.678041578620277,
      "learning_rate": 4.626014688828759e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1549
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 95.96875,
      "epoch": 0.07489369926555856,
      "grad_norm": 13.706873663030853,
      "learning_rate": 4.625773096250483e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.30210787057876587,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1550
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.21875,
      "epoch": 0.07494201778121376,
      "grad_norm": 0.0,
      "learning_rate": 4.625531503672207e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1551
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 97.25,
      "epoch": 0.07499033629686896,
      "grad_norm": 9.515361908560084,
      "learning_rate": 4.625289911093931e-07,
      "loss": 0.0,
      "reward": 1.28125,
      "reward_std": 0.28838545083999634,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 1552
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 93.0,
      "epoch": 0.07503865481252416,
      "grad_norm": 15.272012634465035,
      "learning_rate": 4.6250483185156555e-07,
      "loss": -0.0,
      "reward": 1.3125,
      "reward_std": 0.49441635608673096,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 1553
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.65625,
      "epoch": 0.07508697332817936,
      "grad_norm": 8.517390342324505,
      "learning_rate": 4.624806725937379e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1554
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.84375,
      "epoch": 0.07513529184383455,
      "grad_norm": 16.584457452738075,
      "learning_rate": 4.6245651333591027e-07,
      "loss": -0.0,
      "reward": 2.328125,
      "reward_std": 0.27841317653656006,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1555
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.6875,
      "epoch": 0.07518361035948976,
      "grad_norm": 9.35880291967982,
      "learning_rate": 4.624323540780827e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1556
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 75.90625,
      "epoch": 0.07523192887514496,
      "grad_norm": 18.281748803180275,
      "learning_rate": 4.624081948202551e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.3877224326133728,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1557
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 76.625,
      "epoch": 0.07528024739080015,
      "grad_norm": 9.040145035130658,
      "learning_rate": 4.623840355624275e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.29854315519332886,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1558
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 79.6875,
      "epoch": 0.07532856590645536,
      "grad_norm": 24.51621440030875,
      "learning_rate": 4.6235987630459987e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.2651395797729492,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1559
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.6875,
      "epoch": 0.07537688442211055,
      "grad_norm": 7.049788281644529,
      "learning_rate": 4.6233571704677234e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1560
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 76.15625,
      "epoch": 0.07542520293776575,
      "grad_norm": 12.583482836786722,
      "learning_rate": 4.623115577889447e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.2883424758911133,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1561
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 69.65625,
      "epoch": 0.07547352145342096,
      "grad_norm": 0.0,
      "learning_rate": 4.622873985311171e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1562
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 69.15625,
      "epoch": 0.07552183996907615,
      "grad_norm": 20.991433586793296,
      "learning_rate": 4.622632392732895e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1563
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.1875,
      "epoch": 0.07557015848473135,
      "grad_norm": 0.0,
      "learning_rate": 4.6223908001546194e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 1564
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.09375,
      "epoch": 0.07561847700038654,
      "grad_norm": 9.174868420725726,
      "learning_rate": 4.622149207576343e-07,
      "loss": -0.0,
      "reward": 2.421875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 1565
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.875,
      "epoch": 0.07566679551604175,
      "grad_norm": 0.0,
      "learning_rate": 4.6219076149980666e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 1566
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 64.0,
      "epoch": 0.07571511403169695,
      "grad_norm": 8.093771164412699,
      "learning_rate": 4.621666022419791e-07,
      "loss": -0.0,
      "reward": 2.34375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1567
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 79.25,
      "epoch": 0.07576343254735214,
      "grad_norm": 18.37495859497479,
      "learning_rate": 4.621424429841515e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.5550205111503601,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1568
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.8125,
      "epoch": 0.07581175106300735,
      "grad_norm": 0.0,
      "learning_rate": 4.621182837263239e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1569
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 71.40625,
      "epoch": 0.07586006957866254,
      "grad_norm": 16.628806636710397,
      "learning_rate": 4.620941244684963e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 1570
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 85.53125,
      "epoch": 0.07590838809431774,
      "grad_norm": 5.707547814866577,
      "learning_rate": 4.6206996521066873e-07,
      "loss": -0.0,
      "reward": 2.4375,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 1571
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 66.71875,
      "epoch": 0.07595670660997295,
      "grad_norm": 18.464551288155576,
      "learning_rate": 4.620458059528411e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.23827512562274933,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1572
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.53125,
      "epoch": 0.07600502512562814,
      "grad_norm": 8.410966669657876,
      "learning_rate": 4.6202164669501355e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.2824693024158478,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1573
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 79.25,
      "epoch": 0.07605334364128334,
      "grad_norm": 4.519248004894048,
      "learning_rate": 4.619974874371859e-07,
      "loss": -0.0,
      "reward": 1.40625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1574
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.375,
      "epoch": 0.07610166215693853,
      "grad_norm": 0.0,
      "learning_rate": 4.619733281793583e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1575
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 94.59375,
      "epoch": 0.07614998067259374,
      "grad_norm": 23.71427579369658,
      "learning_rate": 4.619491689215307e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.5810320377349854,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1576
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 84.40625,
      "epoch": 0.07619829918824894,
      "grad_norm": 18.723725321721016,
      "learning_rate": 4.619250096637031e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.4150616228580475,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1577
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 72.34375,
      "epoch": 0.07624661770390413,
      "grad_norm": 12.910463992358569,
      "learning_rate": 4.619008504058755e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.30721208453178406,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1578
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 86.9375,
      "epoch": 0.07629493621955934,
      "grad_norm": 54.35873600842153,
      "learning_rate": 4.618766911480479e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.4209679663181305,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1579
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.4375,
      "epoch": 0.07634325473521453,
      "grad_norm": 0.0,
      "learning_rate": 4.6185253189022034e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1580
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 65.03125,
      "epoch": 0.07639157325086973,
      "grad_norm": 20.96754599663897,
      "learning_rate": 4.618283726323927e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1581
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 71.21875,
      "epoch": 0.07643989176652494,
      "grad_norm": 14.1589184847923,
      "learning_rate": 4.618042133745651e-07,
      "loss": 0.0,
      "reward": 2.1875,
      "reward_std": 0.37457847595214844,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1582
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 83.3125,
      "epoch": 0.07648821028218013,
      "grad_norm": 9.985467278517103,
      "learning_rate": 4.6178005411673753e-07,
      "loss": -0.0,
      "reward": 1.546875,
      "reward_std": 0.4150615930557251,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1583
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.59375,
      "epoch": 0.07653652879783533,
      "grad_norm": 0.0,
      "learning_rate": 4.6175589485890995e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1584
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 84.15625,
      "epoch": 0.07658484731349052,
      "grad_norm": 6.533950961856368,
      "learning_rate": 4.617317356010823e-07,
      "loss": 0.0,
      "reward": 1.40625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1585
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 95.0,
      "epoch": 0.07663316582914573,
      "grad_norm": 12.030161899234207,
      "learning_rate": 4.6170757634325467e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.14562058448791504,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1586
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.6875,
      "epoch": 0.07668148434480093,
      "grad_norm": 11.258532897312401,
      "learning_rate": 4.6168341708542713e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1587
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.75,
      "epoch": 0.07672980286045612,
      "grad_norm": 0.0,
      "learning_rate": 4.616592578275995e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1588
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.78125,
      "epoch": 0.07677812137611133,
      "grad_norm": 33.563720516630184,
      "learning_rate": 4.616350985697719e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1589
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 87.8125,
      "epoch": 0.07682643989176652,
      "grad_norm": 9.802952353990792,
      "learning_rate": 4.616109393119443e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.43191659450531006,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1590
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.53125,
      "epoch": 0.07687475840742172,
      "grad_norm": 19.130207480901717,
      "learning_rate": 4.6158678005411674e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.3051002025604248,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1591
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 92.75,
      "epoch": 0.07692307692307693,
      "grad_norm": 13.352218580803536,
      "learning_rate": 4.615626207962891e-07,
      "loss": -0.0,
      "reward": 1.90625,
      "reward_std": 0.4387563169002533,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1592
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.875,
      "epoch": 0.07697139543873212,
      "grad_norm": 0.0,
      "learning_rate": 4.6153846153846156e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.18174266815185547,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 0.96875,
      "step": 1593
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.40625,
      "epoch": 0.07701971395438732,
      "grad_norm": 10.17606007147161,
      "learning_rate": 4.615143022806339e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1594
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 92.6875,
      "epoch": 0.07706803247004251,
      "grad_norm": 10.137148761313819,
      "learning_rate": 4.6149014302280634e-07,
      "loss": -0.0,
      "reward": 2.09375,
      "reward_std": 0.3389131426811218,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1595
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.0,
      "epoch": 0.07711635098569772,
      "grad_norm": 31.500787245172997,
      "learning_rate": 4.614659837649787e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 1596
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.65625,
      "epoch": 0.07716466950135292,
      "grad_norm": 0.0,
      "learning_rate": 4.614418245071511e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1597
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 104.75,
      "epoch": 0.07721298801700811,
      "grad_norm": 9.374389961268987,
      "learning_rate": 4.614176652493235e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2962459325790405,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1598
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.71875,
      "epoch": 0.07726130653266332,
      "grad_norm": 10.756623308516328,
      "learning_rate": 4.613935059914959e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.34352827072143555,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1599
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 99.4375,
      "epoch": 0.07730962504831851,
      "grad_norm": 7.820648433268644,
      "learning_rate": 4.6136934673366835e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1600
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 108.6875,
      "epoch": 0.07735794356397371,
      "grad_norm": 7.940354471235407,
      "learning_rate": 4.613451874758407e-07,
      "loss": 0.0,
      "reward": 1.5,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1601
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 96.09375,
      "epoch": 0.07740626207962892,
      "grad_norm": 20.2655446998582,
      "learning_rate": 4.613210282180131e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.27564918994903564,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1602
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 102.6875,
      "epoch": 0.07745458059528411,
      "grad_norm": 24.20079082307071,
      "learning_rate": 4.6129686896018554e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.43872106075286865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1603
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 92.59375,
      "epoch": 0.07750289911093931,
      "grad_norm": 10.457865683780788,
      "learning_rate": 4.6127270970235795e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.36722296476364136,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1604
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 101.21875,
      "epoch": 0.0775512176265945,
      "grad_norm": 9.061043886559046,
      "learning_rate": 4.612485504445303e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.41398805379867554,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1605
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.75,
      "epoch": 0.07759953614224971,
      "grad_norm": 0.0,
      "learning_rate": 4.612243911867027e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1606
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.1875,
      "epoch": 0.07764785465790491,
      "grad_norm": 0.0,
      "learning_rate": 4.6120023192887514e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1607
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.34375,
      "epoch": 0.0776961731735601,
      "grad_norm": 0.0,
      "learning_rate": 4.611760726710475e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1608
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.4375,
      "epoch": 0.07774449168921531,
      "grad_norm": 0.0,
      "learning_rate": 4.611519134132199e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1609
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 94.34375,
      "epoch": 0.0777928102048705,
      "grad_norm": 9.188851504594236,
      "learning_rate": 4.6112775415539233e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.36722296476364136,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1610
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 122.03125,
      "epoch": 0.0778411287205257,
      "grad_norm": 6.194860903299831,
      "learning_rate": 4.6110359489756474e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1611
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.96875,
      "epoch": 0.07788944723618091,
      "grad_norm": 0.0,
      "learning_rate": 4.610794356397371e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1612
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.4375,
      "epoch": 0.0779377657518361,
      "grad_norm": 17.785702327810498,
      "learning_rate": 4.6105527638190957e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1613
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 108.71875,
      "epoch": 0.0779860842674913,
      "grad_norm": 4.928886949325705,
      "learning_rate": 4.6103111712408193e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1614
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 113.5,
      "epoch": 0.0780344027831465,
      "grad_norm": 13.468328119101646,
      "learning_rate": 4.6100695786625434e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.39414215087890625,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1615
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 97.53125,
      "epoch": 0.0780827212988017,
      "grad_norm": 8.901083081259161,
      "learning_rate": 4.609827986084267e-07,
      "loss": -0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1616
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 87.75,
      "epoch": 0.0781310398144569,
      "grad_norm": 9.72993649167661,
      "learning_rate": 4.609586393505991e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1617
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 117.0,
      "epoch": 0.0781793583301121,
      "grad_norm": 12.078507363443514,
      "learning_rate": 4.6093448009277153e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1618
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.90625,
      "epoch": 0.0782276768457673,
      "grad_norm": 0.0,
      "learning_rate": 4.609103208349439e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 1619
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 120.03125,
      "epoch": 0.0782759953614225,
      "grad_norm": 5.782655223951448,
      "learning_rate": 4.6088616157711636e-07,
      "loss": -0.0,
      "reward": 1.625,
      "reward_std": 0.21371951699256897,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1620
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 129.53125,
      "epoch": 0.0783243138770777,
      "grad_norm": 18.086987319854305,
      "learning_rate": 4.608620023192887e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.47976505756378174,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1621
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 108.90625,
      "epoch": 0.0783726323927329,
      "grad_norm": 4.026453180492373,
      "learning_rate": 4.6083784306146113e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.308285653591156,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1622
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.90625,
      "epoch": 0.07842095090838809,
      "grad_norm": 0.0,
      "learning_rate": 4.6081368380363355e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1623
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 107.25,
      "epoch": 0.0784692694240433,
      "grad_norm": 30.32364115255762,
      "learning_rate": 4.6078952454580596e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.27467289566993713,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1624
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 99.9375,
      "epoch": 0.0785175879396985,
      "grad_norm": 6.842245951183688,
      "learning_rate": 4.607653652879783e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.2573546767234802,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1625
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.59375,
      "epoch": 0.07856590645535369,
      "grad_norm": 8.665929104694976,
      "learning_rate": 4.607412060301507e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.29713040590286255,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1626
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 104.15625,
      "epoch": 0.0786142249710089,
      "grad_norm": 41.50024298908064,
      "learning_rate": 4.6071704677232315e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.3344728350639343,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 1627
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 123.0625,
      "epoch": 0.07866254348666409,
      "grad_norm": 4.845377368132762,
      "learning_rate": 4.606928875144955e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1628
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 99.75,
      "epoch": 0.07871086200231929,
      "grad_norm": 21.945500424039995,
      "learning_rate": 4.606687282566679e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.3463020324707031,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1629
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 96.59375,
      "epoch": 0.0787591805179745,
      "grad_norm": 20.01504259414508,
      "learning_rate": 4.6064456899884034e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.463489294052124,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1630
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 99.59375,
      "epoch": 0.07880749903362969,
      "grad_norm": 11.774608784796335,
      "learning_rate": 4.6062040974101275e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.30935919284820557,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1631
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 107.1875,
      "epoch": 0.07885581754928489,
      "grad_norm": 28.764177271463712,
      "learning_rate": 4.605962504831851e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1632
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 107.75,
      "epoch": 0.07890413606494008,
      "grad_norm": 7.847952243942356,
      "learning_rate": 4.605720912253576e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.4336168169975281,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1633
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 131.28125,
      "epoch": 0.07895245458059529,
      "grad_norm": 20.524463868036502,
      "learning_rate": 4.6054793196752994e-07,
      "loss": 0.0,
      "reward": 1.484375,
      "reward_std": 0.5198932886123657,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1634
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 112.78125,
      "epoch": 0.07900077309625049,
      "grad_norm": 7.207398589299127,
      "learning_rate": 4.6052377270970235e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1635
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 124.0,
      "epoch": 0.07904909161190568,
      "grad_norm": 7.683308619057884,
      "learning_rate": 4.604996134518747e-07,
      "loss": -0.0,
      "reward": 1.40625,
      "reward_std": 0.28104937076568604,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1636
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 106.1875,
      "epoch": 0.07909741012756089,
      "grad_norm": 7.844491395191924,
      "learning_rate": 4.6047545419404713e-07,
      "loss": -0.0,
      "reward": 1.4375,
      "reward_std": 0.19266104698181152,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1637
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 110.3125,
      "epoch": 0.07914572864321608,
      "grad_norm": 12.41020592263507,
      "learning_rate": 4.6045129493621954e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.3846342861652374,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1638
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 118.625,
      "epoch": 0.07919404715887128,
      "grad_norm": 5.095040817286084,
      "learning_rate": 4.604271356783919e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1639
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.90625,
      "epoch": 0.07924236567452649,
      "grad_norm": 0.0,
      "learning_rate": 4.6040297642056437e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1640
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.375,
      "epoch": 0.07929068419018168,
      "grad_norm": 17.27363309450412,
      "learning_rate": 4.6037881716273673e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.27102428674697876,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1641
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 121.4375,
      "epoch": 0.07933900270583688,
      "grad_norm": 12.553798628562221,
      "learning_rate": 4.6035465790490914e-07,
      "loss": -0.0,
      "reward": 1.625,
      "reward_std": 0.40173304080963135,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1642
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 118.84375,
      "epoch": 0.07938732122149207,
      "grad_norm": 11.168980696495183,
      "learning_rate": 4.6033049864708156e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.5271093845367432,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1643
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 119.6875,
      "epoch": 0.07943563973714728,
      "grad_norm": 10.273444216346327,
      "learning_rate": 4.6030633938925397e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.20785759389400482,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1644
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 99.96875,
      "epoch": 0.07948395825280248,
      "grad_norm": 3.9910564724083333,
      "learning_rate": 4.6028218013142633e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1645
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 102.25,
      "epoch": 0.07953227676845767,
      "grad_norm": 5.1439823473461965,
      "learning_rate": 4.6025802087359874e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1646
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 110.34375,
      "epoch": 0.07958059528411288,
      "grad_norm": 20.893786814604006,
      "learning_rate": 4.6023386161577116e-07,
      "loss": -0.0,
      "reward": 1.1875,
      "reward_std": 0.32666346430778503,
      "rewards/accuracy_reward": 0.125,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 1647
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 101.40625,
      "epoch": 0.07962891379976807,
      "grad_norm": 7.026491164663187,
      "learning_rate": 4.602097023579435e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.3029785752296448,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1648
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 95.3125,
      "epoch": 0.07967723231542327,
      "grad_norm": 6.081131788168349,
      "learning_rate": 4.6018554310011593e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1649
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 101.34375,
      "epoch": 0.07972555083107848,
      "grad_norm": 12.98695382123606,
      "learning_rate": 4.6016138384228835e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.40785470604896545,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1650
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 106.71875,
      "epoch": 0.07977386934673367,
      "grad_norm": 17.41638567281985,
      "learning_rate": 4.6013722458446076e-07,
      "loss": -0.0,
      "reward": 1.546875,
      "reward_std": 0.6487216353416443,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1651
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.15625,
      "epoch": 0.07982218786238887,
      "grad_norm": 0.0,
      "learning_rate": 4.601130653266331e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 1652
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 106.0625,
      "epoch": 0.07987050637804406,
      "grad_norm": 12.80585452950731,
      "learning_rate": 4.600889060688056e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.39135709404945374,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1653
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.53125,
      "epoch": 0.07991882489369927,
      "grad_norm": 0.0,
      "learning_rate": 4.6006474681097795e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1654
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.3125,
      "epoch": 0.07996714340935447,
      "grad_norm": 0.0,
      "learning_rate": 4.6004058755315036e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1655
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 98.875,
      "epoch": 0.08001546192500966,
      "grad_norm": 5.154805118957773,
      "learning_rate": 4.600164282953227e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1656
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 106.59375,
      "epoch": 0.08006378044066487,
      "grad_norm": 4.116957116889562,
      "learning_rate": 4.599922690374952e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1657
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 104.96875,
      "epoch": 0.08011209895632006,
      "grad_norm": 4.661368010515841,
      "learning_rate": 4.5996810977966755e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1658
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.21875,
      "epoch": 0.08016041747197526,
      "grad_norm": 8.568231207817044,
      "learning_rate": 4.599439505218399e-07,
      "loss": -0.0,
      "reward": 1.6875,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1659
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 107.78125,
      "epoch": 0.08020873598763047,
      "grad_norm": 25.18155113572449,
      "learning_rate": 4.599197912640124e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.5068925023078918,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1660
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 102.65625,
      "epoch": 0.08025705450328566,
      "grad_norm": 18.26476497498305,
      "learning_rate": 4.5989563200618474e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.4742468595504761,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1661
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 89.0,
      "epoch": 0.08030537301894086,
      "grad_norm": 11.096225462141703,
      "learning_rate": 4.5987147274835715e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.37405285239219666,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1662
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.4375,
      "epoch": 0.08035369153459605,
      "grad_norm": 0.0,
      "learning_rate": 4.5984731349052956e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1663
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 98.1875,
      "epoch": 0.08040201005025126,
      "grad_norm": 15.061082026223668,
      "learning_rate": 4.59823154232702e-07,
      "loss": -0.0,
      "reward": 1.734375,
      "reward_std": 0.3457430303096771,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1664
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 115.0,
      "epoch": 0.08045032856590646,
      "grad_norm": 16.97540233755971,
      "learning_rate": 4.5979899497487434e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.39774754643440247,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1665
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 96.4375,
      "epoch": 0.08049864708156165,
      "grad_norm": 10.239230859638717,
      "learning_rate": 4.5977483571704675e-07,
      "loss": -0.0,
      "reward": 2.34375,
      "reward_std": 0.24359199404716492,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1666
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 108.59375,
      "epoch": 0.08054696559721686,
      "grad_norm": 10.9276750363965,
      "learning_rate": 4.5975067645921917e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.27564918994903564,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1667
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 93.5625,
      "epoch": 0.08059528411287205,
      "grad_norm": 13.802803499871878,
      "learning_rate": 4.597265172013915e-07,
      "loss": -0.0,
      "reward": 1.6875,
      "reward_std": 0.3563244640827179,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1668
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 109.125,
      "epoch": 0.08064360262852725,
      "grad_norm": 11.969938826699115,
      "learning_rate": 4.5970235794356394e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.35776790976524353,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1669
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 102.1875,
      "epoch": 0.08069192114418246,
      "grad_norm": 10.64080518387802,
      "learning_rate": 4.5967819868573635e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1670
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 109.34375,
      "epoch": 0.08074023965983765,
      "grad_norm": 22.143572382543013,
      "learning_rate": 4.5965403942790877e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.5191493630409241,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1671
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 100.25,
      "epoch": 0.08078855817549285,
      "grad_norm": 9.590772595616562,
      "learning_rate": 4.5962988017008113e-07,
      "loss": -0.0,
      "reward": 2.078125,
      "reward_std": 0.34717273712158203,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1672
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 110.75,
      "epoch": 0.08083687669114804,
      "grad_norm": 7.608662863908091,
      "learning_rate": 4.596057209122536e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.36295419931411743,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1673
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.375,
      "epoch": 0.08088519520680325,
      "grad_norm": 0.0,
      "learning_rate": 4.5958156165442595e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1674
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 86.46875,
      "epoch": 0.08093351372245845,
      "grad_norm": 25.95804619309545,
      "learning_rate": 4.5955740239659837e-07,
      "loss": -0.0,
      "reward": 2.078125,
      "reward_std": 0.43556106090545654,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1675
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 102.25,
      "epoch": 0.08098183223811364,
      "grad_norm": 7.569934737562381,
      "learning_rate": 4.5953324313877073e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.3006556034088135,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1676
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 110.9375,
      "epoch": 0.08103015075376885,
      "grad_norm": 11.288950447711061,
      "learning_rate": 4.595090838809432e-07,
      "loss": 0.0,
      "reward": 1.484375,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1677
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 102.15625,
      "epoch": 0.08107846926942404,
      "grad_norm": 5.786753992386029,
      "learning_rate": 4.5948492462311556e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 1678
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 97.0625,
      "epoch": 0.08112678778507924,
      "grad_norm": 14.262662903631087,
      "learning_rate": 4.594607653652879e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.25069957971572876,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1679
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 114.3125,
      "epoch": 0.08117510630073445,
      "grad_norm": 6.2560396716037445,
      "learning_rate": 4.594366061074604e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1680
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 108.03125,
      "epoch": 0.08122342481638964,
      "grad_norm": 8.338319475363551,
      "learning_rate": 4.5941244684963274e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.4171832799911499,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1681
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.5625,
      "epoch": 0.08127174333204484,
      "grad_norm": 8.691497407797732,
      "learning_rate": 4.5938828759180516e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.29713037610054016,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1682
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 103.09375,
      "epoch": 0.08132006184770003,
      "grad_norm": 7.034890864124357,
      "learning_rate": 4.5936412833397757e-07,
      "loss": -0.0,
      "reward": 2.390625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 1683
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 111.75,
      "epoch": 0.08136838036335524,
      "grad_norm": 10.502514219471289,
      "learning_rate": 4.5933996907615e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1684
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 115.5,
      "epoch": 0.08141669887901044,
      "grad_norm": 18.750201985784418,
      "learning_rate": 4.5931580981832235e-07,
      "loss": 0.0,
      "reward": 1.453125,
      "reward_std": 0.48295044898986816,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1685
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 104.8125,
      "epoch": 0.08146501739466563,
      "grad_norm": 13.790560352698614,
      "learning_rate": 4.5929165056049476e-07,
      "loss": -0.0,
      "reward": 2.25,
      "reward_std": 0.23827512562274933,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1686
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 95.96875,
      "epoch": 0.08151333591032084,
      "grad_norm": 5.333528953908158,
      "learning_rate": 4.5926749130266717e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.28566449880599976,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1687
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.40625,
      "epoch": 0.08156165442597603,
      "grad_norm": 0.0,
      "learning_rate": 4.5924333204483953e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1688
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 122.875,
      "epoch": 0.08160997294163123,
      "grad_norm": 6.558566943601671,
      "learning_rate": 4.5921917278701195e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1689
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 121.90625,
      "epoch": 0.08165829145728644,
      "grad_norm": 16.43961647165623,
      "learning_rate": 4.5919501352918436e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.29693374037742615,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1690
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 98.65625,
      "epoch": 0.08170660997294163,
      "grad_norm": 8.625349396037226,
      "learning_rate": 4.591708542713568e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.22049658000469208,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1691
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 110.03125,
      "epoch": 0.08175492848859683,
      "grad_norm": 10.843577207662287,
      "learning_rate": 4.5914669501352914e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.2715832591056824,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1692
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.9375,
      "epoch": 0.08180324700425202,
      "grad_norm": 0.0,
      "learning_rate": 4.591225357557016e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1693
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.65625,
      "epoch": 0.08185156551990723,
      "grad_norm": 0.0,
      "learning_rate": 4.5909837649787396e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1694
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 96.65625,
      "epoch": 0.08189988403556243,
      "grad_norm": 9.374038193094073,
      "learning_rate": 4.590742172400464e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.3766237795352936,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1695
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 119.90625,
      "epoch": 0.08194820255121762,
      "grad_norm": 8.017156613108494,
      "learning_rate": 4.5905005798221874e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.3185884654521942,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1696
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.21875,
      "epoch": 0.08199652106687283,
      "grad_norm": 12.44929864936802,
      "learning_rate": 4.590258987243912e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1697
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 106.65625,
      "epoch": 0.08204483958252802,
      "grad_norm": 6.751107462184243,
      "learning_rate": 4.5900173946656356e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.2651395797729492,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1698
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 106.71875,
      "epoch": 0.08209315809818322,
      "grad_norm": 7.09387050634781,
      "learning_rate": 4.589775802087359e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1699
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 103.6875,
      "epoch": 0.08214147661383843,
      "grad_norm": 19.229595593159704,
      "learning_rate": 4.589534209509084e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.28247910737991333,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1700
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 117.4375,
      "epoch": 0.08218979512949362,
      "grad_norm": 28.636202193579987,
      "learning_rate": 4.5892926169308075e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.5510252118110657,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1701
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 103.03125,
      "epoch": 0.08223811364514882,
      "grad_norm": 32.699240334281555,
      "learning_rate": 4.5890510243525317e-07,
      "loss": -0.0,
      "reward": 1.515625,
      "reward_std": 0.49705949425697327,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1702
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 126.625,
      "epoch": 0.08228643216080402,
      "grad_norm": 13.349544229215363,
      "learning_rate": 4.588809431774256e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.6164275407791138,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1703
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.03125,
      "epoch": 0.08233475067645922,
      "grad_norm": 0.0,
      "learning_rate": 4.58856783919598e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1704
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 121.03125,
      "epoch": 0.08238306919211442,
      "grad_norm": 181.19889714383467,
      "learning_rate": 4.5883262466177035e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.5214077830314636,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1705
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 122.90625,
      "epoch": 0.08243138770776962,
      "grad_norm": 18.495223544031255,
      "learning_rate": 4.5880846540394277e-07,
      "loss": -0.0,
      "reward": 2.109375,
      "reward_std": 0.3457430303096771,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1706
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 115.09375,
      "epoch": 0.08247970622342482,
      "grad_norm": 32.44071068970081,
      "learning_rate": 4.587843061461152e-07,
      "loss": -0.0,
      "reward": 1.734375,
      "reward_std": 0.5780647993087769,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1707
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 127.34375,
      "epoch": 0.08252802473908001,
      "grad_norm": 15.891605397330334,
      "learning_rate": 4.587601468882876e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.620493471622467,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1708
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.90625,
      "epoch": 0.08257634325473522,
      "grad_norm": 0.0,
      "learning_rate": 4.5873598763045996e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1709
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 116.1875,
      "epoch": 0.08262466177039042,
      "grad_norm": 24.345110520908904,
      "learning_rate": 4.5871182837263237e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.5435781478881836,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1710
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.34375,
      "epoch": 0.08267298028604561,
      "grad_norm": 0.0,
      "learning_rate": 4.586876691148048e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1711
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 112.1875,
      "epoch": 0.08272129880170082,
      "grad_norm": 0.0,
      "learning_rate": 4.5866350985697714e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1712
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 104.53125,
      "epoch": 0.082769617317356,
      "grad_norm": 8.160606806456947,
      "learning_rate": 4.586393505991496e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.3029785752296448,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1713
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 96.46875,
      "epoch": 0.08281793583301121,
      "grad_norm": 8.4850668189143,
      "learning_rate": 4.5861519134132197e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.3255898952484131,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1714
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.21875,
      "epoch": 0.08286625434866642,
      "grad_norm": 0.0,
      "learning_rate": 4.585910320834944e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1715
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 116.4375,
      "epoch": 0.0829145728643216,
      "grad_norm": 20.714248918116798,
      "learning_rate": 4.5856687282566674e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.38993722200393677,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1716
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.5,
      "epoch": 0.08296289137997681,
      "grad_norm": 6.465866036201264,
      "learning_rate": 4.585427135678392e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1717
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 109.4375,
      "epoch": 0.083011209895632,
      "grad_norm": 8.913161849651365,
      "learning_rate": 4.5851855431001157e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1718
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 128.46875,
      "epoch": 0.0830595284112872,
      "grad_norm": 22.151573464832428,
      "learning_rate": 4.5849439505218393e-07,
      "loss": -0.0,
      "reward": 1.46875,
      "reward_std": 0.5886078476905823,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1719
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 119.125,
      "epoch": 0.08310784692694241,
      "grad_norm": 11.875535582326984,
      "learning_rate": 4.584702357943564e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.4228830337524414,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1720
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.84375,
      "epoch": 0.0831561654425976,
      "grad_norm": 0.0,
      "learning_rate": 4.5844607653652876e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 1721
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.0,
      "epoch": 0.0832044839582528,
      "grad_norm": 0.0,
      "learning_rate": 4.5842191727870117e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1722
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.5,
      "epoch": 0.083252802473908,
      "grad_norm": 5.950279180792469,
      "learning_rate": 4.583977580208736e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.24464009702205658,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1723
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 96.5,
      "epoch": 0.0833011209895632,
      "grad_norm": 6.441283940292358,
      "learning_rate": 4.58373598763046e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.24359199404716492,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1724
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 114.375,
      "epoch": 0.0833494395052184,
      "grad_norm": 5.854682965587962,
      "learning_rate": 4.5834943950521836e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1725
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 101.25,
      "epoch": 0.0833977580208736,
      "grad_norm": 20.632909129367665,
      "learning_rate": 4.583252802473908e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.3945523202419281,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1726
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 117.46875,
      "epoch": 0.0834460765365288,
      "grad_norm": 8.287848582168284,
      "learning_rate": 4.583011209895632e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.3330284357070923,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 1727
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 101.71875,
      "epoch": 0.08349439505218399,
      "grad_norm": 13.048809777452052,
      "learning_rate": 4.582769617317356e-07,
      "loss": -0.0,
      "reward": 2.109375,
      "reward_std": 0.34717273712158203,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1728
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 103.34375,
      "epoch": 0.0835427135678392,
      "grad_norm": 8.674115290956607,
      "learning_rate": 4.5825280247390796e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.23243674635887146,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1729
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 124.46875,
      "epoch": 0.0835910320834944,
      "grad_norm": 8.006379388523936,
      "learning_rate": 4.582286432160804e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1730
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 97.4375,
      "epoch": 0.08363935059914959,
      "grad_norm": 16.191268815426465,
      "learning_rate": 4.582044839582528e-07,
      "loss": -0.0,
      "reward": 2.265625,
      "reward_std": 0.3234682083129883,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1731
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.6875,
      "epoch": 0.0836876691148048,
      "grad_norm": 0.0,
      "learning_rate": 4.5818032470042515e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1732
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 98.09375,
      "epoch": 0.08373598763045999,
      "grad_norm": 13.620350813372449,
      "learning_rate": 4.581561654425976e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.22017991542816162,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1733
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.90625,
      "epoch": 0.08378430614611519,
      "grad_norm": 0.0,
      "learning_rate": 4.5813200618477e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1734
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 94.6875,
      "epoch": 0.0838326246617704,
      "grad_norm": 11.939489857381718,
      "learning_rate": 4.581078469269424e-07,
      "loss": -0.0,
      "reward": 1.75,
      "reward_std": 0.3389131426811218,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1735
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 95.09375,
      "epoch": 0.08388094317742559,
      "grad_norm": 15.089314658523675,
      "learning_rate": 4.5808368766911475e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.3908922076225281,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1736
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.71875,
      "epoch": 0.08392926169308079,
      "grad_norm": 0.0,
      "learning_rate": 4.580595284112872e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1737
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 95.9375,
      "epoch": 0.08397758020873598,
      "grad_norm": 16.66142005262079,
      "learning_rate": 4.580353691534596e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.42508673667907715,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1738
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 106.375,
      "epoch": 0.08402589872439119,
      "grad_norm": 11.795072227135998,
      "learning_rate": 4.58011209895632e-07,
      "loss": 0.0,
      "reward": 1.5625,
      "reward_std": 0.32666346430778503,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1739
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.59375,
      "epoch": 0.08407421724004639,
      "grad_norm": 5.397572464656831,
      "learning_rate": 4.579870506378044e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1740
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 93.71875,
      "epoch": 0.08412253575570158,
      "grad_norm": 10.644047692016654,
      "learning_rate": 4.5796289137997677e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 1741
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.125,
      "epoch": 0.08417085427135679,
      "grad_norm": 0.0,
      "learning_rate": 4.579387321221492e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1742
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 113.96875,
      "epoch": 0.08421917278701198,
      "grad_norm": 18.216333760761113,
      "learning_rate": 4.579145728643216e-07,
      "loss": -0.0,
      "reward": 1.671875,
      "reward_std": 0.4692919850349426,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1743
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.46875,
      "epoch": 0.08426749130266718,
      "grad_norm": 0.0,
      "learning_rate": 4.57890413606494e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1744
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 92.40625,
      "epoch": 0.08431580981832239,
      "grad_norm": 27.827746603552395,
      "learning_rate": 4.5786625434866637e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.38764041662216187,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1745
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 72.96875,
      "epoch": 0.08436412833397758,
      "grad_norm": 12.277403178804821,
      "learning_rate": 4.578420950908388e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.31983357667922974,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1746
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 88.75,
      "epoch": 0.08441244684963278,
      "grad_norm": 15.248225467244792,
      "learning_rate": 4.578179358330112e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.3782707154750824,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1747
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.46875,
      "epoch": 0.08446076536528797,
      "grad_norm": 0.0,
      "learning_rate": 4.577937765751836e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1748
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 105.71875,
      "epoch": 0.08450908388094318,
      "grad_norm": 93.34767223402937,
      "learning_rate": 4.5776961731735597e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.36306124925613403,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1749
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.90625,
      "epoch": 0.08455740239659838,
      "grad_norm": 0.0,
      "learning_rate": 4.5774545805952844e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1750
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.8125,
      "epoch": 0.08460572091225357,
      "grad_norm": 8.723040442966461,
      "learning_rate": 4.577212988017008e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.26409149169921875,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1751
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.90625,
      "epoch": 0.08465403942790878,
      "grad_norm": 0.0,
      "learning_rate": 4.5769713954387316e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1752
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.9375,
      "epoch": 0.08470235794356397,
      "grad_norm": 0.0,
      "learning_rate": 4.576729802860456e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1753
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 92.875,
      "epoch": 0.08475067645921917,
      "grad_norm": 7.127772493162616,
      "learning_rate": 4.57648821028218e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 1754
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 94.28125,
      "epoch": 0.08479899497487438,
      "grad_norm": 9.963244248232227,
      "learning_rate": 4.576246617703904e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.2961486577987671,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1755
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.09375,
      "epoch": 0.08484731349052957,
      "grad_norm": 6.761336779365728,
      "learning_rate": 4.5760050251256276e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.24251843988895416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1756
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 92.8125,
      "epoch": 0.08489563200618477,
      "grad_norm": 17.181515169805902,
      "learning_rate": 4.5757634325473523e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.17598573863506317,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1757
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.3125,
      "epoch": 0.08494395052183996,
      "grad_norm": 11.374149263746153,
      "learning_rate": 4.575521839969076e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1758
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.09375,
      "epoch": 0.08499226903749517,
      "grad_norm": 0.0,
      "learning_rate": 4.5752802473908e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1759
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.28125,
      "epoch": 0.08504058755315037,
      "grad_norm": 0.0,
      "learning_rate": 4.575038654812524e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1760
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 92.65625,
      "epoch": 0.08508890606880556,
      "grad_norm": 5.225988938817571,
      "learning_rate": 4.574797062234248e-07,
      "loss": -0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 1761
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 89.59375,
      "epoch": 0.08513722458446077,
      "grad_norm": 9.387596593753335,
      "learning_rate": 4.574555469655972e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1762
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.46875,
      "epoch": 0.08518554310011596,
      "grad_norm": 14.602654732282227,
      "learning_rate": 4.574313877077696e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1763
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 98.09375,
      "epoch": 0.08523386161577116,
      "grad_norm": 12.057277031698886,
      "learning_rate": 4.57407228449942e-07,
      "loss": -0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1764
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 95.28125,
      "epoch": 0.08528218013142637,
      "grad_norm": 11.448590310048417,
      "learning_rate": 4.573830691921144e-07,
      "loss": 0.0,
      "reward": 1.40625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1765
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 96.4375,
      "epoch": 0.08533049864708156,
      "grad_norm": 14.472624106147636,
      "learning_rate": 4.573589099342868e-07,
      "loss": 0.0,
      "reward": 1.5625,
      "reward_std": 0.3471825420856476,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1766
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.125,
      "epoch": 0.08537881716273676,
      "grad_norm": 0.0,
      "learning_rate": 4.573347506764592e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1767
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 71.1875,
      "epoch": 0.08542713567839195,
      "grad_norm": 28.850436896134177,
      "learning_rate": 4.573105914186316e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.308285653591156,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1768
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 83.1875,
      "epoch": 0.08547545419404716,
      "grad_norm": 9.676764422512713,
      "learning_rate": 4.57286432160804e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.32666343450546265,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1769
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.46875,
      "epoch": 0.08552377270970236,
      "grad_norm": 0.0,
      "learning_rate": 4.5726227290297645e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1770
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 77.0,
      "epoch": 0.08557209122535755,
      "grad_norm": 10.865616540757607,
      "learning_rate": 4.572381136451488e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.3599716126918793,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1771
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 86.34375,
      "epoch": 0.08562040974101276,
      "grad_norm": 11.462295029903832,
      "learning_rate": 4.5721395438732117e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1772
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 83.84375,
      "epoch": 0.08566872825666795,
      "grad_norm": 13.634653816138401,
      "learning_rate": 4.5718979512949363e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1773
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.40625,
      "epoch": 0.08571704677232315,
      "grad_norm": 0.0,
      "learning_rate": 4.57165635871666e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1774
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 69.5625,
      "epoch": 0.08576536528797836,
      "grad_norm": 6.7142080861279965,
      "learning_rate": 4.571414766138384e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1775
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 87.46875,
      "epoch": 0.08581368380363355,
      "grad_norm": 10.28335235552911,
      "learning_rate": 4.5711731735601077e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.3198433816432953,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1776
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 76.875,
      "epoch": 0.08586200231928875,
      "grad_norm": 12.160397550420097,
      "learning_rate": 4.5709315809818323e-07,
      "loss": -0.0,
      "reward": 1.234375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.15625,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 1777
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.3125,
      "epoch": 0.08591032083494395,
      "grad_norm": 0.0,
      "learning_rate": 4.570689988403556e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1778
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.9375,
      "epoch": 0.08595863935059915,
      "grad_norm": 11.151414248416922,
      "learning_rate": 4.57044839582528e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2993341088294983,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1779
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.625,
      "epoch": 0.08600695786625435,
      "grad_norm": 11.757581264781276,
      "learning_rate": 4.570206803247004e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19999711215496063,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1780
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 81.59375,
      "epoch": 0.08605527638190955,
      "grad_norm": 26.48937415692162,
      "learning_rate": 4.569965210668728e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1781
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.40625,
      "epoch": 0.08610359489756475,
      "grad_norm": 0.0,
      "learning_rate": 4.569723618090452e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1782
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 95.84375,
      "epoch": 0.08615191341321994,
      "grad_norm": 5.8109740566264625,
      "learning_rate": 4.569482025512176e-07,
      "loss": 0.0,
      "reward": 1.40625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1783
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 60.8125,
      "epoch": 0.08620023192887515,
      "grad_norm": 10.763388499332647,
      "learning_rate": 4.5692404329339e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.3319803476333618,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1784
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.71875,
      "epoch": 0.08624855044453035,
      "grad_norm": 7.765884855668858,
      "learning_rate": 4.568998840355624e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2798827290534973,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1785
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.0,
      "completion_length": 91.375,
      "epoch": 0.08629686896018554,
      "grad_norm": 60.587482056839306,
      "learning_rate": 4.568757247777348e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.6516169309616089,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1786
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 71.40625,
      "epoch": 0.08634518747584075,
      "grad_norm": 9.745038125295352,
      "learning_rate": 4.568515655199072e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.3104073107242584,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1787
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 66.78125,
      "epoch": 0.08639350599149594,
      "grad_norm": 9.462295248843365,
      "learning_rate": 4.568274062620796e-07,
      "loss": -0.0,
      "reward": 2.296875,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1788
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.9375,
      "epoch": 0.08644182450715114,
      "grad_norm": 0.0,
      "learning_rate": 4.56803247004252e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1789
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 69.5,
      "epoch": 0.08649014302280635,
      "grad_norm": 8.862492404426652,
      "learning_rate": 4.5677908774642445e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.3945523202419281,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1790
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 82.15625,
      "epoch": 0.08653846153846154,
      "grad_norm": 72.61853665725118,
      "learning_rate": 4.567549284885968e-07,
      "loss": -0.0,
      "reward": 1.6875,
      "reward_std": 0.35959720611572266,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1791
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.96875,
      "epoch": 0.08658678005411674,
      "grad_norm": 0.0,
      "learning_rate": 4.567307692307692e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1792
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 81.96875,
      "epoch": 0.08663509856977193,
      "grad_norm": 79.76694624661603,
      "learning_rate": 4.5670660997294164e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.4142608642578125,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1793
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 74.03125,
      "epoch": 0.08668341708542714,
      "grad_norm": 10.973294544394932,
      "learning_rate": 4.56682450715114e-07,
      "loss": -0.0,
      "reward": 2.078125,
      "reward_std": 0.32407689094543457,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1794
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 85.15625,
      "epoch": 0.08673173560108234,
      "grad_norm": 35.102609965705824,
      "learning_rate": 4.566582914572864e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1795
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.875,
      "epoch": 0.08678005411673753,
      "grad_norm": 11.551009321215357,
      "learning_rate": 4.566341321994588e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.2461063265800476,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1796
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 86.8125,
      "epoch": 0.08682837263239274,
      "grad_norm": 13.33148826608209,
      "learning_rate": 4.5660997294163124e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 1797
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.15625,
      "epoch": 0.08687669114804793,
      "grad_norm": 0.0,
      "learning_rate": 4.565858136838036e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 1798
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 93.40625,
      "epoch": 0.08692500966370313,
      "grad_norm": 5.31519601672194,
      "learning_rate": 4.56561654425976e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.24251843988895416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1799
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 74.09375,
      "epoch": 0.08697332817935834,
      "grad_norm": 11.344208195227793,
      "learning_rate": 4.5653749516814843e-07,
      "loss": -0.0,
      "reward": 1.125,
      "reward_std": 0.28104937076568604,
      "rewards/accuracy_reward": 0.09375,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 1800
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.09375,
      "epoch": 0.08702164669501353,
      "grad_norm": 0.0,
      "learning_rate": 4.5651333591032084e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1801
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.1875,
      "epoch": 0.08706996521066873,
      "grad_norm": 0.0,
      "learning_rate": 4.564891766524932e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 1802
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 72.03125,
      "epoch": 0.08711828372632392,
      "grad_norm": 5.3124401931244,
      "learning_rate": 4.564650173946656e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.308285653591156,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1803
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 89.71875,
      "epoch": 0.08716660224197913,
      "grad_norm": 4.8470384524796994,
      "learning_rate": 4.5644085813683803e-07,
      "loss": -0.0,
      "reward": 2.40625,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 1804
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.5,
      "epoch": 0.08721492075763433,
      "grad_norm": 0.0,
      "learning_rate": 4.564166988790104e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 1805
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 75.25,
      "epoch": 0.08726323927328952,
      "grad_norm": 9.743026118908094,
      "learning_rate": 4.563925396211828e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.41823136806488037,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1806
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.53125,
      "epoch": 0.08731155778894473,
      "grad_norm": 0.0,
      "learning_rate": 4.563683803633552e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1807
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.65625,
      "epoch": 0.08735987630459992,
      "grad_norm": 0.0,
      "learning_rate": 4.5634422110552763e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1808
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 68.84375,
      "epoch": 0.08740819482025512,
      "grad_norm": 0.0,
      "learning_rate": 4.563200618477e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 1809
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.71875,
      "epoch": 0.08745651333591033,
      "grad_norm": 0.0,
      "learning_rate": 4.5629590258987246e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1810
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 93.78125,
      "epoch": 0.08750483185156552,
      "grad_norm": 9.299551001979788,
      "learning_rate": 4.562717433320448e-07,
      "loss": -0.0,
      "reward": 2.421875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 1811
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.21875,
      "epoch": 0.08755315036722072,
      "grad_norm": 0.0,
      "learning_rate": 4.562475840742172e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1812
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.71875,
      "epoch": 0.08760146888287591,
      "grad_norm": 6.127786473010066,
      "learning_rate": 4.5622342481638965e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1813
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 81.65625,
      "epoch": 0.08764978739853112,
      "grad_norm": 15.008040103685852,
      "learning_rate": 4.56199265558562e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.41505181789398193,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1814
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.875,
      "epoch": 0.08769810591418632,
      "grad_norm": 10.180500470180116,
      "learning_rate": 4.561751063007344e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.3029785752296448,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1815
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.53125,
      "epoch": 0.08774642442984151,
      "grad_norm": 5.569988859860578,
      "learning_rate": 4.561509470429068e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.33090677857398987,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1816
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.1875,
      "epoch": 0.08779474294549672,
      "grad_norm": 0.0,
      "learning_rate": 4.5612678778507925e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1817
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.34375,
      "epoch": 0.08784306146115191,
      "grad_norm": 0.0,
      "learning_rate": 4.561026285272516e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1818
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 93.1875,
      "epoch": 0.08789137997680711,
      "grad_norm": 12.447766956526769,
      "learning_rate": 4.56078469269424e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.3864519000053406,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1819
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 93.78125,
      "epoch": 0.08793969849246232,
      "grad_norm": 8.832946031539494,
      "learning_rate": 4.5605431001159644e-07,
      "loss": -0.0,
      "reward": 2.390625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 1820
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 103.6875,
      "epoch": 0.08798801700811751,
      "grad_norm": 18.61534219145018,
      "learning_rate": 4.5603015075376885e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.5508139133453369,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1821
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.46875,
      "epoch": 0.08803633552377271,
      "grad_norm": 15.280403402542413,
      "learning_rate": 4.560059914959412e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.2651395797729492,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1822
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.0,
      "epoch": 0.0880846540394279,
      "grad_norm": 0.0,
      "learning_rate": 4.559818322381136e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1823
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.1875,
      "epoch": 0.08813297255508311,
      "grad_norm": 9.176831348683647,
      "learning_rate": 4.5595767298028604e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23634064197540283,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1824
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 99.0625,
      "epoch": 0.08818129107073831,
      "grad_norm": 10.946225040275007,
      "learning_rate": 4.559335137224584e-07,
      "loss": 0.0,
      "reward": 1.28125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 1825
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 89.40625,
      "epoch": 0.0882296095863935,
      "grad_norm": 11.110321735500047,
      "learning_rate": 4.559093544646308e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.31432515382766724,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1826
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 87.4375,
      "epoch": 0.08827792810204871,
      "grad_norm": 12.39709980407758,
      "learning_rate": 4.5588519520680323e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.30935919284820557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1827
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.4375,
      "epoch": 0.0883262466177039,
      "grad_norm": 0.0,
      "learning_rate": 4.5586103594897564e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1828
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 96.625,
      "epoch": 0.0883745651333591,
      "grad_norm": 9.809546710915395,
      "learning_rate": 4.55836876691148e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1829
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 105.625,
      "epoch": 0.08842288364901431,
      "grad_norm": 18.98315772423421,
      "learning_rate": 4.5581271743332047e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.43236586451530457,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1830
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 100.9375,
      "epoch": 0.0884712021646695,
      "grad_norm": 13.124883311674406,
      "learning_rate": 4.5578855817549283e-07,
      "loss": 0.0,
      "reward": 1.453125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1831
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.0625,
      "epoch": 0.0885195206803247,
      "grad_norm": 0.0,
      "learning_rate": 4.557643989176652e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1832
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.6875,
      "epoch": 0.0885678391959799,
      "grad_norm": 0.0,
      "learning_rate": 4.5574023965983766e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1833
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.625,
      "epoch": 0.0886161577116351,
      "grad_norm": 0.0,
      "learning_rate": 4.5571608040201e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1834
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 90.0625,
      "epoch": 0.0886644762272903,
      "grad_norm": 27.988250812704237,
      "learning_rate": 4.5569192114418243e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.49562233686447144,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1835
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.71875,
      "epoch": 0.0887127947429455,
      "grad_norm": 0.0,
      "learning_rate": 4.556677618863548e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1836
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.875,
      "epoch": 0.0887611132586007,
      "grad_norm": 22.89247381957586,
      "learning_rate": 4.5564360262852726e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.26726123690605164,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1837
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.40625,
      "epoch": 0.08880943177425589,
      "grad_norm": 0.0,
      "learning_rate": 4.556194433706996e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1838
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 94.34375,
      "epoch": 0.0888577502899111,
      "grad_norm": 5.6073321395919855,
      "learning_rate": 4.5559528411287203e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1839
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.84375,
      "epoch": 0.0889060688055663,
      "grad_norm": 0.0,
      "learning_rate": 4.5557112485504445e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1840
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.21875,
      "epoch": 0.08895438732122149,
      "grad_norm": 0.0,
      "learning_rate": 4.5554696559721686e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1841
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.125,
      "epoch": 0.0890027058368767,
      "grad_norm": 0.0,
      "learning_rate": 4.555228063393892e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1842
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.78125,
      "epoch": 0.08905102435253189,
      "grad_norm": 5.342278313019069,
      "learning_rate": 4.5549864708156163e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1843
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 96.65625,
      "epoch": 0.08909934286818709,
      "grad_norm": 4.216714607162707,
      "learning_rate": 4.5547448782373405e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1844
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.09375,
      "epoch": 0.0891476613838423,
      "grad_norm": 24.97200703160206,
      "learning_rate": 4.554503285659064e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.3255898952484131,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1845
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.875,
      "epoch": 0.08919597989949749,
      "grad_norm": 0.0,
      "learning_rate": 4.554261693080788e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1846
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.25,
      "epoch": 0.08924429841515269,
      "grad_norm": 0.0,
      "learning_rate": 4.5540201005025124e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1847
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 88.4375,
      "epoch": 0.08929261693080788,
      "grad_norm": 5.41036953823109,
      "learning_rate": 4.5537785079242365e-07,
      "loss": -0.0,
      "reward": 2.40625,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 1848
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.21875,
      "epoch": 0.08934093544646309,
      "grad_norm": 0.0,
      "learning_rate": 4.55353691534596e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1849
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 89.9375,
      "epoch": 0.08938925396211829,
      "grad_norm": 16.27128369986106,
      "learning_rate": 4.553295322767685e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.36278265714645386,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1850
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.125,
      "epoch": 0.08943757247777348,
      "grad_norm": 5.718600093420521,
      "learning_rate": 4.5530537301894084e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.23568855226039886,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1851
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 82.40625,
      "epoch": 0.08948589099342869,
      "grad_norm": 25.684388525806202,
      "learning_rate": 4.5528121376111325e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.35307908058166504,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1852
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.5,
      "epoch": 0.08953420950908388,
      "grad_norm": 0.0,
      "learning_rate": 4.5525705450328566e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1853
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.78125,
      "epoch": 0.08958252802473908,
      "grad_norm": 0.0,
      "learning_rate": 4.55232895245458e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1854
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 86.75,
      "epoch": 0.08963084654039429,
      "grad_norm": 8.04282049587361,
      "learning_rate": 4.5520873598763044e-07,
      "loss": -0.0,
      "reward": 2.296875,
      "reward_std": 0.3235850930213928,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1855
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.84375,
      "epoch": 0.08967916505604948,
      "grad_norm": 7.697332884322218,
      "learning_rate": 4.551845767298028e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.3740372061729431,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1856
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.09375,
      "epoch": 0.08972748357170468,
      "grad_norm": 0.0,
      "learning_rate": 4.5516041747197527e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1857
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 89.15625,
      "epoch": 0.08977580208735987,
      "grad_norm": 20.33945029464952,
      "learning_rate": 4.5513625821414763e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.32407689094543457,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1858
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.71875,
      "epoch": 0.08982412060301508,
      "grad_norm": 0.0,
      "learning_rate": 4.5511209895632004e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1859
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 84.40625,
      "epoch": 0.08987243911867028,
      "grad_norm": 33.03118390884739,
      "learning_rate": 4.5508793969849245e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1860
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 93.1875,
      "epoch": 0.08992075763432547,
      "grad_norm": 8.021189156673525,
      "learning_rate": 4.5506378044066487e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1861
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.375,
      "epoch": 0.08996907614998068,
      "grad_norm": 21.022917886459933,
      "learning_rate": 4.5503962118283723e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1862
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 91.96875,
      "epoch": 0.09001739466563587,
      "grad_norm": 8.672270329018273,
      "learning_rate": 4.550154619250097e-07,
      "loss": 0.0,
      "reward": 1.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.03125,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 1863
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 93.46875,
      "epoch": 0.09006571318129107,
      "grad_norm": 7.9914581102826645,
      "learning_rate": 4.5499130266718206e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1864
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 91.03125,
      "epoch": 0.09011403169694628,
      "grad_norm": 6.502114975594971,
      "learning_rate": 4.549671434093544e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1865
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 88.5,
      "epoch": 0.09016235021260147,
      "grad_norm": 15.826683613522905,
      "learning_rate": 4.5494298415152683e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.3503679633140564,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1866
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.78125,
      "epoch": 0.09021066872825667,
      "grad_norm": 0.0,
      "learning_rate": 4.5491882489369924e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1867
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.03125,
      "epoch": 0.09025898724391186,
      "grad_norm": 0.0,
      "learning_rate": 4.5489466563587166e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1868
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.96875,
      "epoch": 0.09030730575956707,
      "grad_norm": 0.0,
      "learning_rate": 4.54870506378044e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1869
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.28125,
      "epoch": 0.09035562427522227,
      "grad_norm": 0.0,
      "learning_rate": 4.548463471202165e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1870
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 82.9375,
      "epoch": 0.09040394279087746,
      "grad_norm": 35.36011515987192,
      "learning_rate": 4.5482218786238885e-07,
      "loss": 0.0,
      "reward": 1.40625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 1871
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.59375,
      "epoch": 0.09045226130653267,
      "grad_norm": 0.0,
      "learning_rate": 4.5479802860456126e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1872
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 77.6875,
      "epoch": 0.09050057982218786,
      "grad_norm": 15.601028338635171,
      "learning_rate": 4.5477386934673367e-07,
      "loss": -0.0,
      "reward": 2.3125,
      "reward_std": 0.3808925449848175,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1873
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 94.8125,
      "epoch": 0.09054889833784306,
      "grad_norm": 7.284824739335697,
      "learning_rate": 4.5474971008890603e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1874
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.4375,
      "epoch": 0.09059721685349827,
      "grad_norm": 0.0,
      "learning_rate": 4.5472555083107845e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1875
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 85.65625,
      "epoch": 0.09064553536915346,
      "grad_norm": 11.151697441491127,
      "learning_rate": 4.547013915732508e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.39879563450813293,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1876
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.75,
      "epoch": 0.09069385388480866,
      "grad_norm": 0.0,
      "learning_rate": 4.546772323154233e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 1877
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.71875,
      "epoch": 0.09074217240046385,
      "grad_norm": 0.0,
      "learning_rate": 4.5465307305759563e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1878
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 92.3125,
      "epoch": 0.09079049091611906,
      "grad_norm": 45.03083392112769,
      "learning_rate": 4.5462891379976805e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.3708576560020447,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1879
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.65625,
      "epoch": 0.09083880943177426,
      "grad_norm": 21.077363905486948,
      "learning_rate": 4.5460475454194046e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1880
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.59375,
      "epoch": 0.09088712794742945,
      "grad_norm": 16.316185818595304,
      "learning_rate": 4.545805952841129e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1881
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.71875,
      "epoch": 0.09093544646308466,
      "grad_norm": 0.0,
      "learning_rate": 4.5455643602628524e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1882
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.3125,
      "epoch": 0.09098376497873985,
      "grad_norm": 11.220301861755367,
      "learning_rate": 4.545322767684577e-07,
      "loss": -0.0,
      "reward": 2.21875,
      "reward_std": 0.2529362142086029,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1883
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.4375,
      "epoch": 0.09103208349439505,
      "grad_norm": 0.0,
      "learning_rate": 4.5450811751063006e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1884
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.75,
      "epoch": 0.09108040201005026,
      "grad_norm": 0.0,
      "learning_rate": 4.544839582528024e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1885
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.75,
      "epoch": 0.09112872052570545,
      "grad_norm": 0.0,
      "learning_rate": 4.5445979899497484e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1886
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 78.78125,
      "epoch": 0.09117703904136065,
      "grad_norm": 23.443516417219634,
      "learning_rate": 4.5443563973714725e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.23243674635887146,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1887
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 70.1875,
      "epoch": 0.09122535755701584,
      "grad_norm": 8.204023274237446,
      "learning_rate": 4.5441148047931967e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.2993341088294983,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1888
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 80.34375,
      "epoch": 0.09127367607267105,
      "grad_norm": 39.326689840430625,
      "learning_rate": 4.54387321221492e-07,
      "loss": -0.0,
      "reward": 2.265625,
      "reward_std": 0.45561128854751587,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1889
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.15625,
      "epoch": 0.09132199458832625,
      "grad_norm": 0.0,
      "learning_rate": 4.543631619636645e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 1890
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 83.28125,
      "epoch": 0.09137031310398144,
      "grad_norm": 15.329375236170575,
      "learning_rate": 4.5433900270583685e-07,
      "loss": 0.0,
      "reward": 1.5,
      "reward_std": 0.28104937076568604,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1891
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 78.75,
      "epoch": 0.09141863161963665,
      "grad_norm": 20.715850519188255,
      "learning_rate": 4.5431484344800927e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.5090347528457642,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1892
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.375,
      "epoch": 0.09146695013529184,
      "grad_norm": 14.074918660060481,
      "learning_rate": 4.542906841901817e-07,
      "loss": -0.0,
      "reward": 1.75,
      "reward_std": 0.27102428674697876,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1893
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.625,
      "epoch": 0.09151526865094704,
      "grad_norm": 8.342515810877542,
      "learning_rate": 4.5426652493235404e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2920827269554138,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1894
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.09375,
      "epoch": 0.09156358716660225,
      "grad_norm": 19.733171525628013,
      "learning_rate": 4.5424236567452645e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.317629873752594,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1895
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 79.09375,
      "epoch": 0.09161190568225744,
      "grad_norm": 34.32634617153069,
      "learning_rate": 4.542182064166988e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.4555858373641968,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1896
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 84.875,
      "epoch": 0.09166022419791264,
      "grad_norm": 13.745553204763308,
      "learning_rate": 4.541940471588713e-07,
      "loss": -0.0,
      "reward": 1.90625,
      "reward_std": 0.3662523031234741,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1897
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.84375,
      "epoch": 0.09170854271356783,
      "grad_norm": 0.0,
      "learning_rate": 4.5416988790104364e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1898
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.09375,
      "epoch": 0.09175686122922304,
      "grad_norm": 0.0,
      "learning_rate": 4.5414572864321606e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 1899
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.65625,
      "epoch": 0.09180517974487824,
      "grad_norm": 8.719981597302516,
      "learning_rate": 4.5412156938538847e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1900
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 94.9375,
      "epoch": 0.09185349826053343,
      "grad_norm": 5.116851682176304,
      "learning_rate": 4.540974101275609e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 1901
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 81.5625,
      "epoch": 0.09190181677618864,
      "grad_norm": 19.993148590695537,
      "learning_rate": 4.5407325086973324e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.29693374037742615,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1902
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.4375,
      "epoch": 0.09195013529184383,
      "grad_norm": 0.0,
      "learning_rate": 4.540490916119057e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1903
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.96875,
      "epoch": 0.09199845380749903,
      "grad_norm": 0.0,
      "learning_rate": 4.5402493235407807e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 1904
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 86.6875,
      "epoch": 0.09204677232315424,
      "grad_norm": 7.421866001831724,
      "learning_rate": 4.5400077309625043e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.19999711215496063,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1905
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.1875,
      "epoch": 0.09209509083880943,
      "grad_norm": 17.118524227183233,
      "learning_rate": 4.5397661383842285e-07,
      "loss": -0.0,
      "reward": 2.25,
      "reward_std": 0.29030048847198486,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1906
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.71875,
      "epoch": 0.09214340935446463,
      "grad_norm": 14.151528640494687,
      "learning_rate": 4.5395245458059526e-07,
      "loss": -0.0,
      "reward": 2.375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1907
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.625,
      "epoch": 0.09219172787011982,
      "grad_norm": 5.064717088825263,
      "learning_rate": 4.5392829532276767e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1908
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.53125,
      "epoch": 0.09224004638577503,
      "grad_norm": 7.021057882614672,
      "learning_rate": 4.5390413606494003e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.21528217196464539,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1909
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 86.96875,
      "epoch": 0.09228836490143023,
      "grad_norm": 13.832633135054499,
      "learning_rate": 4.538799768071125e-07,
      "loss": -0.0,
      "reward": 1.484375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1910
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.71875,
      "epoch": 0.09233668341708542,
      "grad_norm": 9.747030822149004,
      "learning_rate": 4.5385581754928486e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1911
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 93.03125,
      "epoch": 0.09238500193274063,
      "grad_norm": 9.179910428684181,
      "learning_rate": 4.538316582914573e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.19999711215496063,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1912
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 95.21875,
      "epoch": 0.09243332044839582,
      "grad_norm": 14.484695791595648,
      "learning_rate": 4.538074990336297e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1913
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 76.59375,
      "epoch": 0.09248163896405102,
      "grad_norm": 9.880040675573344,
      "learning_rate": 4.537833397758021e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.3051002025604248,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1914
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.75,
      "epoch": 0.09252995747970623,
      "grad_norm": 43.721963434820104,
      "learning_rate": 4.5375918051797446e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.3197515308856964,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1915
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 86.59375,
      "epoch": 0.09257827599536142,
      "grad_norm": 16.622363584787355,
      "learning_rate": 4.537350212601468e-07,
      "loss": -0.0,
      "reward": 2.34375,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1916
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 84.625,
      "epoch": 0.09262659451101662,
      "grad_norm": 94.31162118512583,
      "learning_rate": 4.537108620023193e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.4037913978099823,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1917
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.71875,
      "epoch": 0.09267491302667182,
      "grad_norm": 23.179275706988715,
      "learning_rate": 4.5368670274449165e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1918
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.875,
      "epoch": 0.09272323154232702,
      "grad_norm": 33.92657147825696,
      "learning_rate": 4.5366254348666406e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.27467289566993713,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1919
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 89.125,
      "epoch": 0.09277155005798222,
      "grad_norm": 23.5158650366441,
      "learning_rate": 4.536383842288365e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.3104073107242584,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1920
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.375,
      "epoch": 0.09281986857363742,
      "grad_norm": 8.078567421576462,
      "learning_rate": 4.536142249710089e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.308285653591156,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1921
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 85.28125,
      "epoch": 0.09286818708929262,
      "grad_norm": 9.179748700464053,
      "learning_rate": 4.5359006571318125e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 1922
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.75,
      "epoch": 0.09291650560494781,
      "grad_norm": 101.67148278136489,
      "learning_rate": 4.535659064553537e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.3899371922016144,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1923
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.75,
      "epoch": 0.09296482412060302,
      "grad_norm": 0.0,
      "learning_rate": 4.535417471975261e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1924
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.21875,
      "epoch": 0.09301314263625822,
      "grad_norm": 0.0,
      "learning_rate": 4.5351758793969844e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1925
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.96875,
      "epoch": 0.09306146115191341,
      "grad_norm": 0.0,
      "learning_rate": 4.5349342868187085e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1926
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.625,
      "epoch": 0.09310977966756862,
      "grad_norm": 0.0,
      "learning_rate": 4.5346926942404327e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1927
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.59375,
      "epoch": 0.0931580981832238,
      "grad_norm": 0.0,
      "learning_rate": 4.534451101662157e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 1928
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.59375,
      "epoch": 0.09320641669887901,
      "grad_norm": 6.371100680425705,
      "learning_rate": 4.5342095090838804e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1929
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.875,
      "epoch": 0.09325473521453422,
      "grad_norm": 0.0,
      "learning_rate": 4.533967916505605e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1930
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 79.4375,
      "epoch": 0.0933030537301894,
      "grad_norm": 15.570968449538062,
      "learning_rate": 4.5337263239273287e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.36213356256484985,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1931
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.03125,
      "epoch": 0.09335137224584461,
      "grad_norm": 0.0,
      "learning_rate": 4.533484731349053e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 1932
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.21875,
      "epoch": 0.0933996907614998,
      "grad_norm": 18.117081887428764,
      "learning_rate": 4.533243138770777e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.20633062720298767,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1933
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 68.4375,
      "epoch": 0.093448009277155,
      "grad_norm": 5.355720504802393,
      "learning_rate": 4.533001546192501e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.31795984506607056,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1934
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.40625,
      "epoch": 0.09349632779281021,
      "grad_norm": 10.62880073897276,
      "learning_rate": 4.5327599536142247e-07,
      "loss": -0.0,
      "reward": 2.296875,
      "reward_std": 0.188242569565773,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 1935
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.09375,
      "epoch": 0.0935446463084654,
      "grad_norm": 0.0,
      "learning_rate": 4.5325183610359483e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1936
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 93.28125,
      "epoch": 0.0935929648241206,
      "grad_norm": 8.969197099975576,
      "learning_rate": 4.532276768457673e-07,
      "loss": -0.0,
      "reward": 1.625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1937
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.625,
      "epoch": 0.0936412833397758,
      "grad_norm": 5.756532373364212,
      "learning_rate": 4.5320351758793966e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 1938
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.0,
      "epoch": 0.093689601855431,
      "grad_norm": 0.0,
      "learning_rate": 4.5317935833011207e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1939
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.59375,
      "epoch": 0.0937379203710862,
      "grad_norm": 4.302551715753777,
      "learning_rate": 4.531551990722845e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.308285653591156,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1940
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.28125,
      "epoch": 0.0937862388867414,
      "grad_norm": 0.0,
      "learning_rate": 4.531310398144569e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1941
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 95.3125,
      "epoch": 0.0938345574023966,
      "grad_norm": 12.192858901281063,
      "learning_rate": 4.5310688055662926e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.3006555736064911,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1942
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.875,
      "epoch": 0.09388287591805179,
      "grad_norm": 35.01606673582863,
      "learning_rate": 4.5308272129880173e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.33150607347488403,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1943
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.21875,
      "epoch": 0.093931194433707,
      "grad_norm": 0.0,
      "learning_rate": 4.530585620409741e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1944
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 83.5,
      "epoch": 0.0939795129493622,
      "grad_norm": 15.176616306323357,
      "learning_rate": 4.5303440278314645e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.33725470304489136,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1945
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.53125,
      "epoch": 0.09402783146501739,
      "grad_norm": 0.0,
      "learning_rate": 4.5301024352531886e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 1946
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.15625,
      "epoch": 0.0940761499806726,
      "grad_norm": 10.636518587732041,
      "learning_rate": 4.529860842674913e-07,
      "loss": -0.0,
      "reward": 2.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1947
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.34375,
      "epoch": 0.09412446849632779,
      "grad_norm": 0.0,
      "learning_rate": 4.529619250096637e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1948
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 100.65625,
      "epoch": 0.09417278701198299,
      "grad_norm": 20.964295145374635,
      "learning_rate": 4.5293776575183605e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1949
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.3125,
      "epoch": 0.0942211055276382,
      "grad_norm": 6.681484809664242,
      "learning_rate": 4.529136064940085e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1950
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.25,
      "epoch": 0.09426942404329339,
      "grad_norm": 0.0,
      "learning_rate": 4.528894472361809e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1951
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 93.09375,
      "epoch": 0.09431774255894859,
      "grad_norm": 17.107273543200286,
      "learning_rate": 4.528652879783533e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.35091447830200195,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1952
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.625,
      "epoch": 0.09436606107460378,
      "grad_norm": 5.337225639384561,
      "learning_rate": 4.528411287205257e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1953
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.84375,
      "epoch": 0.09441437959025899,
      "grad_norm": 0.0,
      "learning_rate": 4.528169694626981e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1954
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 87.625,
      "epoch": 0.09446269810591419,
      "grad_norm": 6.9719993434351855,
      "learning_rate": 4.527928102048705e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1955
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.21875,
      "epoch": 0.09451101662156938,
      "grad_norm": 10.841002744281786,
      "learning_rate": 4.5276865094704284e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.3029785752296448,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1956
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.03125,
      "epoch": 0.09455933513722459,
      "grad_norm": 0.0,
      "learning_rate": 4.527444916892153e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 1957
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.375,
      "epoch": 0.09460765365287978,
      "grad_norm": 0.0,
      "learning_rate": 4.5272033243138767e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1958
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.21875,
      "epoch": 0.09465597216853498,
      "grad_norm": 29.975523286086737,
      "learning_rate": 4.526961731735601e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.32346823811531067,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1959
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 89.9375,
      "epoch": 0.09470429068419019,
      "grad_norm": 20.90775726456193,
      "learning_rate": 4.526720139157325e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.3977220952510834,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1960
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 117.625,
      "epoch": 0.09475260919984538,
      "grad_norm": 8.473778857019466,
      "learning_rate": 4.526478546579049e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.3252435624599457,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1961
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.75,
      "epoch": 0.09480092771550058,
      "grad_norm": 0.0,
      "learning_rate": 4.5262369540007727e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 1962
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 106.0,
      "epoch": 0.09484924623115577,
      "grad_norm": 10.042854150084665,
      "learning_rate": 4.5259953614224973e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1963
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.34375,
      "epoch": 0.09489756474681098,
      "grad_norm": 0.0,
      "learning_rate": 4.525753768844221e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1964
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 103.3125,
      "epoch": 0.09494588326246618,
      "grad_norm": 12.881254133144466,
      "learning_rate": 4.525512176265945e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.3676624298095703,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1965
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 106.90625,
      "epoch": 0.09499420177812137,
      "grad_norm": 13.994275649099844,
      "learning_rate": 4.5252705836876687e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.32187318801879883,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1966
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.03125,
      "epoch": 0.09504252029377658,
      "grad_norm": 0.0,
      "learning_rate": 4.525028991109393e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1967
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 100.71875,
      "epoch": 0.09509083880943177,
      "grad_norm": 25.301754986918386,
      "learning_rate": 4.524787398531117e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.558118462562561,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1968
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.09375,
      "epoch": 0.09513915732508697,
      "grad_norm": 0.0,
      "learning_rate": 4.5245458059528406e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 1969
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 123.3125,
      "epoch": 0.09518747584074218,
      "grad_norm": 15.4631342311353,
      "learning_rate": 4.524304213374565e-07,
      "loss": -0.0,
      "reward": 1.578125,
      "reward_std": 0.494406521320343,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 1970
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 100.40625,
      "epoch": 0.09523579435639737,
      "grad_norm": 12.297852252821754,
      "learning_rate": 4.524062620796289e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.26726123690605164,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1971
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.59375,
      "epoch": 0.09528411287205257,
      "grad_norm": 0.0,
      "learning_rate": 4.523821028218013e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1972
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 105.3125,
      "epoch": 0.09533243138770776,
      "grad_norm": 44.060935231254064,
      "learning_rate": 4.523579435639737e-07,
      "loss": -0.0,
      "reward": 1.625,
      "reward_std": 0.4491388201713562,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1973
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 104.59375,
      "epoch": 0.09538074990336297,
      "grad_norm": 10.097242687241637,
      "learning_rate": 4.523337843061461e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.37190574407577515,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1974
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.46875,
      "epoch": 0.09542906841901817,
      "grad_norm": 0.0,
      "learning_rate": 4.523096250483185e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 1975
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 101.5625,
      "epoch": 0.09547738693467336,
      "grad_norm": 8.006989678307564,
      "learning_rate": 4.5228546579049085e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.2982703447341919,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 1976
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 97.5625,
      "epoch": 0.09552570545032857,
      "grad_norm": 13.191071449972014,
      "learning_rate": 4.522613065326633e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1977
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.84375,
      "epoch": 0.09557402396598376,
      "grad_norm": 0.0,
      "learning_rate": 4.522371472748357e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1978
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.5,
      "epoch": 0.09562234248163896,
      "grad_norm": 0.0,
      "learning_rate": 4.522129880170081e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1979
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 99.875,
      "epoch": 0.09567066099729417,
      "grad_norm": 11.683488743232468,
      "learning_rate": 4.521888287591805e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.4150615930557251,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1980
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 103.9375,
      "epoch": 0.09571897951294936,
      "grad_norm": 5.1490494411525,
      "learning_rate": 4.521646695013529e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1981
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.5,
      "epoch": 0.09576729802860456,
      "grad_norm": 0.0,
      "learning_rate": 4.521405102435253e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 1982
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 111.65625,
      "epoch": 0.09581561654425975,
      "grad_norm": 18.659344395127953,
      "learning_rate": 4.5211635098569774e-07,
      "loss": 0.0,
      "reward": 1.390625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 1983
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 117.5,
      "epoch": 0.09586393505991496,
      "grad_norm": 9.55722981461499,
      "learning_rate": 4.520921917278701e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.23827512562274933,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1984
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 115.3125,
      "epoch": 0.09591225357557016,
      "grad_norm": 6.669649129546688,
      "learning_rate": 4.520680324700425e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 1985
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 91.625,
      "epoch": 0.09596057209122535,
      "grad_norm": 13.707808984399119,
      "learning_rate": 4.520438732122149e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.24831002950668335,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1986
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.3125,
      "epoch": 0.09600889060688056,
      "grad_norm": 0.0,
      "learning_rate": 4.520197139543873e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 1987
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 108.46875,
      "epoch": 0.09605720912253575,
      "grad_norm": 4.870248103150213,
      "learning_rate": 4.519955546965597e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1988
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.125,
      "epoch": 0.09610552763819095,
      "grad_norm": 0.0,
      "learning_rate": 4.5197139543873207e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 1989
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.59375,
      "epoch": 0.09615384615384616,
      "grad_norm": 0.0,
      "learning_rate": 4.5194723618090453e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 1990
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.875,
      "epoch": 0.09620216466950135,
      "grad_norm": 0.0,
      "learning_rate": 4.519230769230769e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 1991
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.25,
      "epoch": 0.09625048318515655,
      "grad_norm": 0.0,
      "learning_rate": 4.518989176652493e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 1992
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 95.0625,
      "epoch": 0.09629880170081175,
      "grad_norm": 6.057861106756051,
      "learning_rate": 4.518747584074217e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.3608423173427582,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 1993
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 110.875,
      "epoch": 0.09634712021646695,
      "grad_norm": 5.312712060732674,
      "learning_rate": 4.5185059914959413e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1994
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 103.03125,
      "epoch": 0.09639543873212215,
      "grad_norm": 6.251470577012096,
      "learning_rate": 4.518264398917665e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 1995
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 107.5,
      "epoch": 0.09644375724777735,
      "grad_norm": 5.59795658895245,
      "learning_rate": 4.518022806339389e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 1996
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 107.625,
      "epoch": 0.09649207576343255,
      "grad_norm": 5.7821118000549365,
      "learning_rate": 4.517781213761113e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 1997
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.84375,
      "epoch": 0.09654039427908774,
      "grad_norm": 0.0,
      "learning_rate": 4.517539621182837e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 1998
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 123.75,
      "epoch": 0.09658871279474295,
      "grad_norm": 0.0,
      "learning_rate": 4.517298028604561e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 1999
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.0625,
      "epoch": 0.09663703131039815,
      "grad_norm": 0.0,
      "learning_rate": 4.517056436026285e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2000
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 109.28125,
      "epoch": 0.09668534982605334,
      "grad_norm": 0.0,
      "learning_rate": 4.516814843448009e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2001
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 106.09375,
      "epoch": 0.09673366834170855,
      "grad_norm": 4.653542522968168,
      "learning_rate": 4.516573250869733e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2002
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 116.03125,
      "epoch": 0.09678198685736374,
      "grad_norm": 13.815832834512769,
      "learning_rate": 4.5163316582914575e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2003
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.40625,
      "epoch": 0.09683030537301894,
      "grad_norm": 0.0,
      "learning_rate": 4.516090065713181e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 2004
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 115.375,
      "epoch": 0.09687862388867415,
      "grad_norm": 11.550412871749417,
      "learning_rate": 4.515848473134905e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2005
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.46875,
      "epoch": 0.09692694240432934,
      "grad_norm": 0.0,
      "learning_rate": 4.515606880556629e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2006
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 106.75,
      "epoch": 0.09697526091998454,
      "grad_norm": 7.057741939336327,
      "learning_rate": 4.5153652879783535e-07,
      "loss": -0.0,
      "reward": 2.140625,
      "reward_std": 0.2961486577987671,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2007
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.75,
      "epoch": 0.09702357943563973,
      "grad_norm": 0.0,
      "learning_rate": 4.515123695400077e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2008
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.625,
      "epoch": 0.09707189795129494,
      "grad_norm": 0.0,
      "learning_rate": 4.5148821028218007e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2009
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 115.28125,
      "epoch": 0.09712021646695014,
      "grad_norm": 16.708504148440728,
      "learning_rate": 4.5146405102435254e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2010
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 104.21875,
      "epoch": 0.09716853498260533,
      "grad_norm": 12.609657335282169,
      "learning_rate": 4.514398917665249e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2011
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 92.34375,
      "epoch": 0.09721685349826054,
      "grad_norm": 8.200333316195767,
      "learning_rate": 4.514157325086973e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.3230033218860626,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2012
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.8125,
      "epoch": 0.09726517201391573,
      "grad_norm": 0.0,
      "learning_rate": 4.5139157325086973e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2013
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 116.9375,
      "epoch": 0.09731349052957093,
      "grad_norm": 43.74068083103278,
      "learning_rate": 4.5136741399304214e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.30509042739868164,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2014
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.78125,
      "epoch": 0.09736180904522614,
      "grad_norm": 0.0,
      "learning_rate": 4.513432547352145e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2015
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 119.75,
      "epoch": 0.09741012756088133,
      "grad_norm": 18.81752899245148,
      "learning_rate": 4.513190954773869e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.23827512562274933,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2016
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.59375,
      "epoch": 0.09745844607653653,
      "grad_norm": 0.0,
      "learning_rate": 4.5129493621955933e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2017
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.6875,
      "epoch": 0.09750676459219172,
      "grad_norm": 0.0,
      "learning_rate": 4.512707769617317e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2018
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 108.4375,
      "epoch": 0.09755508310784693,
      "grad_norm": 6.012948886622389,
      "learning_rate": 4.512466177039041e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2019
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 106.6875,
      "epoch": 0.09760340162350213,
      "grad_norm": 19.571669200468694,
      "learning_rate": 4.512224584460765e-07,
      "loss": -0.0,
      "reward": 2.28125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2020
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 110.375,
      "epoch": 0.09765172013915732,
      "grad_norm": 13.836508367533739,
      "learning_rate": 4.5119829918824893e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.37405282258987427,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2021
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 113.59375,
      "epoch": 0.09770003865481253,
      "grad_norm": 0.0,
      "learning_rate": 4.511741399304213e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2022
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.4375,
      "epoch": 0.09774835717046772,
      "grad_norm": 0.0,
      "learning_rate": 4.5114998067259376e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2023
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 125.65625,
      "epoch": 0.09779667568612292,
      "grad_norm": 6.562198942864914,
      "learning_rate": 4.511258214147661e-07,
      "loss": -0.0,
      "reward": 1.453125,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 2024
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.5625,
      "epoch": 0.09784499420177813,
      "grad_norm": 0.0,
      "learning_rate": 4.5110166215693853e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2025
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.25,
      "epoch": 0.09789331271743332,
      "grad_norm": 42.45439148273885,
      "learning_rate": 4.510775028991109e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.29305070638656616,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2026
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 128.5625,
      "epoch": 0.09794163123308852,
      "grad_norm": 17.057389827832235,
      "learning_rate": 4.5105334364128336e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.30935919284820557,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2027
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 122.1875,
      "epoch": 0.09798994974874371,
      "grad_norm": 14.852871540368774,
      "learning_rate": 4.510291843834557e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.28104937076568604,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2028
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 98.875,
      "epoch": 0.09803826826439892,
      "grad_norm": 9.211254195594304,
      "learning_rate": 4.510050251256281e-07,
      "loss": 0.0,
      "reward": 1.453125,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 2029
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 119.75,
      "epoch": 0.09808658678005412,
      "grad_norm": 16.10546607268595,
      "learning_rate": 4.5098086586780055e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2030
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 104.8125,
      "epoch": 0.09813490529570931,
      "grad_norm": 9.162759483354264,
      "learning_rate": 4.509567066099729e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.43874651193618774,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2031
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 120.40625,
      "epoch": 0.09818322381136452,
      "grad_norm": 19.752941583094916,
      "learning_rate": 4.509325473521453e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.22738909721374512,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2032
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 111.40625,
      "epoch": 0.09823154232701971,
      "grad_norm": 11.872104424674536,
      "learning_rate": 4.5090838809431774e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22738909721374512,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2033
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 109.21875,
      "epoch": 0.09827986084267491,
      "grad_norm": 8.916515257895217,
      "learning_rate": 4.5088422883649015e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2034
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 110.03125,
      "epoch": 0.09832817935833012,
      "grad_norm": 18.89783146714711,
      "learning_rate": 4.508600695786625e-07,
      "loss": -0.0,
      "reward": 1.90625,
      "reward_std": 0.39265817403793335,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2035
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 110.3125,
      "epoch": 0.09837649787398531,
      "grad_norm": 11.22099442867859,
      "learning_rate": 4.508359103208349e-07,
      "loss": -0.0,
      "reward": 2.078125,
      "reward_std": 0.42190128564834595,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2036
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 128.28125,
      "epoch": 0.09842481638964051,
      "grad_norm": 34.93024569260681,
      "learning_rate": 4.5081175106300734e-07,
      "loss": -0.0,
      "reward": 1.515625,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2037
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 102.78125,
      "epoch": 0.0984731349052957,
      "grad_norm": 10.57138057814358,
      "learning_rate": 4.507875918051797e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.28247910737991333,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2038
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.5,
      "epoch": 0.09852145342095091,
      "grad_norm": 0.0,
      "learning_rate": 4.507634325473521e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2039
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 109.875,
      "epoch": 0.09856977193660611,
      "grad_norm": 93.13107996381433,
      "learning_rate": 4.507392732895245e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.27883464097976685,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2040
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 108.15625,
      "epoch": 0.0986180904522613,
      "grad_norm": 9.120896162421596,
      "learning_rate": 4.5071511403169694e-07,
      "loss": -0.0,
      "reward": 2.21875,
      "reward_std": 0.38816186785697937,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2041
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.59375,
      "epoch": 0.09866640896791651,
      "grad_norm": 0.0,
      "learning_rate": 4.506909547738693e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2042
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 101.25,
      "epoch": 0.0987147274835717,
      "grad_norm": 9.791233720412189,
      "learning_rate": 4.5066679551604177e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2043
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 117.59375,
      "epoch": 0.0987630459992269,
      "grad_norm": 11.123087559988372,
      "learning_rate": 4.5064263625821413e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.37108510732650757,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2044
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 107.25,
      "epoch": 0.09881136451488211,
      "grad_norm": 6.60336835636231,
      "learning_rate": 4.5061847700038654e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.30173346400260925,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2045
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 115.28125,
      "epoch": 0.0988596830305373,
      "grad_norm": 19.43379820366063,
      "learning_rate": 4.505943177425589e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.2302001267671585,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2046
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.78125,
      "epoch": 0.0989080015461925,
      "grad_norm": 0.0,
      "learning_rate": 4.5057015848473137e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2047
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 108.90625,
      "epoch": 0.0989563200618477,
      "grad_norm": 13.178138332628382,
      "learning_rate": 4.5054599922690373e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2048
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.75,
      "epoch": 0.0990046385775029,
      "grad_norm": 0.0,
      "learning_rate": 4.505218399690761e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2049
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 95.3125,
      "epoch": 0.0990529570931581,
      "grad_norm": 7.612696477806553,
      "learning_rate": 4.5049768071124856e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.3061639666557312,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2050
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 96.4375,
      "epoch": 0.0991012756088133,
      "grad_norm": 7.706089126486334,
      "learning_rate": 4.504735214534209e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.3330284357070923,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2051
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.90625,
      "epoch": 0.0991495941244685,
      "grad_norm": 0.0,
      "learning_rate": 4.5044936219559333e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2052
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.5625,
      "epoch": 0.09919791264012369,
      "grad_norm": 0.0,
      "learning_rate": 4.5042520293776574e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2053
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 99.3125,
      "epoch": 0.0992462311557789,
      "grad_norm": 22.8748225810299,
      "learning_rate": 4.5040104367993816e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.37722262740135193,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2054
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.34375,
      "epoch": 0.0992945496714341,
      "grad_norm": 0.0,
      "learning_rate": 4.503768844221105e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2055
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 110.9375,
      "epoch": 0.09934286818708929,
      "grad_norm": 14.061406576342746,
      "learning_rate": 4.5035272516428293e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.5412536859512329,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2056
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.03125,
      "epoch": 0.0993911867027445,
      "grad_norm": 0.0,
      "learning_rate": 4.5032856590645534e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 2057
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.90625,
      "epoch": 0.09943950521839968,
      "grad_norm": 0.0,
      "learning_rate": 4.5030440664862776e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2058
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.59375,
      "epoch": 0.09948782373405489,
      "grad_norm": 0.0,
      "learning_rate": 4.502802473908001e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2059
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.65625,
      "epoch": 0.0995361422497101,
      "grad_norm": 0.0,
      "learning_rate": 4.5025608813297253e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2060
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 115.03125,
      "epoch": 0.09958446076536528,
      "grad_norm": 0.0,
      "learning_rate": 4.5023192887514495e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 2061
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 105.90625,
      "epoch": 0.09963277928102049,
      "grad_norm": 21.90345890556263,
      "learning_rate": 4.502077696173173e-07,
      "loss": 0.0,
      "reward": 1.21875,
      "reward_std": 0.18600594997406006,
      "rewards/accuracy_reward": 0.15625,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 2062
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 119.71875,
      "epoch": 0.09968109779667568,
      "grad_norm": 10.276444611808696,
      "learning_rate": 4.501836103594898e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2063
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.40625,
      "epoch": 0.09972941631233088,
      "grad_norm": 0.0,
      "learning_rate": 4.5015945110166213e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2064
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 111.3125,
      "epoch": 0.09977773482798609,
      "grad_norm": 0.0,
      "learning_rate": 4.5013529184383455e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2065
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.40625,
      "epoch": 0.09982605334364128,
      "grad_norm": 0.0,
      "learning_rate": 4.501111325860069e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2066
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 109.875,
      "epoch": 0.09987437185929648,
      "grad_norm": 7.4462032965073615,
      "learning_rate": 4.500869733281794e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2951723635196686,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2067
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 129.875,
      "epoch": 0.09992269037495168,
      "grad_norm": 8.2133187297616,
      "learning_rate": 4.5006281407035174e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.17598573863506317,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2068
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.25,
      "epoch": 0.09997100889060688,
      "grad_norm": 0.0,
      "learning_rate": 4.500386548125241e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2069
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 122.09375,
      "epoch": 0.10001932740626208,
      "grad_norm": 0.0,
      "learning_rate": 4.5001449555469656e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 2070
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 107.125,
      "epoch": 0.10006764592191728,
      "grad_norm": 7.089739517591716,
      "learning_rate": 4.499903362968689e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.31539323925971985,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2071
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.09375,
      "epoch": 0.10011596443757248,
      "grad_norm": 0.0,
      "learning_rate": 4.4996617703904134e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2072
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 114.9375,
      "epoch": 0.10016428295322767,
      "grad_norm": 10.886338038625025,
      "learning_rate": 4.4994201778121375e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.23384949564933777,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2073
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 132.8125,
      "epoch": 0.10021260146888288,
      "grad_norm": 5.379890903503586,
      "learning_rate": 4.4991785852338616e-07,
      "loss": -0.0,
      "reward": 1.3125,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2074
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.25,
      "epoch": 0.10026091998453808,
      "grad_norm": 0.0,
      "learning_rate": 4.498936992655585e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2075
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 125.25,
      "epoch": 0.10030923850019327,
      "grad_norm": 10.33719085636481,
      "learning_rate": 4.4986954000773094e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.4282999336719513,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2076
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 130.6875,
      "epoch": 0.10035755701584848,
      "grad_norm": 0.0,
      "learning_rate": 4.4984538074990335e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2077
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.96875,
      "epoch": 0.10040587553150367,
      "grad_norm": 0.0,
      "learning_rate": 4.4982122149207577e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2078
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 130.21875,
      "epoch": 0.10045419404715887,
      "grad_norm": 17.265551779025277,
      "learning_rate": 4.4979706223424813e-07,
      "loss": -0.0,
      "reward": 2.078125,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2079
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 109.71875,
      "epoch": 0.10050251256281408,
      "grad_norm": 12.748997039330726,
      "learning_rate": 4.4977290297642054e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2080
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 114.5,
      "epoch": 0.10055083107846927,
      "grad_norm": 0.0,
      "learning_rate": 4.4974874371859295e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2081
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.375,
      "epoch": 0.10059914959412447,
      "grad_norm": 0.0,
      "learning_rate": 4.497245844607653e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2082
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 121.53125,
      "epoch": 0.10064746810977966,
      "grad_norm": 7.707900587173797,
      "learning_rate": 4.497004252029378e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2083
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 132.3125,
      "epoch": 0.10069578662543487,
      "grad_norm": 0.0,
      "learning_rate": 4.4967626594511014e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2084
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 124.90625,
      "epoch": 0.10074410514109007,
      "grad_norm": 6.97913540455979,
      "learning_rate": 4.4965210668728256e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.35168886184692383,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2085
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 114.03125,
      "epoch": 0.10079242365674526,
      "grad_norm": 9.29694122710033,
      "learning_rate": 4.496279474294549e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2086
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.71875,
      "epoch": 0.10084074217240047,
      "grad_norm": 0.0,
      "learning_rate": 4.496037881716274e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2087
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.21875,
      "epoch": 0.10088906068805566,
      "grad_norm": 0.0,
      "learning_rate": 4.4957962891379974e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2088
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 123.53125,
      "epoch": 0.10093737920371086,
      "grad_norm": 20.25609135428287,
      "learning_rate": 4.495554696559721e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.2529362142086029,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2089
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 127.15625,
      "epoch": 0.10098569771936607,
      "grad_norm": 18.37711290076137,
      "learning_rate": 4.4953131039814457e-07,
      "loss": -0.0,
      "reward": 1.21875,
      "reward_std": 0.18600594997406006,
      "rewards/accuracy_reward": 0.15625,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 2090
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.875,
      "epoch": 0.10103401623502126,
      "grad_norm": 0.0,
      "learning_rate": 4.4950715114031693e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2091
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.28125,
      "epoch": 0.10108233475067646,
      "grad_norm": 0.0,
      "learning_rate": 4.4948299188248935e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2092
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.90625,
      "epoch": 0.10113065326633165,
      "grad_norm": 0.0,
      "learning_rate": 4.4945883262466176e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2093
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 118.9375,
      "epoch": 0.10117897178198686,
      "grad_norm": 0.0,
      "learning_rate": 4.4943467336683417e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2094
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 118.46875,
      "epoch": 0.10122729029764206,
      "grad_norm": 7.43586762251969,
      "learning_rate": 4.4941051410900653e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2095
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 120.15625,
      "epoch": 0.10127560881329725,
      "grad_norm": 13.172685559696395,
      "learning_rate": 4.4938635485117895e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2096
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.5,
      "epoch": 0.10132392732895246,
      "grad_norm": 0.0,
      "learning_rate": 4.4936219559335136e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2097
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 114.65625,
      "epoch": 0.10137224584460765,
      "grad_norm": 4.786140563152984,
      "learning_rate": 4.493380363355238e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2098
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.65625,
      "epoch": 0.10142056436026285,
      "grad_norm": 0.0,
      "learning_rate": 4.4931387707769613e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2099
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 132.0625,
      "epoch": 0.10146888287591806,
      "grad_norm": 4.054701638767518,
      "learning_rate": 4.4928971781986855e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2100
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.84375,
      "epoch": 0.10151720139157325,
      "grad_norm": 0.0,
      "learning_rate": 4.4926555856204096e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2101
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 123.46875,
      "epoch": 0.10156551990722845,
      "grad_norm": 0.0,
      "learning_rate": 4.492413993042133e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2102
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 115.46875,
      "epoch": 0.10161383842288364,
      "grad_norm": 38.04516538949195,
      "learning_rate": 4.492172400463858e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.22835706174373627,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2103
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 129.125,
      "epoch": 0.10166215693853885,
      "grad_norm": 20.055901002552805,
      "learning_rate": 4.4919308078855815e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.3998437523841858,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2104
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 120.90625,
      "epoch": 0.10171047545419405,
      "grad_norm": 11.275569325778296,
      "learning_rate": 4.4916892153073056e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.27564918994903564,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2105
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 124.25,
      "epoch": 0.10175879396984924,
      "grad_norm": 0.0,
      "learning_rate": 4.491447622729029e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2106
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.15625,
      "epoch": 0.10180711248550445,
      "grad_norm": 0.0,
      "learning_rate": 4.491206030150754e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2107
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.28125,
      "epoch": 0.10185543100115964,
      "grad_norm": 0.0,
      "learning_rate": 4.4909644375724775e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2108
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 132.46875,
      "epoch": 0.10190374951681484,
      "grad_norm": 0.0,
      "learning_rate": 4.4907228449942017e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2109
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 129.65625,
      "epoch": 0.10195206803247005,
      "grad_norm": 0.0,
      "learning_rate": 4.490481252415926e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2110
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.0625,
      "epoch": 0.10200038654812524,
      "grad_norm": 0.0,
      "learning_rate": 4.4902396598376494e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2111
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.96875,
      "epoch": 0.10204870506378044,
      "grad_norm": 0.0,
      "learning_rate": 4.4899980672593735e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2112
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 126.21875,
      "epoch": 0.10209702357943563,
      "grad_norm": 0.0,
      "learning_rate": 4.4897564746810977e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2113
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 124.65625,
      "epoch": 0.10214534209509084,
      "grad_norm": 0.0,
      "learning_rate": 4.489514882102822e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2114
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 124.84375,
      "epoch": 0.10219366061074604,
      "grad_norm": 0.0,
      "learning_rate": 4.4892732895245454e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2115
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 136.375,
      "epoch": 0.10224197912640123,
      "grad_norm": 15.532907865701908,
      "learning_rate": 4.4890316969462695e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2116
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 136.65625,
      "epoch": 0.10229029764205644,
      "grad_norm": 0.0,
      "learning_rate": 4.4887901043679937e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2117
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.6875,
      "epoch": 0.10233861615771163,
      "grad_norm": 0.0,
      "learning_rate": 4.488548511789718e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2118
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 136.84375,
      "epoch": 0.10238693467336683,
      "grad_norm": 0.0,
      "learning_rate": 4.4883069192114414e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2119
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 122.8125,
      "epoch": 0.10243525318902204,
      "grad_norm": 0.0,
      "learning_rate": 4.488065326633166e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2120
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 125.71875,
      "epoch": 0.10248357170467723,
      "grad_norm": 0.0,
      "learning_rate": 4.4878237340548897e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2121
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 127.875,
      "epoch": 0.10253189022033243,
      "grad_norm": 0.0,
      "learning_rate": 4.4875821414766133e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2122
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 123.53125,
      "epoch": 0.10258020873598762,
      "grad_norm": 0.0,
      "learning_rate": 4.487340548898338e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2123
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 127.40625,
      "epoch": 0.10262852725164283,
      "grad_norm": 0.0,
      "learning_rate": 4.4870989563200616e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2124
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 123.8125,
      "epoch": 0.10267684576729803,
      "grad_norm": 0.0,
      "learning_rate": 4.4868573637417857e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2125
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.21875,
      "epoch": 0.10272516428295322,
      "grad_norm": 0.0,
      "learning_rate": 4.4866157711635093e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2126
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 148.9375,
      "epoch": 0.10277348279860843,
      "grad_norm": 0.0,
      "learning_rate": 4.486374178585234e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2127
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 129.875,
      "epoch": 0.10282180131426362,
      "grad_norm": 13.824943870968717,
      "learning_rate": 4.4861325860069576e-07,
      "loss": -0.0,
      "reward": 2.265625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2128
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 135.8125,
      "epoch": 0.10287011982991882,
      "grad_norm": 0.0,
      "learning_rate": 4.4858909934286817e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2129
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 133.21875,
      "epoch": 0.10291843834557403,
      "grad_norm": 9.41366515749206,
      "learning_rate": 4.485649400850406e-07,
      "loss": 0.0,
      "reward": 1.28125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2130
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 127.625,
      "epoch": 0.10296675686122922,
      "grad_norm": 0.0,
      "learning_rate": 4.4854078082721295e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2131
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 124.1875,
      "epoch": 0.10301507537688442,
      "grad_norm": 0.0,
      "learning_rate": 4.4851662156938536e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2132
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 132.5,
      "epoch": 0.10306339389253961,
      "grad_norm": 0.0,
      "learning_rate": 4.484924623115578e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2133
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 125.875,
      "epoch": 0.10311171240819482,
      "grad_norm": 10.194687987251415,
      "learning_rate": 4.484683030537302e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.4592459797859192,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2134
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.875,
      "epoch": 0.10316003092385002,
      "grad_norm": 0.0,
      "learning_rate": 4.4844414379590255e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2135
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 123.4375,
      "epoch": 0.10320834943950521,
      "grad_norm": 0.0,
      "learning_rate": 4.4841998453807496e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2136
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 126.40625,
      "epoch": 0.10325666795516042,
      "grad_norm": 7.456606825132452,
      "learning_rate": 4.483958252802474e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2137
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.21875,
      "epoch": 0.10330498647081561,
      "grad_norm": 0.0,
      "learning_rate": 4.483716660224198e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2138
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.21875,
      "epoch": 0.10335330498647081,
      "grad_norm": 0.0,
      "learning_rate": 4.4834750676459215e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2139
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.25,
      "epoch": 0.10340162350212602,
      "grad_norm": 0.0,
      "learning_rate": 4.483233475067646e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 2140
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.53125,
      "epoch": 0.10344994201778121,
      "grad_norm": 0.0,
      "learning_rate": 4.48299188248937e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2141
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 112.78125,
      "epoch": 0.10349826053343641,
      "grad_norm": 12.25446601702919,
      "learning_rate": 4.4827502899110934e-07,
      "loss": -0.0,
      "reward": 2.296875,
      "reward_std": 0.308285653591156,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2142
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 127.6875,
      "epoch": 0.1035465790490916,
      "grad_norm": 11.495676769437864,
      "learning_rate": 4.482508697332818e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2143
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 113.03125,
      "epoch": 0.10359489756474681,
      "grad_norm": 37.46156860670908,
      "learning_rate": 4.4822671047545417e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2144
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 126.4375,
      "epoch": 0.10364321608040201,
      "grad_norm": 9.566151322960103,
      "learning_rate": 4.482025512176266e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.21371951699256897,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2145
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.46875,
      "epoch": 0.1036915345960572,
      "grad_norm": 0.0,
      "learning_rate": 4.4817839195979894e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2146
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 130.3125,
      "epoch": 0.10373985311171241,
      "grad_norm": 0.0,
      "learning_rate": 4.481542327019714e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2147
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 135.90625,
      "epoch": 0.1037881716273676,
      "grad_norm": 0.0,
      "learning_rate": 4.4813007344414377e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2148
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.59375,
      "epoch": 0.1038364901430228,
      "grad_norm": 0.0,
      "learning_rate": 4.481059141863162e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2149
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 133.3125,
      "epoch": 0.10388480865867801,
      "grad_norm": 18.2522603222964,
      "learning_rate": 4.480817549284886e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.36278265714645386,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2150
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 122.0,
      "epoch": 0.1039331271743332,
      "grad_norm": 0.0,
      "learning_rate": 4.4805759567066096e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 2151
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 119.34375,
      "epoch": 0.1039814456899884,
      "grad_norm": 5.5514720856029625,
      "learning_rate": 4.4803343641283337e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2152
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 134.0,
      "epoch": 0.1040297642056436,
      "grad_norm": 0.0,
      "learning_rate": 4.480092771550058e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2153
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 114.90625,
      "epoch": 0.1040780827212988,
      "grad_norm": 4.571992182411719,
      "learning_rate": 4.479851178971782e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2154
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.46875,
      "epoch": 0.104126401236954,
      "grad_norm": 0.0,
      "learning_rate": 4.4796095863935056e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2155
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.34375,
      "epoch": 0.1041747197526092,
      "grad_norm": 0.0,
      "learning_rate": 4.4793679938152297e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 2156
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 109.0625,
      "epoch": 0.1042230382682644,
      "grad_norm": 0.0,
      "learning_rate": 4.479126401236954e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2157
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.65625,
      "epoch": 0.10427135678391959,
      "grad_norm": 0.0,
      "learning_rate": 4.478884808658678e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2158
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.71875,
      "epoch": 0.1043196752995748,
      "grad_norm": 0.0,
      "learning_rate": 4.4786432160804016e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2159
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.25,
      "epoch": 0.10436799381523,
      "grad_norm": 0.0,
      "learning_rate": 4.478401623502126e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2160
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 113.03125,
      "epoch": 0.10441631233088519,
      "grad_norm": 17.261818285147562,
      "learning_rate": 4.47816003092385e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.21528217196464539,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2161
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.21875,
      "epoch": 0.1044646308465404,
      "grad_norm": 0.0,
      "learning_rate": 4.4779184383455735e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2162
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 109.375,
      "epoch": 0.10451294936219559,
      "grad_norm": 0.0,
      "learning_rate": 4.477676845767298e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2163
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 131.4375,
      "epoch": 0.10456126787785079,
      "grad_norm": 10.0826073627544,
      "learning_rate": 4.477435253189022e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2164
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.71875,
      "epoch": 0.104609586393506,
      "grad_norm": 0.0,
      "learning_rate": 4.477193660610746e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2165
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 124.46875,
      "epoch": 0.10465790490916119,
      "grad_norm": 9.549246205239871,
      "learning_rate": 4.4769520680324695e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2166
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.125,
      "epoch": 0.10470622342481639,
      "grad_norm": 0.0,
      "learning_rate": 4.476710475454194e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2167
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.78125,
      "epoch": 0.10475454194047158,
      "grad_norm": 0.0,
      "learning_rate": 4.476468882875918e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2168
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.0625,
      "epoch": 0.10480286045612679,
      "grad_norm": 0.0,
      "learning_rate": 4.476227290297642e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2169
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 127.09375,
      "epoch": 0.10485117897178199,
      "grad_norm": 0.0,
      "learning_rate": 4.475985697719366e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2170
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 131.84375,
      "epoch": 0.10489949748743718,
      "grad_norm": 5.409245210422641,
      "learning_rate": 4.47574410514109e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.15580293536186218,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2171
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.25,
      "epoch": 0.10494781600309239,
      "grad_norm": 0.0,
      "learning_rate": 4.475502512562814e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2172
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.59375,
      "epoch": 0.10499613451874758,
      "grad_norm": 0.0,
      "learning_rate": 4.475260919984538e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2173
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 132.71875,
      "epoch": 0.10504445303440278,
      "grad_norm": 5.116953014699152,
      "learning_rate": 4.475019327406262e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2174
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.59375,
      "epoch": 0.10509277155005799,
      "grad_norm": 0.0,
      "learning_rate": 4.4747777348279856e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2175
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.125,
      "epoch": 0.10514109006571318,
      "grad_norm": 0.0,
      "learning_rate": 4.47453614224971e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2176
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 117.5625,
      "epoch": 0.10518940858136838,
      "grad_norm": 8.413974193495752,
      "learning_rate": 4.474294549671434e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2177
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 116.9375,
      "epoch": 0.10523772709702357,
      "grad_norm": 0.0,
      "learning_rate": 4.474052957093158e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2178
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.59375,
      "epoch": 0.10528604561267878,
      "grad_norm": 0.0,
      "learning_rate": 4.4738113645148817e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2179
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.40625,
      "epoch": 0.10533436412833398,
      "grad_norm": 0.0,
      "learning_rate": 4.4735697719366063e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2180
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 125.0625,
      "epoch": 0.10538268264398917,
      "grad_norm": 0.0,
      "learning_rate": 4.47332817935833e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2181
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 115.46875,
      "epoch": 0.10543100115964438,
      "grad_norm": 0.0,
      "learning_rate": 4.4730865867800535e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 2182
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.75,
      "epoch": 0.10547931967529957,
      "grad_norm": 0.0,
      "learning_rate": 4.472844994201778e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2183
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 121.5,
      "epoch": 0.10552763819095477,
      "grad_norm": 5.264137257369632,
      "learning_rate": 4.472603401623502e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2184
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 122.03125,
      "epoch": 0.10557595670660998,
      "grad_norm": 5.601170120294653,
      "learning_rate": 4.472361809045226e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2185
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.375,
      "epoch": 0.10562427522226517,
      "grad_norm": 0.0,
      "learning_rate": 4.4721202164669496e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2186
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 119.25,
      "epoch": 0.10567259373792037,
      "grad_norm": 0.0,
      "learning_rate": 4.471878623888674e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2187
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.0,
      "epoch": 0.10572091225357556,
      "grad_norm": 0.0,
      "learning_rate": 4.471637031310398e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2188
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.625,
      "epoch": 0.10576923076923077,
      "grad_norm": 0.0,
      "learning_rate": 4.471395438732122e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2189
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.625,
      "epoch": 0.10581754928488597,
      "grad_norm": 0.0,
      "learning_rate": 4.471153846153846e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2190
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 125.46875,
      "epoch": 0.10586586780054116,
      "grad_norm": 0.0,
      "learning_rate": 4.47091225357557e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 2191
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.78125,
      "epoch": 0.10591418631619637,
      "grad_norm": 0.0,
      "learning_rate": 4.470670660997294e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2192
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.21875,
      "epoch": 0.10596250483185156,
      "grad_norm": 0.0,
      "learning_rate": 4.470429068419018e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2193
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 112.9375,
      "epoch": 0.10601082334750676,
      "grad_norm": 22.759169705034846,
      "learning_rate": 4.470187475840742e-07,
      "loss": -0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2194
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.21875,
      "epoch": 0.10605914186316197,
      "grad_norm": 0.0,
      "learning_rate": 4.4699458832624657e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2195
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 103.8125,
      "epoch": 0.10610746037881716,
      "grad_norm": 12.548217215195194,
      "learning_rate": 4.46970429068419e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.317629873752594,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2196
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.5,
      "epoch": 0.10615577889447236,
      "grad_norm": 0.0,
      "learning_rate": 4.469462698105914e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2197
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 103.96875,
      "epoch": 0.10620409741012755,
      "grad_norm": 7.329449725960835,
      "learning_rate": 4.469221105527638e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2198
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 116.03125,
      "epoch": 0.10625241592578276,
      "grad_norm": 4.503623148592177,
      "learning_rate": 4.468979512949362e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2199
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.09375,
      "epoch": 0.10630073444143796,
      "grad_norm": 0.0,
      "learning_rate": 4.4687379203710864e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2200
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 102.0,
      "epoch": 0.10634905295709315,
      "grad_norm": 10.4195957241259,
      "learning_rate": 4.46849632779281e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.3541097044944763,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2201
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 115.3125,
      "epoch": 0.10639737147274836,
      "grad_norm": 10.927663205066635,
      "learning_rate": 4.468254735214534e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.27564918994903564,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2202
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.65625,
      "epoch": 0.10644568998840355,
      "grad_norm": 0.0,
      "learning_rate": 4.4680131426362583e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2203
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 119.8125,
      "epoch": 0.10649400850405875,
      "grad_norm": 0.0,
      "learning_rate": 4.467771550057982e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2204
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 118.46875,
      "epoch": 0.10654232701971396,
      "grad_norm": 8.323975685922115,
      "learning_rate": 4.467529957479706e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2205
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 125.15625,
      "epoch": 0.10659064553536915,
      "grad_norm": 0.0,
      "learning_rate": 4.4672883649014296e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2206
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.90625,
      "epoch": 0.10663896405102435,
      "grad_norm": 0.0,
      "learning_rate": 4.4670467723231543e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2207
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.0,
      "epoch": 0.10668728256667954,
      "grad_norm": 0.0,
      "learning_rate": 4.466805179744878e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2208
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.34375,
      "epoch": 0.10673560108233475,
      "grad_norm": 0.0,
      "learning_rate": 4.466563587166602e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2209
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.25,
      "epoch": 0.10678391959798995,
      "grad_norm": 0.0,
      "learning_rate": 4.466321994588326e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2210
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.0625,
      "epoch": 0.10683223811364514,
      "grad_norm": 0.0,
      "learning_rate": 4.4660804020100503e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2211
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.9375,
      "epoch": 0.10688055662930035,
      "grad_norm": 0.0,
      "learning_rate": 4.465838809431774e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2212
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 116.28125,
      "epoch": 0.10692887514495554,
      "grad_norm": 11.86154529241539,
      "learning_rate": 4.4655972168534986e-07,
      "loss": 0.0,
      "reward": 1.390625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 2213
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 101.4375,
      "epoch": 0.10697719366061074,
      "grad_norm": 12.10985384017556,
      "learning_rate": 4.465355624275222e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2651395797729492,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2214
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.71875,
      "epoch": 0.10702551217626595,
      "grad_norm": 0.0,
      "learning_rate": 4.465114031696946e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2215
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 119.46875,
      "epoch": 0.10707383069192114,
      "grad_norm": 22.2400836960511,
      "learning_rate": 4.46487243911867e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.18600594997406006,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2216
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.25,
      "epoch": 0.10712214920757634,
      "grad_norm": 0.0,
      "learning_rate": 4.464630846540394e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 2217
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.125,
      "epoch": 0.10717046772323154,
      "grad_norm": 0.0,
      "learning_rate": 4.464389253962118e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2218
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.875,
      "epoch": 0.10721878623888674,
      "grad_norm": 0.0,
      "learning_rate": 4.464147661383842e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2219
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 114.9375,
      "epoch": 0.10726710475454194,
      "grad_norm": 0.0,
      "learning_rate": 4.4639060688055665e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2220
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.25,
      "epoch": 0.10731542327019714,
      "grad_norm": 0.0,
      "learning_rate": 4.46366447622729e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2221
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.21875,
      "epoch": 0.10736374178585234,
      "grad_norm": 0.0,
      "learning_rate": 4.463422883649014e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2222
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 124.15625,
      "epoch": 0.10741206030150753,
      "grad_norm": 0.0,
      "learning_rate": 4.4631812910707384e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2223
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.8125,
      "epoch": 0.10746037881716274,
      "grad_norm": 0.0,
      "learning_rate": 4.462939698492462e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2224
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.875,
      "epoch": 0.10750869733281794,
      "grad_norm": 0.0,
      "learning_rate": 4.462698105914186e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2225
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.09375,
      "epoch": 0.10755701584847313,
      "grad_norm": 0.0,
      "learning_rate": 4.4624565133359097e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2226
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 103.5625,
      "epoch": 0.10760533436412834,
      "grad_norm": 4.486693253649576,
      "learning_rate": 4.4622149207576344e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.18174266815185547,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2227
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.4375,
      "epoch": 0.10765365287978353,
      "grad_norm": 0.0,
      "learning_rate": 4.461973328179358e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2228
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.875,
      "epoch": 0.10770197139543873,
      "grad_norm": 0.0,
      "learning_rate": 4.461731735601082e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2229
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 110.3125,
      "epoch": 0.10775028991109394,
      "grad_norm": 9.361341985317075,
      "learning_rate": 4.461490143022806e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.20633062720298767,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2230
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.25,
      "completion_length": 107.875,
      "epoch": 0.10779860842674913,
      "grad_norm": 11.390827331181477,
      "learning_rate": 4.4612485504445304e-07,
      "loss": -0.0,
      "reward": 1.671875,
      "reward_std": 0.39774754643440247,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2231
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 123.84375,
      "epoch": 0.10784692694240433,
      "grad_norm": 0.0,
      "learning_rate": 4.461006957866254e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2232
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 112.03125,
      "epoch": 0.10789524545805952,
      "grad_norm": 7.963468246304003,
      "learning_rate": 4.4607653652879787e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.2657213509082794,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2233
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 103.84375,
      "epoch": 0.10794356397371473,
      "grad_norm": 5.859164329463464,
      "learning_rate": 4.4605237727097023e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.28247910737991333,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2234
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.375,
      "epoch": 0.10799188248936993,
      "grad_norm": 0.0,
      "learning_rate": 4.460282180131426e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2235
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.875,
      "epoch": 0.10804020100502512,
      "grad_norm": 0.0,
      "learning_rate": 4.46004058755315e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2236
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.09375,
      "epoch": 0.10808851952068033,
      "grad_norm": 0.0,
      "learning_rate": 4.459798994974874e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2237
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 111.46875,
      "epoch": 0.10813683803633552,
      "grad_norm": 0.0,
      "learning_rate": 4.4595574023965983e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2238
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.625,
      "epoch": 0.10818515655199072,
      "grad_norm": 0.0,
      "learning_rate": 4.459315809818322e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2239
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 131.65625,
      "epoch": 0.10823347506764593,
      "grad_norm": 0.0,
      "learning_rate": 4.4590742172400466e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2240
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.59375,
      "epoch": 0.10828179358330112,
      "grad_norm": 0.0,
      "learning_rate": 4.45883262466177e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2241
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.375,
      "epoch": 0.10833011209895632,
      "grad_norm": 0.0,
      "learning_rate": 4.4585910320834943e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2242
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 96.5,
      "epoch": 0.10837843061461151,
      "grad_norm": 25.950289902984824,
      "learning_rate": 4.4583494395052184e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.37325698137283325,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2243
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.15625,
      "epoch": 0.10842674913026672,
      "grad_norm": 0.0,
      "learning_rate": 4.458107846926942e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2244
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.34375,
      "epoch": 0.10847506764592192,
      "grad_norm": 0.0,
      "learning_rate": 4.457866254348666e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2245
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.09375,
      "epoch": 0.10852338616157711,
      "grad_norm": 0.0,
      "learning_rate": 4.45762466177039e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2246
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.375,
      "epoch": 0.10857170467723232,
      "grad_norm": 0.0,
      "learning_rate": 4.4573830691921145e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2247
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.125,
      "epoch": 0.10862002319288751,
      "grad_norm": 0.0,
      "learning_rate": 4.457141476613838e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2248
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.0625,
      "epoch": 0.10866834170854271,
      "grad_norm": 0.0,
      "learning_rate": 4.456899884035562e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2249
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.0,
      "epoch": 0.10871666022419792,
      "grad_norm": 0.0,
      "learning_rate": 4.4566582914572863e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2250
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.6875,
      "epoch": 0.10876497873985311,
      "grad_norm": 0.0,
      "learning_rate": 4.4564166988790105e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 2251
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 121.625,
      "epoch": 0.10881329725550831,
      "grad_norm": 0.0,
      "learning_rate": 4.456175106300734e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2252
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 115.34375,
      "epoch": 0.1088616157711635,
      "grad_norm": 13.77760078597136,
      "learning_rate": 4.455933513722459e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.31663835048675537,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2253
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.375,
      "epoch": 0.10890993428681871,
      "grad_norm": 0.0,
      "learning_rate": 4.4556919211441824e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2254
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 119.28125,
      "epoch": 0.10895825280247391,
      "grad_norm": 0.0,
      "learning_rate": 4.455450328565906e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2255
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 118.9375,
      "epoch": 0.1090065713181291,
      "grad_norm": 0.0,
      "learning_rate": 4.45520873598763e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 2256
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 129.6875,
      "epoch": 0.10905488983378431,
      "grad_norm": 0.0,
      "learning_rate": 4.454967143409354e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 2257
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 132.84375,
      "epoch": 0.1091032083494395,
      "grad_norm": 8.958565343537183,
      "learning_rate": 4.4547255508310784e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.2662131190299988,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2258
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 130.40625,
      "epoch": 0.1091515268650947,
      "grad_norm": 11.060134465932581,
      "learning_rate": 4.454483958252802e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.2651395797729492,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 2259
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.78125,
      "epoch": 0.10919984538074991,
      "grad_norm": 0.0,
      "learning_rate": 4.4542423656745266e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2260
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 117.5625,
      "epoch": 0.1092481638964051,
      "grad_norm": 7.655551704191156,
      "learning_rate": 4.45400077309625e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2261
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.84375,
      "epoch": 0.1092964824120603,
      "grad_norm": 0.0,
      "learning_rate": 4.4537591805179744e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2262
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 119.90625,
      "epoch": 0.1093448009277155,
      "grad_norm": 11.091939931824625,
      "learning_rate": 4.4535175879396985e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.22835706174373627,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2263
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.1875,
      "epoch": 0.1093931194433707,
      "grad_norm": 0.0,
      "learning_rate": 4.4532759953614227e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2264
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.8125,
      "epoch": 0.1094414379590259,
      "grad_norm": 0.0,
      "learning_rate": 4.4530344027831463e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2265
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 124.125,
      "epoch": 0.1094897564746811,
      "grad_norm": 0.0,
      "learning_rate": 4.45279281020487e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2266
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.75,
      "epoch": 0.1095380749903363,
      "grad_norm": 0.0,
      "learning_rate": 4.4525512176265945e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2267
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.34375,
      "epoch": 0.1095863935059915,
      "grad_norm": 0.0,
      "learning_rate": 4.452309625048318e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2268
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 123.6875,
      "epoch": 0.1096347120216467,
      "grad_norm": 0.0,
      "learning_rate": 4.4520680324700423e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2269
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 128.96875,
      "epoch": 0.1096830305373019,
      "grad_norm": 0.0,
      "learning_rate": 4.4518264398917664e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2270
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 134.1875,
      "epoch": 0.10973134905295709,
      "grad_norm": 0.0,
      "learning_rate": 4.4515848473134906e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2271
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 109.21875,
      "epoch": 0.1097796675686123,
      "grad_norm": 0.0,
      "learning_rate": 4.451343254735214e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2272
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.625,
      "epoch": 0.1098279860842675,
      "grad_norm": 122.20159058697233,
      "learning_rate": 4.451101662156939e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2273
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 127.5625,
      "epoch": 0.10987630459992269,
      "grad_norm": 0.0,
      "learning_rate": 4.4508600695786624e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2274
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 136.15625,
      "epoch": 0.1099246231155779,
      "grad_norm": 15.763872396576591,
      "learning_rate": 4.450618477000386e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2275
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.53125,
      "epoch": 0.10997294163123308,
      "grad_norm": 0.0,
      "learning_rate": 4.45037688442211e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2276
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 125.15625,
      "epoch": 0.11002126014688829,
      "grad_norm": 0.0,
      "learning_rate": 4.4501352918438343e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2277
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 128.65625,
      "epoch": 0.1100695786625435,
      "grad_norm": 0.0,
      "learning_rate": 4.4498936992655585e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2278
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 106.5625,
      "epoch": 0.11011789717819868,
      "grad_norm": 8.871924025809783,
      "learning_rate": 4.449652106687282e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.20785759389400482,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2279
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 130.125,
      "epoch": 0.11016621569385389,
      "grad_norm": 0.0,
      "learning_rate": 4.4494105141090067e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 2280
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.4375,
      "epoch": 0.11021453420950908,
      "grad_norm": 0.0,
      "learning_rate": 4.4491689215307303e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2281
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 130.25,
      "epoch": 0.11026285272516428,
      "grad_norm": 0.0,
      "learning_rate": 4.4489273289524545e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2282
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.9375,
      "epoch": 0.11031117124081949,
      "grad_norm": 0.0,
      "learning_rate": 4.4486857363741786e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2283
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 109.3125,
      "epoch": 0.11035948975647468,
      "grad_norm": 5.134154974592399,
      "learning_rate": 4.448444143795903e-07,
      "loss": -0.0,
      "reward": 2.359375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2284
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.71875,
      "epoch": 0.11040780827212988,
      "grad_norm": 0.0,
      "learning_rate": 4.4482025512176263e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2285
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 120.9375,
      "epoch": 0.11045612678778507,
      "grad_norm": 0.0,
      "learning_rate": 4.44796095863935e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2286
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 118.0625,
      "epoch": 0.11050444530344028,
      "grad_norm": 10.18492427266892,
      "learning_rate": 4.4477193660610746e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.31539320945739746,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2287
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 116.25,
      "epoch": 0.11055276381909548,
      "grad_norm": 8.508765341311667,
      "learning_rate": 4.447477773482798e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2288
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.3125,
      "epoch": 0.11060108233475067,
      "grad_norm": 0.0,
      "learning_rate": 4.4472361809045224e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2289
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.3125,
      "epoch": 0.11064940085040588,
      "grad_norm": 0.0,
      "learning_rate": 4.4469945883262465e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2290
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.6875,
      "epoch": 0.11069771936606107,
      "grad_norm": 0.0,
      "learning_rate": 4.4467529957479706e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2291
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.65625,
      "epoch": 0.11074603788171627,
      "grad_norm": 0.0,
      "learning_rate": 4.446511403169694e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2292
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.1875,
      "epoch": 0.11079435639737148,
      "grad_norm": 0.0,
      "learning_rate": 4.446269810591419e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2293
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.96875,
      "epoch": 0.11084267491302667,
      "grad_norm": 0.0,
      "learning_rate": 4.4460282180131425e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2294
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.5,
      "epoch": 0.11089099342868187,
      "grad_norm": 0.0,
      "learning_rate": 4.445786625434866e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2295
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 125.21875,
      "epoch": 0.11093931194433707,
      "grad_norm": 0.0,
      "learning_rate": 4.44554503285659e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2296
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 117.375,
      "epoch": 0.11098763045999227,
      "grad_norm": 0.0,
      "learning_rate": 4.4453034402783144e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2297
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 117.09375,
      "epoch": 0.11103594897564747,
      "grad_norm": 4.0799746680515865,
      "learning_rate": 4.4450618477000385e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2298
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 127.25,
      "epoch": 0.11108426749130267,
      "grad_norm": 0.0,
      "learning_rate": 4.444820255121762e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2299
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.90625,
      "epoch": 0.11113258600695787,
      "grad_norm": 0.0,
      "learning_rate": 4.444578662543487e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2300
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.6875,
      "epoch": 0.11118090452261306,
      "grad_norm": 0.0,
      "learning_rate": 4.4443370699652104e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2301
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 114.65625,
      "epoch": 0.11122922303826827,
      "grad_norm": 4.798614783808365,
      "learning_rate": 4.4440954773869345e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2302
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.96875,
      "epoch": 0.11127754155392347,
      "grad_norm": 0.0,
      "learning_rate": 4.4438538848086587e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2303
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.75,
      "epoch": 0.11132586006957866,
      "grad_norm": 0.0,
      "learning_rate": 4.443612292230383e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2304
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.09375,
      "epoch": 0.11137417858523387,
      "grad_norm": 0.0,
      "learning_rate": 4.4433706996521064e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2305
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.34375,
      "epoch": 0.11142249710088906,
      "grad_norm": 0.0,
      "learning_rate": 4.44312910707383e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2306
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.40625,
      "epoch": 0.11147081561654426,
      "grad_norm": 0.0,
      "learning_rate": 4.4428875144955547e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2307
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 125.28125,
      "epoch": 0.11151913413219947,
      "grad_norm": 0.0,
      "learning_rate": 4.4426459219172783e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2308
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.53125,
      "epoch": 0.11156745264785466,
      "grad_norm": 0.0,
      "learning_rate": 4.4424043293390024e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2309
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.375,
      "epoch": 0.11161577116350986,
      "grad_norm": 0.0,
      "learning_rate": 4.4421627367607266e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2310
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.34375,
      "epoch": 0.11166408967916505,
      "grad_norm": 0.0,
      "learning_rate": 4.4419211441824507e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2311
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.71875,
      "epoch": 0.11171240819482026,
      "grad_norm": 0.0,
      "learning_rate": 4.4416795516041743e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2312
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.90625,
      "epoch": 0.11176072671047546,
      "grad_norm": 0.0,
      "learning_rate": 4.441437959025899e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 2313
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 114.65625,
      "epoch": 0.11180904522613065,
      "grad_norm": 17.002890338230056,
      "learning_rate": 4.4411963664476226e-07,
      "loss": -0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2314
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.21875,
      "epoch": 0.11185736374178586,
      "grad_norm": 0.0,
      "learning_rate": 4.4409547738693467e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2315
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 112.6875,
      "epoch": 0.11190568225744105,
      "grad_norm": 0.0,
      "learning_rate": 4.4407131812910703e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2316
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 100.71875,
      "epoch": 0.11195400077309625,
      "grad_norm": 11.835089327640436,
      "learning_rate": 4.4404715887127945e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.3561340868473053,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2317
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.0625,
      "epoch": 0.11200231928875146,
      "grad_norm": 0.0,
      "learning_rate": 4.4402299961345186e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2318
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 117.40625,
      "epoch": 0.11205063780440665,
      "grad_norm": 0.0,
      "learning_rate": 4.439988403556242e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2319
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.46875,
      "epoch": 0.11209895632006185,
      "grad_norm": 0.0,
      "learning_rate": 4.439746810977967e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 2320
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 116.75,
      "epoch": 0.11214727483571704,
      "grad_norm": 16.814583041826317,
      "learning_rate": 4.4395052183996905e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2321
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.375,
      "epoch": 0.11219559335137225,
      "grad_norm": 0.0,
      "learning_rate": 4.4392636258214146e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2322
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 126.0625,
      "epoch": 0.11224391186702745,
      "grad_norm": 0.0,
      "learning_rate": 4.439022033243139e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2323
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 109.875,
      "epoch": 0.11229223038268264,
      "grad_norm": 36.915162593679334,
      "learning_rate": 4.438780440664863e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.24359199404716492,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2324
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.96875,
      "epoch": 0.11234054889833785,
      "grad_norm": 0.0,
      "learning_rate": 4.4385388480865865e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 2325
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 123.625,
      "epoch": 0.11238886741399304,
      "grad_norm": 5.635728875145441,
      "learning_rate": 4.43829725550831e-07,
      "loss": -0.0,
      "reward": 2.40625,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2326
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.875,
      "epoch": 0.11243718592964824,
      "grad_norm": 0.0,
      "learning_rate": 4.438055662930035e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2327
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.65625,
      "epoch": 0.11248550444530345,
      "grad_norm": 0.0,
      "learning_rate": 4.4378140703517584e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2328
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.09375,
      "epoch": 0.11253382296095864,
      "grad_norm": 0.0,
      "learning_rate": 4.4375724777734825e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2329
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 112.3125,
      "epoch": 0.11258214147661384,
      "grad_norm": 11.329796171566937,
      "learning_rate": 4.4373308851952067e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.27564918994903564,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2330
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.90625,
      "epoch": 0.11263045999226903,
      "grad_norm": 0.0,
      "learning_rate": 4.437089292616931e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 2331
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 114.4375,
      "epoch": 0.11267877850792424,
      "grad_norm": 16.99880866393937,
      "learning_rate": 4.4368477000386544e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2332
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.625,
      "epoch": 0.11272709702357944,
      "grad_norm": 0.0,
      "learning_rate": 4.436606107460379e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2333
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.84375,
      "epoch": 0.11277541553923463,
      "grad_norm": 0.0,
      "learning_rate": 4.4363645148821027e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2334
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.75,
      "epoch": 0.11282373405488984,
      "grad_norm": 0.0,
      "learning_rate": 4.436122922303827e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2335
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.84375,
      "epoch": 0.11287205257054503,
      "grad_norm": 0.0,
      "learning_rate": 4.4358813297255504e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2336
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.9375,
      "epoch": 0.11292037108620023,
      "grad_norm": 0.0,
      "learning_rate": 4.4356397371472746e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 2337
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.78125,
      "epoch": 0.11296868960185544,
      "grad_norm": 0.0,
      "learning_rate": 4.4353981445689987e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2338
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.65625,
      "epoch": 0.11301700811751063,
      "grad_norm": 0.0,
      "learning_rate": 4.4351565519907223e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2339
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.40625,
      "epoch": 0.11306532663316583,
      "grad_norm": 0.0,
      "learning_rate": 4.434914959412447e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2340
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.09375,
      "epoch": 0.11311364514882102,
      "grad_norm": 0.0,
      "learning_rate": 4.4346733668341706e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2341
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.96875,
      "epoch": 0.11316196366447623,
      "grad_norm": 0.0,
      "learning_rate": 4.4344317742558947e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2342
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 120.875,
      "epoch": 0.11321028218013143,
      "grad_norm": 0.0,
      "learning_rate": 4.434190181677619e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2343
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 106.09375,
      "epoch": 0.11325860069578662,
      "grad_norm": 46.95395588267377,
      "learning_rate": 4.433948589099343e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2344
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 130.15625,
      "epoch": 0.11330691921144183,
      "grad_norm": 0.0,
      "learning_rate": 4.4337069965210666e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2345
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 120.125,
      "epoch": 0.11335523772709702,
      "grad_norm": 9.461986219351061,
      "learning_rate": 4.43346540394279e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.20633062720298767,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2346
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.84375,
      "epoch": 0.11340355624275222,
      "grad_norm": 0.0,
      "learning_rate": 4.433223811364515e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2347
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.09375,
      "epoch": 0.11345187475840743,
      "grad_norm": 0.0,
      "learning_rate": 4.4329822187862385e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2348
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.9375,
      "epoch": 0.11350019327406262,
      "grad_norm": 0.0,
      "learning_rate": 4.4327406262079626e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2349
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 127.25,
      "epoch": 0.11354851178971782,
      "grad_norm": 5.242756517325904,
      "learning_rate": 4.4324990336296867e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2350
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 122.21875,
      "epoch": 0.11359683030537301,
      "grad_norm": 0.0,
      "learning_rate": 4.432257441051411e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2351
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.40625,
      "epoch": 0.11364514882102822,
      "grad_norm": 0.0,
      "learning_rate": 4.4320158484731345e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2352
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.9375,
      "epoch": 0.11369346733668342,
      "grad_norm": 0.0,
      "learning_rate": 4.431774255894859e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2353
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 122.71875,
      "epoch": 0.11374178585233861,
      "grad_norm": 4.693169919515703,
      "learning_rate": 4.431532663316583e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2354
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.71875,
      "epoch": 0.11379010436799382,
      "grad_norm": 0.0,
      "learning_rate": 4.431291070738307e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2355
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.75,
      "epoch": 0.11383842288364901,
      "grad_norm": 0.0,
      "learning_rate": 4.4310494781600305e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2356
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.1875,
      "epoch": 0.11388674139930421,
      "grad_norm": 0.0,
      "learning_rate": 4.4308078855817546e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 2357
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.75,
      "epoch": 0.11393505991495942,
      "grad_norm": 0.0,
      "learning_rate": 4.430566293003479e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 2358
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.4375,
      "epoch": 0.11398337843061461,
      "grad_norm": 0.0,
      "learning_rate": 4.4303247004252024e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2359
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.40625,
      "epoch": 0.11403169694626981,
      "grad_norm": 0.0,
      "learning_rate": 4.430083107846927e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2360
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.03125,
      "epoch": 0.114080015461925,
      "grad_norm": 0.0,
      "learning_rate": 4.4298415152686506e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2361
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 121.0625,
      "epoch": 0.11412833397758021,
      "grad_norm": 0.0,
      "learning_rate": 4.429599922690375e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 2362
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.09375,
      "epoch": 0.11417665249323541,
      "grad_norm": 0.0,
      "learning_rate": 4.429358330112099e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2363
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.6875,
      "epoch": 0.1142249710088906,
      "grad_norm": 0.0,
      "learning_rate": 4.429116737533823e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2364
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.78125,
      "epoch": 0.11427328952454581,
      "grad_norm": 0.0,
      "learning_rate": 4.4288751449555467e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2365
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 117.03125,
      "epoch": 0.114321608040201,
      "grad_norm": 17.094166302492297,
      "learning_rate": 4.428633552377271e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.2529362142086029,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2366
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.71875,
      "epoch": 0.1143699265558562,
      "grad_norm": 0.0,
      "learning_rate": 4.428391959798995e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2367
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.25,
      "epoch": 0.11441824507151141,
      "grad_norm": 0.0,
      "learning_rate": 4.4281503672207185e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2368
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.46875,
      "epoch": 0.1144665635871666,
      "grad_norm": 0.0,
      "learning_rate": 4.4279087746424427e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2369
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.75,
      "epoch": 0.1145148821028218,
      "grad_norm": 0.0,
      "learning_rate": 4.427667182064167e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2370
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.71875,
      "epoch": 0.114563200618477,
      "grad_norm": 0.0,
      "learning_rate": 4.427425589485891e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2371
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 106.875,
      "epoch": 0.1146115191341322,
      "grad_norm": 29.275022496641334,
      "learning_rate": 4.4271839969076146e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2372
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.65625,
      "epoch": 0.1146598376497874,
      "grad_norm": 0.0,
      "learning_rate": 4.426942404329339e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2373
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 113.96875,
      "epoch": 0.1147081561654426,
      "grad_norm": 0.0,
      "learning_rate": 4.426700811751063e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 2374
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.3125,
      "epoch": 0.1147564746810978,
      "grad_norm": 0.0,
      "learning_rate": 4.426459219172787e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2375
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.875,
      "epoch": 0.11480479319675299,
      "grad_norm": 0.0,
      "learning_rate": 4.4262176265945106e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2376
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.15625,
      "epoch": 0.1148531117124082,
      "grad_norm": 0.0,
      "learning_rate": 4.425976034016235e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2377
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.0625,
      "epoch": 0.1149014302280634,
      "grad_norm": 0.0,
      "learning_rate": 4.425734441437959e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 2378
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.0,
      "epoch": 0.11494974874371859,
      "grad_norm": 0.0,
      "learning_rate": 4.4254928488596825e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2379
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.21875,
      "epoch": 0.1149980672593738,
      "grad_norm": 0.0,
      "learning_rate": 4.425251256281407e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2380
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 100.34375,
      "epoch": 0.11504638577502899,
      "grad_norm": 0.0,
      "learning_rate": 4.4250096637031307e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2381
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.0625,
      "epoch": 0.11509470429068419,
      "grad_norm": 0.0,
      "learning_rate": 4.424768071124855e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2382
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.0625,
      "epoch": 0.1151430228063394,
      "grad_norm": 0.0,
      "learning_rate": 4.424526478546579e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2383
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.65625,
      "epoch": 0.11519134132199459,
      "grad_norm": 0.0,
      "learning_rate": 4.424284885968303e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2384
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.5625,
      "epoch": 0.11523965983764979,
      "grad_norm": 0.0,
      "learning_rate": 4.424043293390027e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2385
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.40625,
      "epoch": 0.11528797835330498,
      "grad_norm": 0.0,
      "learning_rate": 4.423801700811751e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2386
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.21875,
      "epoch": 0.11533629686896019,
      "grad_norm": 0.0,
      "learning_rate": 4.423560108233475e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 2387
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.78125,
      "epoch": 0.11538461538461539,
      "grad_norm": 0.0,
      "learning_rate": 4.4233185156551986e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2388
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.0625,
      "epoch": 0.11543293390027058,
      "grad_norm": 0.0,
      "learning_rate": 4.423076923076923e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24251842498779297,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2389
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.75,
      "epoch": 0.11548125241592579,
      "grad_norm": 0.0,
      "learning_rate": 4.422835330498647e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2390
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.53125,
      "epoch": 0.11552957093158098,
      "grad_norm": 0.0,
      "learning_rate": 4.422593737920371e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2391
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 109.25,
      "epoch": 0.11557788944723618,
      "grad_norm": 5.776384718464013,
      "learning_rate": 4.4223521453420946e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2392
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.34375,
      "epoch": 0.11562620796289139,
      "grad_norm": 0.0,
      "learning_rate": 4.4221105527638193e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2393
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.78125,
      "epoch": 0.11567452647854658,
      "grad_norm": 0.0,
      "learning_rate": 4.421868960185543e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2394
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 102.875,
      "epoch": 0.11572284499420178,
      "grad_norm": 11.08274707599,
      "learning_rate": 4.421627367607267e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2395
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.65625,
      "epoch": 0.11577116350985697,
      "grad_norm": 0.0,
      "learning_rate": 4.4213857750289906e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2396
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.0625,
      "epoch": 0.11581948202551218,
      "grad_norm": 0.0,
      "learning_rate": 4.4211441824507153e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2397
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 96.6875,
      "epoch": 0.11586780054116738,
      "grad_norm": 7.642375367435261,
      "learning_rate": 4.420902589872439e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.33351296186447144,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2398
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.40625,
      "epoch": 0.11591611905682257,
      "grad_norm": 0.0,
      "learning_rate": 4.4206609972941625e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2399
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 95.46875,
      "epoch": 0.11596443757247778,
      "grad_norm": 11.972319944638311,
      "learning_rate": 4.420419404715887e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2982703447341919,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2400
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 98.0625,
      "epoch": 0.11601275608813297,
      "grad_norm": 12.617585074383653,
      "learning_rate": 4.420177812137611e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2401
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.375,
      "epoch": 0.11606107460378817,
      "grad_norm": 0.0,
      "learning_rate": 4.419936219559335e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 2402
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.84375,
      "epoch": 0.11610939311944338,
      "grad_norm": 0.0,
      "learning_rate": 4.419694626981059e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2403
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.84375,
      "epoch": 0.11615771163509857,
      "grad_norm": 0.0,
      "learning_rate": 4.419453034402783e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2404
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.4375,
      "epoch": 0.11620603015075377,
      "grad_norm": 0.0,
      "learning_rate": 4.419211441824507e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2405
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.6875,
      "epoch": 0.11625434866640896,
      "grad_norm": 0.0,
      "learning_rate": 4.418969849246231e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2406
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.4375,
      "epoch": 0.11630266718206417,
      "grad_norm": 0.0,
      "learning_rate": 4.418728256667955e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2407
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.15625,
      "epoch": 0.11635098569771937,
      "grad_norm": 0.0,
      "learning_rate": 4.4184866640896787e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2408
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.5,
      "epoch": 0.11639930421337456,
      "grad_norm": 0.0,
      "learning_rate": 4.418245071511403e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2409
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 105.90625,
      "epoch": 0.11644762272902977,
      "grad_norm": 0.0,
      "learning_rate": 4.418003478933127e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 2410
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 100.0,
      "epoch": 0.11649594124468496,
      "grad_norm": 8.278001375909298,
      "learning_rate": 4.417761886354851e-07,
      "loss": -0.0,
      "reward": 1.90625,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2411
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.875,
      "epoch": 0.11654425976034016,
      "grad_norm": 0.0,
      "learning_rate": 4.4175202937765747e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2412
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.59375,
      "epoch": 0.11659257827599537,
      "grad_norm": 0.0,
      "learning_rate": 4.4172787011982994e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2413
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.9375,
      "epoch": 0.11664089679165056,
      "grad_norm": 0.0,
      "learning_rate": 4.417037108620023e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2414
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 100.21875,
      "epoch": 0.11668921530730576,
      "grad_norm": 9.301609866539755,
      "learning_rate": 4.416795516041747e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2415
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.25,
      "epoch": 0.11673753382296095,
      "grad_norm": 0.0,
      "learning_rate": 4.4165539234634707e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 2416
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.6875,
      "epoch": 0.11678585233861616,
      "grad_norm": 0.0,
      "learning_rate": 4.4163123308851954e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2417
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 113.3125,
      "epoch": 0.11683417085427136,
      "grad_norm": 8.5123249221937,
      "learning_rate": 4.416070738306919e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2418
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.875,
      "epoch": 0.11688248936992655,
      "grad_norm": 0.0,
      "learning_rate": 4.4158291457286426e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2419
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.28125,
      "epoch": 0.11693080788558176,
      "grad_norm": 0.0,
      "learning_rate": 4.4155875531503673e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2420
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 110.65625,
      "epoch": 0.11697912640123695,
      "grad_norm": 0.0,
      "learning_rate": 4.415345960572091e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2421
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.71875,
      "epoch": 0.11702744491689215,
      "grad_norm": 0.0,
      "learning_rate": 4.415104367993815e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2422
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 112.15625,
      "epoch": 0.11707576343254736,
      "grad_norm": 0.0,
      "learning_rate": 4.414862775415539e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 2423
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.03125,
      "epoch": 0.11712408194820255,
      "grad_norm": 0.0,
      "learning_rate": 4.4146211828372633e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2424
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.875,
      "epoch": 0.11717240046385775,
      "grad_norm": 0.0,
      "learning_rate": 4.414379590258987e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2425
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.3125,
      "epoch": 0.11722071897951294,
      "grad_norm": 0.0,
      "learning_rate": 4.414137997680711e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2426
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.90625,
      "epoch": 0.11726903749516815,
      "grad_norm": 0.0,
      "learning_rate": 4.413896405102435e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2427
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.0,
      "epoch": 0.11731735601082335,
      "grad_norm": 0.0,
      "learning_rate": 4.4136548125241593e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2428
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.09375,
      "epoch": 0.11736567452647854,
      "grad_norm": 0.0,
      "learning_rate": 4.413413219945883e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2429
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.15625,
      "epoch": 0.11741399304213375,
      "grad_norm": 0.0,
      "learning_rate": 4.413171627367607e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 2430
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.3125,
      "epoch": 0.11746231155778894,
      "grad_norm": 0.0,
      "learning_rate": 4.412930034789331e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2431
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 106.90625,
      "epoch": 0.11751063007344414,
      "grad_norm": 0.0,
      "learning_rate": 4.412688442211055e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 2432
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.3125,
      "epoch": 0.11755894858909935,
      "grad_norm": 0.0,
      "learning_rate": 4.4124468496327795e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2433
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.84375,
      "epoch": 0.11760726710475454,
      "grad_norm": 0.0,
      "learning_rate": 4.412205257054503e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2434
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.875,
      "epoch": 0.11765558562040974,
      "grad_norm": 0.0,
      "learning_rate": 4.411963664476227e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2435
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 104.65625,
      "epoch": 0.11770390413606494,
      "grad_norm": 0.0,
      "learning_rate": 4.411722071897951e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 2436
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.84375,
      "epoch": 0.11775222265172014,
      "grad_norm": 0.0,
      "learning_rate": 4.4114804793196755e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2437
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.6875,
      "epoch": 0.11780054116737534,
      "grad_norm": 0.0,
      "learning_rate": 4.411238886741399e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2438
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 115.9375,
      "epoch": 0.11784885968303054,
      "grad_norm": 6.991498058645009,
      "learning_rate": 4.4109972941631227e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.24039678275585175,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2439
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.1875,
      "epoch": 0.11789717819868574,
      "grad_norm": 0.0,
      "learning_rate": 4.4107557015848474e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2440
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.75,
      "epoch": 0.11794549671434093,
      "grad_norm": 0.0,
      "learning_rate": 4.410514109006571e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2441
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 106.34375,
      "epoch": 0.11799381522999614,
      "grad_norm": 14.321564516733394,
      "learning_rate": 4.410272516428295e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.323003351688385,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2442
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 94.1875,
      "epoch": 0.11804213374565134,
      "grad_norm": 20.202851883663786,
      "learning_rate": 4.410030923850019e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.3255898952484131,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2443
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 116.5625,
      "epoch": 0.11809045226130653,
      "grad_norm": 16.86063834170027,
      "learning_rate": 4.4097893312717434e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.35774436593055725,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2444
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.09375,
      "epoch": 0.11813877077696174,
      "grad_norm": 0.0,
      "learning_rate": 4.409547738693467e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2445
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 111.8125,
      "epoch": 0.11818708929261693,
      "grad_norm": 11.271775026486102,
      "learning_rate": 4.409306146115191e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2446
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 105.4375,
      "epoch": 0.11823540780827213,
      "grad_norm": 51.83608061991362,
      "learning_rate": 4.409064553536915e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.24280104041099548,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 2447
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.90625,
      "epoch": 0.11828372632392734,
      "grad_norm": 0.0,
      "learning_rate": 4.4088229609586394e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2448
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.1875,
      "epoch": 0.11833204483958253,
      "grad_norm": 0.0,
      "learning_rate": 4.408581368380363e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2449
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.9375,
      "epoch": 0.11838036335523773,
      "grad_norm": 0.0,
      "learning_rate": 4.408339775802087e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2450
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.75,
      "epoch": 0.11842868187089292,
      "grad_norm": 0.0,
      "learning_rate": 4.408098183223811e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2451
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.03125,
      "epoch": 0.11847700038654813,
      "grad_norm": 0.0,
      "learning_rate": 4.407856590645535e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2452
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.9375,
      "epoch": 0.11852531890220333,
      "grad_norm": 0.0,
      "learning_rate": 4.4076149980672595e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2453
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.5625,
      "epoch": 0.11857363741785852,
      "grad_norm": 0.0,
      "learning_rate": 4.407373405488983e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2454
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.28125,
      "epoch": 0.11862195593351373,
      "grad_norm": 0.0,
      "learning_rate": 4.4071318129107073e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2455
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.25,
      "epoch": 0.11867027444916892,
      "grad_norm": 0.0,
      "learning_rate": 4.406890220332431e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2456
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.59375,
      "epoch": 0.11871859296482412,
      "grad_norm": 0.0,
      "learning_rate": 4.4066486277541556e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2457
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.03125,
      "epoch": 0.11876691148047933,
      "grad_norm": 0.0,
      "learning_rate": 4.406407035175879e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2458
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.46875,
      "epoch": 0.11881522999613452,
      "grad_norm": 0.0,
      "learning_rate": 4.4061654425976033e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2459
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.125,
      "epoch": 0.11886354851178972,
      "grad_norm": 0.0,
      "learning_rate": 4.4059238500193274e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2460
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 107.53125,
      "epoch": 0.11891186702744491,
      "grad_norm": 0.0,
      "learning_rate": 4.405682257441051e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2461
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 106.46875,
      "epoch": 0.11896018554310012,
      "grad_norm": 11.737405844439477,
      "learning_rate": 4.405440664862775e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2462
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.03125,
      "epoch": 0.11900850405875532,
      "grad_norm": 0.0,
      "learning_rate": 4.4051990722844993e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2463
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.34375,
      "epoch": 0.11905682257441051,
      "grad_norm": 0.0,
      "learning_rate": 4.4049574797062234e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2464
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 118.34375,
      "epoch": 0.11910514109006572,
      "grad_norm": 14.269977397717227,
      "learning_rate": 4.404715887127947e-07,
      "loss": -0.0,
      "reward": 1.1875,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.125,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 2465
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.9375,
      "epoch": 0.11915345960572091,
      "grad_norm": 0.0,
      "learning_rate": 4.404474294549671e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 2466
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 122.53125,
      "epoch": 0.11920177812137611,
      "grad_norm": 0.0,
      "learning_rate": 4.4042327019713953e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 2467
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 112.0625,
      "epoch": 0.11925009663703132,
      "grad_norm": 0.0,
      "learning_rate": 4.4039911093931195e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2468
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 101.71875,
      "epoch": 0.11929841515268651,
      "grad_norm": 8.353648234376093,
      "learning_rate": 4.403749516814843e-07,
      "loss": -0.0,
      "reward": 2.09375,
      "reward_std": 0.2799758315086365,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2469
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.6875,
      "epoch": 0.11934673366834171,
      "grad_norm": 0.0,
      "learning_rate": 4.4035079242365677e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2470
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 111.5625,
      "epoch": 0.1193950521839969,
      "grad_norm": 0.0,
      "learning_rate": 4.4032663316582913e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2471
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 116.1875,
      "epoch": 0.11944337069965211,
      "grad_norm": 0.0,
      "learning_rate": 4.403024739080015e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2472
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.53125,
      "epoch": 0.11949168921530731,
      "grad_norm": 0.0,
      "learning_rate": 4.4027831465017396e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2473
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.625,
      "epoch": 0.1195400077309625,
      "grad_norm": 0.0,
      "learning_rate": 4.402541553923463e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2474
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.65625,
      "epoch": 0.11958832624661771,
      "grad_norm": 0.0,
      "learning_rate": 4.4022999613451874e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2475
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.90625,
      "epoch": 0.1196366447622729,
      "grad_norm": 0.0,
      "learning_rate": 4.402058368766911e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2476
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.96875,
      "epoch": 0.1196849632779281,
      "grad_norm": 0.0,
      "learning_rate": 4.4018167761886356e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 2477
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.125,
      "epoch": 0.11973328179358331,
      "grad_norm": 0.0,
      "learning_rate": 4.401575183610359e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2478
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.53125,
      "epoch": 0.1197816003092385,
      "grad_norm": 0.0,
      "learning_rate": 4.4013335910320834e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2479
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.53125,
      "epoch": 0.1198299188248937,
      "grad_norm": 0.0,
      "learning_rate": 4.4010919984538075e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2480
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.96875,
      "epoch": 0.1198782373405489,
      "grad_norm": 0.0,
      "learning_rate": 4.400850405875531e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2481
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.5625,
      "epoch": 0.1199265558562041,
      "grad_norm": 0.0,
      "learning_rate": 4.400608813297255e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2482
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.3125,
      "epoch": 0.1199748743718593,
      "grad_norm": 0.0,
      "learning_rate": 4.4003672207189794e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2483
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.9375,
      "epoch": 0.1200231928875145,
      "grad_norm": 0.0,
      "learning_rate": 4.4001256281407035e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2484
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.53125,
      "epoch": 0.1200715114031697,
      "grad_norm": 0.0,
      "learning_rate": 4.399884035562427e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2485
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.84375,
      "epoch": 0.12011982991882489,
      "grad_norm": 0.0,
      "learning_rate": 4.3996424429841513e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2486
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.875,
      "epoch": 0.1201681484344801,
      "grad_norm": 17.79740164741392,
      "learning_rate": 4.3994008504058754e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.2848735451698303,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2487
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.21875,
      "epoch": 0.1202164669501353,
      "grad_norm": 0.0,
      "learning_rate": 4.3991592578275995e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2488
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.9375,
      "epoch": 0.12026478546579049,
      "grad_norm": 0.0,
      "learning_rate": 4.398917665249323e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2489
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.75,
      "epoch": 0.1203131039814457,
      "grad_norm": 0.0,
      "learning_rate": 4.398676072671048e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2490
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.8125,
      "epoch": 0.12036142249710088,
      "grad_norm": 0.0,
      "learning_rate": 4.3984344800927714e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2491
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.5,
      "epoch": 0.12040974101275609,
      "grad_norm": 0.0,
      "learning_rate": 4.398192887514495e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2492
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 103.34375,
      "epoch": 0.1204580595284113,
      "grad_norm": 28.599515431536375,
      "learning_rate": 4.3979512949362197e-07,
      "loss": -0.0,
      "reward": 2.25,
      "reward_std": 0.26726123690605164,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2493
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.03125,
      "epoch": 0.12050637804406648,
      "grad_norm": 0.0,
      "learning_rate": 4.3977097023579433e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2494
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.875,
      "epoch": 0.12055469655972169,
      "grad_norm": 0.0,
      "learning_rate": 4.3974681097796674e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2495
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.375,
      "epoch": 0.12060301507537688,
      "grad_norm": 0.0,
      "learning_rate": 4.397226517201391e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2496
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.40625,
      "epoch": 0.12065133359103208,
      "grad_norm": 0.0,
      "learning_rate": 4.3969849246231157e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2497
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.0,
      "epoch": 0.12069965210668729,
      "grad_norm": 0.0,
      "learning_rate": 4.3967433320448393e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2498
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 130.40625,
      "epoch": 0.12074797062234248,
      "grad_norm": 0.0,
      "learning_rate": 4.3965017394665635e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 2499
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.5,
      "epoch": 0.12079628913799768,
      "grad_norm": 0.0,
      "learning_rate": 4.3962601468882876e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2500
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 114.3125,
      "epoch": 0.12084460765365287,
      "grad_norm": 9.439823269516593,
      "learning_rate": 4.396018554310011e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.30935919284820557,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2501
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 128.4375,
      "epoch": 0.12089292616930808,
      "grad_norm": 20.800473323168994,
      "learning_rate": 4.3957769617317353e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2502
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.09375,
      "epoch": 0.12094124468496328,
      "grad_norm": 0.0,
      "learning_rate": 4.3955353691534595e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2503
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.1875,
      "epoch": 0.12098956320061847,
      "grad_norm": 0.0,
      "learning_rate": 4.3952937765751836e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2504
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.71875,
      "epoch": 0.12103788171627368,
      "grad_norm": 0.0,
      "learning_rate": 4.395052183996907e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2505
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.09375,
      "epoch": 0.12108620023192887,
      "grad_norm": 0.0,
      "learning_rate": 4.3948105914186313e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2506
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.71875,
      "epoch": 0.12113451874758407,
      "grad_norm": 0.0,
      "learning_rate": 4.3945689988403555e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2507
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.9375,
      "epoch": 0.12118283726323928,
      "grad_norm": 0.0,
      "learning_rate": 4.3943274062620796e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2508
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.3125,
      "epoch": 0.12123115577889447,
      "grad_norm": 0.0,
      "learning_rate": 4.394085813683803e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2509
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 110.40625,
      "epoch": 0.12127947429454967,
      "grad_norm": 13.768412183216858,
      "learning_rate": 4.393844221105528e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.3966739773750305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2510
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.03125,
      "epoch": 0.12132779281020487,
      "grad_norm": 0.0,
      "learning_rate": 4.3936026285272515e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2511
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.65625,
      "epoch": 0.12137611132586007,
      "grad_norm": 0.0,
      "learning_rate": 4.393361035948975e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2512
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.53125,
      "epoch": 0.12142442984151527,
      "grad_norm": 0.0,
      "learning_rate": 4.3931194433707e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2513
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.8125,
      "epoch": 0.12147274835717047,
      "grad_norm": 0.0,
      "learning_rate": 4.3928778507924234e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2514
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 114.21875,
      "epoch": 0.12152106687282567,
      "grad_norm": 0.0,
      "learning_rate": 4.3926362582141475e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2515
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.46875,
      "epoch": 0.12156938538848086,
      "grad_norm": 0.0,
      "learning_rate": 4.392394665635871e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2516
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.71875,
      "epoch": 0.12161770390413607,
      "grad_norm": 0.0,
      "learning_rate": 4.392153073057596e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2517
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 111.75,
      "epoch": 0.12166602241979127,
      "grad_norm": 7.643451970273937,
      "learning_rate": 4.3919114804793194e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2518
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.46875,
      "epoch": 0.12171434093544646,
      "grad_norm": 0.0,
      "learning_rate": 4.3916698879010435e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2519
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.5,
      "epoch": 0.12176265945110167,
      "grad_norm": 0.0,
      "learning_rate": 4.3914282953227677e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2520
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.65625,
      "epoch": 0.12181097796675686,
      "grad_norm": 0.0,
      "learning_rate": 4.391186702744492e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2521
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.75,
      "epoch": 0.12185929648241206,
      "grad_norm": 0.0,
      "learning_rate": 4.3909451101662154e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2522
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.4375,
      "epoch": 0.12190761499806727,
      "grad_norm": 0.0,
      "learning_rate": 4.3907035175879395e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2523
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 114.40625,
      "epoch": 0.12195593351372246,
      "grad_norm": 0.0,
      "learning_rate": 4.3904619250096637e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2524
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.65625,
      "epoch": 0.12200425202937766,
      "grad_norm": 0.0,
      "learning_rate": 4.3902203324313873e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2525
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.5,
      "epoch": 0.12205257054503285,
      "grad_norm": 0.0,
      "learning_rate": 4.3899787398531114e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2526
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.53125,
      "epoch": 0.12210088906068806,
      "grad_norm": 0.0,
      "learning_rate": 4.3897371472748356e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2527
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.0,
      "epoch": 0.12214920757634326,
      "grad_norm": 0.0,
      "learning_rate": 4.3894955546965597e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2528
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 111.09375,
      "epoch": 0.12219752609199845,
      "grad_norm": 36.24516414159874,
      "learning_rate": 4.3892539621182833e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.32512497901916504,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2529
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 98.9375,
      "epoch": 0.12224584460765366,
      "grad_norm": 24.62841068203343,
      "learning_rate": 4.389012369540008e-07,
      "loss": -0.0,
      "reward": 2.296875,
      "reward_std": 0.188242569565773,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2530
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.75,
      "epoch": 0.12229416312330885,
      "grad_norm": 0.0,
      "learning_rate": 4.3887707769617316e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2531
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.5,
      "epoch": 0.12234248163896405,
      "grad_norm": 0.0,
      "learning_rate": 4.388529184383455e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2532
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 108.03125,
      "epoch": 0.12239080015461926,
      "grad_norm": 7.550644585351254,
      "learning_rate": 4.38828759180518e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2533
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.3125,
      "epoch": 0.12243911867027445,
      "grad_norm": 0.0,
      "learning_rate": 4.3880459992269035e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2534
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 96.34375,
      "epoch": 0.12248743718592965,
      "grad_norm": 4.5196232402624075,
      "learning_rate": 4.3878044066486276e-07,
      "loss": -0.0,
      "reward": 2.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 2535
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 112.53125,
      "epoch": 0.12253575570158484,
      "grad_norm": 0.0,
      "learning_rate": 4.387562814070351e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 2536
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 107.21875,
      "epoch": 0.12258407421724005,
      "grad_norm": 0.0,
      "learning_rate": 4.387321221492076e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 2537
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.53125,
      "epoch": 0.12263239273289525,
      "grad_norm": 0.0,
      "learning_rate": 4.3870796289137995e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2538
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 108.75,
      "epoch": 0.12268071124855044,
      "grad_norm": 9.541419140579942,
      "learning_rate": 4.3868380363355236e-07,
      "loss": -0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2539
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.09375,
      "epoch": 0.12272902976420565,
      "grad_norm": 0.0,
      "learning_rate": 4.386596443757248e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2540
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 98.25,
      "epoch": 0.12277734827986084,
      "grad_norm": 5.1370191944160535,
      "learning_rate": 4.386354851178972e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2541
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.125,
      "epoch": 0.12282566679551604,
      "grad_norm": 0.0,
      "learning_rate": 4.3861132586006955e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2542
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.4375,
      "epoch": 0.12287398531117125,
      "grad_norm": 0.0,
      "learning_rate": 4.3858716660224196e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2543
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.40625,
      "epoch": 0.12292230382682644,
      "grad_norm": 0.0,
      "learning_rate": 4.385630073444144e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2544
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.8125,
      "epoch": 0.12297062234248164,
      "grad_norm": 0.0,
      "learning_rate": 4.3853884808658674e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2545
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 105.78125,
      "epoch": 0.12301894085813683,
      "grad_norm": 0.0,
      "learning_rate": 4.3851468882875915e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2546
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 106.125,
      "epoch": 0.12306725937379204,
      "grad_norm": 4.613150050306641,
      "learning_rate": 4.3849052957093156e-07,
      "loss": -0.0,
      "reward": 1.625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2547
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.6875,
      "epoch": 0.12311557788944724,
      "grad_norm": 0.0,
      "learning_rate": 4.38466370313104e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2548
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.84375,
      "epoch": 0.12316389640510243,
      "grad_norm": 0.0,
      "learning_rate": 4.3844221105527634e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2549
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.34375,
      "epoch": 0.12321221492075764,
      "grad_norm": 0.0,
      "learning_rate": 4.384180517974488e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2550
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.5625,
      "epoch": 0.12326053343641283,
      "grad_norm": 0.0,
      "learning_rate": 4.3839389253962117e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2551
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.96875,
      "epoch": 0.12330885195206803,
      "grad_norm": 0.0,
      "learning_rate": 4.383697332817935e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2552
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.5625,
      "epoch": 0.12335717046772324,
      "grad_norm": 0.0,
      "learning_rate": 4.38345574023966e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2553
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.25,
      "epoch": 0.12340548898337843,
      "grad_norm": 0.0,
      "learning_rate": 4.3832141476613835e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2554
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.5625,
      "epoch": 0.12345380749903363,
      "grad_norm": 0.0,
      "learning_rate": 4.3829725550831077e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2555
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.875,
      "epoch": 0.12350212601468882,
      "grad_norm": 21.311557296831953,
      "learning_rate": 4.3827309625048313e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.24386970698833466,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2556
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 107.9375,
      "epoch": 0.12355044453034403,
      "grad_norm": 8.952281660832371,
      "learning_rate": 4.382489369926556e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.3234682083129883,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2557
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.5625,
      "epoch": 0.12359876304599923,
      "grad_norm": 0.0,
      "learning_rate": 4.3822477773482796e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2558
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 107.6875,
      "epoch": 0.12364708156165442,
      "grad_norm": 7.114000581012142,
      "learning_rate": 4.3820061847700037e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.31300365924835205,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2559
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.65625,
      "epoch": 0.12369540007730963,
      "grad_norm": 0.0,
      "learning_rate": 4.381764592191728e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2560
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 114.03125,
      "epoch": 0.12374371859296482,
      "grad_norm": 7.218091757249254,
      "learning_rate": 4.381522999613452e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2561
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.6875,
      "epoch": 0.12379203710862002,
      "grad_norm": 0.0,
      "learning_rate": 4.3812814070351756e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2562
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 95.9375,
      "epoch": 0.12384035562427523,
      "grad_norm": 26.168108733445656,
      "learning_rate": 4.3810398144568997e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.2961388826370239,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2563
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.28125,
      "epoch": 0.12388867413993042,
      "grad_norm": 0.0,
      "learning_rate": 4.380798221878624e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2564
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 110.65625,
      "epoch": 0.12393699265558562,
      "grad_norm": 5.9800066699129335,
      "learning_rate": 4.3805566293003474e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2565
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 112.09375,
      "epoch": 0.12398531117124081,
      "grad_norm": 0.0,
      "learning_rate": 4.3803150367220716e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2566
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.0625,
      "epoch": 0.12403362968689602,
      "grad_norm": 0.0,
      "learning_rate": 4.3800734441437957e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2567
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.9375,
      "epoch": 0.12408194820255122,
      "grad_norm": 0.0,
      "learning_rate": 4.37983185156552e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2568
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.8125,
      "epoch": 0.12413026671820641,
      "grad_norm": 0.0,
      "learning_rate": 4.3795902589872435e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 2569
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.15625,
      "epoch": 0.12417858523386162,
      "grad_norm": 0.0,
      "learning_rate": 4.379348666408968e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2570
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.0625,
      "epoch": 0.12422690374951681,
      "grad_norm": 0.0,
      "learning_rate": 4.3791070738306917e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2571
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.21875,
      "epoch": 0.12427522226517201,
      "grad_norm": 0.0,
      "learning_rate": 4.378865481252416e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2572
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.0,
      "epoch": 0.12432354078082722,
      "grad_norm": 0.0,
      "learning_rate": 4.37862388867414e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2573
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.65625,
      "epoch": 0.12437185929648241,
      "grad_norm": 0.0,
      "learning_rate": 4.3783822960958636e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2574
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.875,
      "epoch": 0.12442017781213761,
      "grad_norm": 0.0,
      "learning_rate": 4.378140703517588e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2575
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.5,
      "epoch": 0.1244684963277928,
      "grad_norm": 0.0,
      "learning_rate": 4.3778991109393114e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2576
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.4375,
      "epoch": 0.12451681484344801,
      "grad_norm": 0.0,
      "learning_rate": 4.377657518361036e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2577
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 109.5625,
      "epoch": 0.12456513335910321,
      "grad_norm": 10.770670739790905,
      "learning_rate": 4.3774159257827596e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.29713040590286255,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2578
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.46875,
      "epoch": 0.1246134518747584,
      "grad_norm": 0.0,
      "learning_rate": 4.377174333204484e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2579
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.15625,
      "epoch": 0.12466177039041361,
      "grad_norm": 0.0,
      "learning_rate": 4.376932740626208e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2580
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.0,
      "epoch": 0.1247100889060688,
      "grad_norm": 0.0,
      "learning_rate": 4.376691148047932e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2581
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.78125,
      "epoch": 0.124758407421724,
      "grad_norm": 0.0,
      "learning_rate": 4.3764495554696556e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2582
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.6875,
      "epoch": 0.12480672593737921,
      "grad_norm": 0.0,
      "learning_rate": 4.3762079628913803e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2583
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.78125,
      "epoch": 0.1248550444530344,
      "grad_norm": 0.0,
      "learning_rate": 4.375966370313104e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2584
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.71875,
      "epoch": 0.1249033629686896,
      "grad_norm": 0.0,
      "learning_rate": 4.3757247777348275e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2585
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.4375,
      "epoch": 0.1249516814843448,
      "grad_norm": 0.0,
      "learning_rate": 4.3754831851565517e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2586
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.28125,
      "epoch": 0.125,
      "grad_norm": 0.0,
      "learning_rate": 4.375241592578276e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2587
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 124.25,
      "epoch": 0.1250483185156552,
      "grad_norm": 0.0,
      "learning_rate": 4.375e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2588
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.9375,
      "epoch": 0.1250966370313104,
      "grad_norm": 0.0,
      "learning_rate": 4.3747584074217235e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2589
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.1875,
      "epoch": 0.12514495554696559,
      "grad_norm": 0.0,
      "learning_rate": 4.374516814843448e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2590
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.71875,
      "epoch": 0.1251932740626208,
      "grad_norm": 0.0,
      "learning_rate": 4.374275222265172e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2591
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 122.9375,
      "epoch": 0.125241592578276,
      "grad_norm": 15.6490201987675,
      "learning_rate": 4.374033629686896e-07,
      "loss": -0.0,
      "reward": 1.375,
      "reward_std": 0.38816189765930176,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2592
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.75,
      "epoch": 0.1252899110939312,
      "grad_norm": 0.0,
      "learning_rate": 4.3737920371086196e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2593
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.8125,
      "epoch": 0.1253382296095864,
      "grad_norm": 0.0,
      "learning_rate": 4.3735504445303437e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2594
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.15625,
      "epoch": 0.12538654812524158,
      "grad_norm": 0.0,
      "learning_rate": 4.373308851952068e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2595
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.0,
      "epoch": 0.12543486664089679,
      "grad_norm": 0.0,
      "learning_rate": 4.3730672593737914e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2596
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 120.3125,
      "epoch": 0.125483185156552,
      "grad_norm": 0.0,
      "learning_rate": 4.372825666795516e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2597
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.59375,
      "epoch": 0.1255315036722072,
      "grad_norm": 0.0,
      "learning_rate": 4.3725840742172397e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2598
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 122.90625,
      "epoch": 0.1255798221878624,
      "grad_norm": 0.0,
      "learning_rate": 4.372342481638964e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2599
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 123.4375,
      "epoch": 0.12562814070351758,
      "grad_norm": 0.0,
      "learning_rate": 4.372100889060688e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 2600
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.09375,
      "epoch": 0.12567645921917278,
      "grad_norm": 0.0,
      "learning_rate": 4.371859296482412e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2601
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.5,
      "epoch": 0.12572477773482799,
      "grad_norm": 0.0,
      "learning_rate": 4.3716177039041357e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2602
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.03125,
      "epoch": 0.1257730962504832,
      "grad_norm": 0.0,
      "learning_rate": 4.3713761113258593e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2603
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.25,
      "epoch": 0.1258214147661384,
      "grad_norm": 0.0,
      "learning_rate": 4.371134518747584e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2604
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.96875,
      "epoch": 0.12586973328179357,
      "grad_norm": 0.0,
      "learning_rate": 4.3708929261693076e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2605
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.09375,
      "epoch": 0.12591805179744878,
      "grad_norm": 0.0,
      "learning_rate": 4.370651333591032e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2606
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 122.8125,
      "epoch": 0.12596637031310398,
      "grad_norm": 0.0,
      "learning_rate": 4.370409741012756e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2607
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.75,
      "epoch": 0.12601468882875919,
      "grad_norm": 0.0,
      "learning_rate": 4.37016814843448e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2608
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.09375,
      "epoch": 0.1260630073444144,
      "grad_norm": 0.0,
      "learning_rate": 4.3699265558562036e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2609
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.71875,
      "epoch": 0.12611132586006957,
      "grad_norm": 0.0,
      "learning_rate": 4.3696849632779283e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2610
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 114.0625,
      "epoch": 0.12615964437572477,
      "grad_norm": 7.657501199484485,
      "learning_rate": 4.369443370699652e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2611
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.125,
      "epoch": 0.12620796289137998,
      "grad_norm": 0.0,
      "learning_rate": 4.369201778121376e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2612
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 113.21875,
      "epoch": 0.12625628140703518,
      "grad_norm": 107.42102741284893,
      "learning_rate": 4.3689601855430996e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.30103427171707153,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2613
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.40625,
      "epoch": 0.12630459992269039,
      "grad_norm": 0.0,
      "learning_rate": 4.368718592964824e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2614
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 124.71875,
      "epoch": 0.12635291843834556,
      "grad_norm": 0.0,
      "learning_rate": 4.368477000386548e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 2615
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 124.8125,
      "epoch": 0.12640123695400077,
      "grad_norm": 0.0,
      "learning_rate": 4.3682354078082715e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2616
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 127.78125,
      "epoch": 0.12644955546965597,
      "grad_norm": 34.15336567242997,
      "learning_rate": 4.367993815229996e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.21371951699256897,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2617
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.21875,
      "epoch": 0.12649787398531118,
      "grad_norm": 0.0,
      "learning_rate": 4.36775222265172e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2618
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 108.53125,
      "epoch": 0.12654619250096638,
      "grad_norm": 8.778857724620236,
      "learning_rate": 4.367510630073444e-07,
      "loss": -0.0,
      "reward": 2.09375,
      "reward_std": 0.2799758315086365,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2619
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 130.09375,
      "epoch": 0.12659451101662156,
      "grad_norm": 4.7075625638076986,
      "learning_rate": 4.367269037495168e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2620
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 116.5625,
      "epoch": 0.12664282953227676,
      "grad_norm": 5.675927948795552,
      "learning_rate": 4.367027444916892e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2621
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.53125,
      "epoch": 0.12669114804793197,
      "grad_norm": 0.0,
      "learning_rate": 4.366785852338616e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2622
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 117.375,
      "epoch": 0.12673946656358717,
      "grad_norm": 0.0,
      "learning_rate": 4.36654425976034e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2623
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 124.0625,
      "epoch": 0.12678778507924238,
      "grad_norm": 0.0,
      "learning_rate": 4.366302667182064e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2624
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.1875,
      "epoch": 0.12683610359489755,
      "grad_norm": 0.0,
      "learning_rate": 4.3660610746037877e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2625
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.78125,
      "epoch": 0.12688442211055276,
      "grad_norm": 0.0,
      "learning_rate": 4.365819482025512e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2626
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.03125,
      "epoch": 0.12693274062620796,
      "grad_norm": 0.0,
      "learning_rate": 4.365577889447236e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2627
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.9375,
      "epoch": 0.12698105914186317,
      "grad_norm": 0.0,
      "learning_rate": 4.36533629686896e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2628
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.15625,
      "epoch": 0.12702937765751837,
      "grad_norm": 0.0,
      "learning_rate": 4.3650947042906837e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2629
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.375,
      "epoch": 0.12707769617317355,
      "grad_norm": 0.0,
      "learning_rate": 4.3648531117124084e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2630
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.5625,
      "epoch": 0.12712601468882875,
      "grad_norm": 0.0,
      "learning_rate": 4.364611519134132e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2631
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 114.9375,
      "epoch": 0.12717433320448396,
      "grad_norm": 6.488018490504179,
      "learning_rate": 4.364369926555856e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2632
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 114.9375,
      "epoch": 0.12722265172013916,
      "grad_norm": 0.0,
      "learning_rate": 4.3641283339775797e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2633
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.5625,
      "epoch": 0.12727097023579437,
      "grad_norm": 0.0,
      "learning_rate": 4.3638867413993044e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23568853735923767,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2634
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.84375,
      "epoch": 0.12731928875144954,
      "grad_norm": 0.0,
      "learning_rate": 4.363645148821028e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2635
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.34375,
      "epoch": 0.12736760726710475,
      "grad_norm": 0.0,
      "learning_rate": 4.3634035562427516e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2636
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 105.6875,
      "epoch": 0.12741592578275995,
      "grad_norm": 5.020997767175584,
      "learning_rate": 4.363161963664476e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.3093337416648865,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2637
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.03125,
      "epoch": 0.12746424429841516,
      "grad_norm": 0.0,
      "learning_rate": 4.3629203710862e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2638
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.09375,
      "epoch": 0.12751256281407036,
      "grad_norm": 0.0,
      "learning_rate": 4.362678778507924e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2639
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 111.0,
      "epoch": 0.12756088132972554,
      "grad_norm": 10.663744096452264,
      "learning_rate": 4.362437185929648e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2640
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.46875,
      "epoch": 0.12760919984538074,
      "grad_norm": 0.0,
      "learning_rate": 4.3621955933513723e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2641
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 121.875,
      "epoch": 0.12765751836103595,
      "grad_norm": 0.0,
      "learning_rate": 4.361954000773096e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2642
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.625,
      "epoch": 0.12770583687669115,
      "grad_norm": 0.0,
      "learning_rate": 4.36171240819482e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 2643
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.8125,
      "epoch": 0.12775415539234636,
      "grad_norm": 0.0,
      "learning_rate": 4.361470815616544e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2644
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 107.0,
      "epoch": 0.12780247390800153,
      "grad_norm": 0.0,
      "learning_rate": 4.361229223038268e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 2645
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 109.65625,
      "epoch": 0.12785079242365674,
      "grad_norm": 0.0,
      "learning_rate": 4.360987630459992e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2646
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 98.8125,
      "epoch": 0.12789911093931194,
      "grad_norm": 6.270722056974805,
      "learning_rate": 4.360746037881716e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2647
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 123.8125,
      "epoch": 0.12794742945496715,
      "grad_norm": 0.0,
      "learning_rate": 4.36050444530344e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2648
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 104.0,
      "epoch": 0.12799574797062235,
      "grad_norm": 39.10029155981003,
      "learning_rate": 4.360262852725164e-07,
      "loss": 0.0,
      "reward": 2.1875,
      "reward_std": 0.3653584420681,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2649
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.625,
      "epoch": 0.12804406648627753,
      "grad_norm": 0.0,
      "learning_rate": 4.3600212601468884e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2650
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.09375,
      "epoch": 0.12809238500193273,
      "grad_norm": 0.0,
      "learning_rate": 4.359779667568612e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2651
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.1875,
      "epoch": 0.12814070351758794,
      "grad_norm": 0.0,
      "learning_rate": 4.359538074990336e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2652
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.53125,
      "epoch": 0.12818902203324314,
      "grad_norm": 0.0,
      "learning_rate": 4.35929648241206e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2653
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.375,
      "epoch": 0.12823734054889835,
      "grad_norm": 0.0,
      "learning_rate": 4.3590548898337845e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2654
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 110.59375,
      "epoch": 0.12828565906455353,
      "grad_norm": 0.0,
      "learning_rate": 4.358813297255508e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 2655
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.75,
      "epoch": 0.12833397758020873,
      "grad_norm": 0.0,
      "learning_rate": 4.3585717046772317e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2656
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.90625,
      "epoch": 0.12838229609586393,
      "grad_norm": 0.0,
      "learning_rate": 4.3583301120989563e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2657
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 123.84375,
      "epoch": 0.12843061461151914,
      "grad_norm": 7.335342110601018,
      "learning_rate": 4.35808851952068e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2658
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.78125,
      "epoch": 0.12847893312717434,
      "grad_norm": 0.0,
      "learning_rate": 4.357846926942404e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2659
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.8125,
      "epoch": 0.12852725164282952,
      "grad_norm": 0.0,
      "learning_rate": 4.357605334364128e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2660
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.40625,
      "epoch": 0.12857557015848473,
      "grad_norm": 0.0,
      "learning_rate": 4.3573637417858524e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2661
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.15625,
      "epoch": 0.12862388867413993,
      "grad_norm": 0.0,
      "learning_rate": 4.357122149207576e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2662
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.28125,
      "epoch": 0.12867220718979513,
      "grad_norm": 0.0,
      "learning_rate": 4.3568805566293e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2663
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.5,
      "epoch": 0.12872052570545034,
      "grad_norm": 0.0,
      "learning_rate": 4.356638964051024e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2664
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.28125,
      "epoch": 0.12876884422110552,
      "grad_norm": 0.0,
      "learning_rate": 4.3563973714727484e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2665
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 102.6875,
      "epoch": 0.12881716273676072,
      "grad_norm": 15.83115272916861,
      "learning_rate": 4.356155778894472e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.24280104041099548,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2666
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.28125,
      "epoch": 0.12886548125241593,
      "grad_norm": 0.0,
      "learning_rate": 4.355914186316196e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2667
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 111.3125,
      "epoch": 0.12891379976807113,
      "grad_norm": 21.357000326296912,
      "learning_rate": 4.35567259373792e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.28778618574142456,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2668
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.1875,
      "epoch": 0.12896211828372633,
      "grad_norm": 0.0,
      "learning_rate": 4.355431001159644e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2669
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.3125,
      "epoch": 0.1290104367993815,
      "grad_norm": 0.0,
      "learning_rate": 4.3551894085813685e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2670
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.8125,
      "epoch": 0.12905875531503672,
      "grad_norm": 0.0,
      "learning_rate": 4.354947816003092e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2671
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.125,
      "epoch": 0.12910707383069192,
      "grad_norm": 0.0,
      "learning_rate": 4.3547062234248163e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2672
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.75,
      "epoch": 0.12915539234634713,
      "grad_norm": 0.0,
      "learning_rate": 4.35446463084654e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2673
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.09375,
      "epoch": 0.12920371086200233,
      "grad_norm": 0.0,
      "learning_rate": 4.3542230382682645e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2674
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.71875,
      "epoch": 0.1292520293776575,
      "grad_norm": 0.0,
      "learning_rate": 4.353981445689988e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2675
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 101.25,
      "epoch": 0.1293003478933127,
      "grad_norm": 6.09571455688844,
      "learning_rate": 4.353739853111712e-07,
      "loss": -0.0,
      "reward": 1.671875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2676
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.25,
      "epoch": 0.12934866640896792,
      "grad_norm": 0.0,
      "learning_rate": 4.3534982605334364e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2677
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.0625,
      "epoch": 0.12939698492462312,
      "grad_norm": 0.0,
      "learning_rate": 4.35325666795516e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2678
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 104.09375,
      "epoch": 0.12944530344027833,
      "grad_norm": 14.858694873940607,
      "learning_rate": 4.353015075376884e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.2529362142086029,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2679
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.1875,
      "epoch": 0.1294936219559335,
      "grad_norm": 0.0,
      "learning_rate": 4.3527734827986083e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 2680
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.4375,
      "epoch": 0.1295419404715887,
      "grad_norm": 0.0,
      "learning_rate": 4.3525318902203324e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2681
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.9375,
      "epoch": 0.1295902589872439,
      "grad_norm": 0.0,
      "learning_rate": 4.352290297642056e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2682
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.875,
      "epoch": 0.12963857750289912,
      "grad_norm": 0.0,
      "learning_rate": 4.35204870506378e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2683
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.0,
      "epoch": 0.12968689601855432,
      "grad_norm": 0.0,
      "learning_rate": 4.3518071124855043e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2684
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.375,
      "epoch": 0.1297352145342095,
      "grad_norm": 0.0,
      "learning_rate": 4.3515655199072284e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2685
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.1875,
      "epoch": 0.1297835330498647,
      "grad_norm": 0.0,
      "learning_rate": 4.351323927328952e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2686
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 115.625,
      "epoch": 0.1298318515655199,
      "grad_norm": 19.924919540043295,
      "learning_rate": 4.351082334750676e-07,
      "loss": -0.0,
      "reward": 2.3125,
      "reward_std": 0.2961388826370239,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2687
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 94.46875,
      "epoch": 0.1298801700811751,
      "grad_norm": 19.738715691255404,
      "learning_rate": 4.3508407421724003e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.22017991542816162,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2688
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.09375,
      "epoch": 0.12992848859683032,
      "grad_norm": 0.0,
      "learning_rate": 4.350599149594124e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2689
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 109.40625,
      "epoch": 0.1299768071124855,
      "grad_norm": 0.0,
      "learning_rate": 4.3503575570158486e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2690
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.8125,
      "epoch": 0.1300251256281407,
      "grad_norm": 0.0,
      "learning_rate": 4.350115964437572e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2691
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.34375,
      "epoch": 0.1300734441437959,
      "grad_norm": 0.0,
      "learning_rate": 4.3498743718592963e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 2692
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 109.53125,
      "epoch": 0.1301217626594511,
      "grad_norm": 10.91738130435164,
      "learning_rate": 4.34963277928102e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2693
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.96875,
      "epoch": 0.1301700811751063,
      "grad_norm": 0.0,
      "learning_rate": 4.3493911867027446e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 2694
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.28125,
      "epoch": 0.1302183996907615,
      "grad_norm": 0.0,
      "learning_rate": 4.349149594124468e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2695
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.21875,
      "epoch": 0.1302667182064167,
      "grad_norm": 0.0,
      "learning_rate": 4.348908001546192e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2696
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.4375,
      "epoch": 0.1303150367220719,
      "grad_norm": 0.0,
      "learning_rate": 4.3486664089679165e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2697
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.90625,
      "epoch": 0.1303633552377271,
      "grad_norm": 0.0,
      "learning_rate": 4.34842481638964e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2698
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 122.0,
      "epoch": 0.1304116737533823,
      "grad_norm": 8.854005149809984,
      "learning_rate": 4.348183223811364e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.2766309082508087,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2699
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.59375,
      "epoch": 0.13045999226903748,
      "grad_norm": 0.0,
      "learning_rate": 4.3479416312330884e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2700
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.125,
      "epoch": 0.1305083107846927,
      "grad_norm": 0.0,
      "learning_rate": 4.3477000386548125e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2701
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 103.28125,
      "epoch": 0.1305566293003479,
      "grad_norm": 7.375179113785617,
      "learning_rate": 4.347458446076536e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 2702
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.6875,
      "epoch": 0.1306049478160031,
      "grad_norm": 0.0,
      "learning_rate": 4.34721685349826e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2703
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.5625,
      "epoch": 0.1306532663316583,
      "grad_norm": 36.897651358547776,
      "learning_rate": 4.3469752609199844e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.24280104041099548,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2704
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.75,
      "epoch": 0.13070158484731348,
      "grad_norm": 0.0,
      "learning_rate": 4.3467336683417085e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2705
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.90625,
      "epoch": 0.13074990336296868,
      "grad_norm": 0.0,
      "learning_rate": 4.346492075763432e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2706
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 107.4375,
      "epoch": 0.1307982218786239,
      "grad_norm": 9.114092464783322,
      "learning_rate": 4.3462504831851563e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2707
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.15625,
      "epoch": 0.1308465403942791,
      "grad_norm": 0.0,
      "learning_rate": 4.3460088906068804e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2708
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.96875,
      "epoch": 0.1308948589099343,
      "grad_norm": 0.0,
      "learning_rate": 4.345767298028604e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2709
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 98.3125,
      "epoch": 0.13094317742558947,
      "grad_norm": 13.341177914436075,
      "learning_rate": 4.3455257054503287e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.278043657541275,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2710
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.0,
      "epoch": 0.13099149594124468,
      "grad_norm": 0.0,
      "learning_rate": 4.3452841128720523e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 2711
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 107.4375,
      "epoch": 0.13103981445689988,
      "grad_norm": 9.954197650071205,
      "learning_rate": 4.3450425202937764e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2712
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.4375,
      "epoch": 0.1310881329725551,
      "grad_norm": 0.0,
      "learning_rate": 4.3448009277155e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 2713
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.03125,
      "epoch": 0.1311364514882103,
      "grad_norm": 0.0,
      "learning_rate": 4.3445593351372247e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2714
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.96875,
      "epoch": 0.13118477000386547,
      "grad_norm": 0.0,
      "learning_rate": 4.3443177425589483e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2715
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 102.09375,
      "epoch": 0.13123308851952067,
      "grad_norm": 0.0,
      "learning_rate": 4.3440761499806724e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 2716
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.40625,
      "epoch": 0.13128140703517588,
      "grad_norm": 0.0,
      "learning_rate": 4.3438345574023966e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2717
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.875,
      "epoch": 0.13132972555083108,
      "grad_norm": 0.0,
      "learning_rate": 4.34359296482412e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2718
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 100.78125,
      "epoch": 0.1313780440664863,
      "grad_norm": 5.5230778231323185,
      "learning_rate": 4.3433513722458443e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2719
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.4375,
      "epoch": 0.13142636258214146,
      "grad_norm": 0.0,
      "learning_rate": 4.3431097796675685e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2720
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 115.3125,
      "epoch": 0.13147468109779667,
      "grad_norm": 0.0,
      "learning_rate": 4.3428681870892926e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2721
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 108.65625,
      "epoch": 0.13152299961345187,
      "grad_norm": 0.0,
      "learning_rate": 4.342626594511016e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2722
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.25,
      "epoch": 0.13157131812910708,
      "grad_norm": 0.0,
      "learning_rate": 4.3423850019327403e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2723
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.46875,
      "epoch": 0.13161963664476228,
      "grad_norm": 0.0,
      "learning_rate": 4.3421434093544645e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2724
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.25,
      "epoch": 0.13166795516041746,
      "grad_norm": 0.0,
      "learning_rate": 4.3419018167761886e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2725
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.4375,
      "epoch": 0.13171627367607266,
      "grad_norm": 0.0,
      "learning_rate": 4.341660224197912e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2726
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.59375,
      "epoch": 0.13176459219172787,
      "grad_norm": 0.0,
      "learning_rate": 4.341418631619637e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2727
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.46875,
      "epoch": 0.13181291070738307,
      "grad_norm": 0.0,
      "learning_rate": 4.3411770390413605e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2728
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.0,
      "epoch": 0.13186122922303828,
      "grad_norm": 0.0,
      "learning_rate": 4.340935446463084e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2729
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.125,
      "epoch": 0.13190954773869346,
      "grad_norm": 0.0,
      "learning_rate": 4.340693853884809e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2730
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.9375,
      "epoch": 0.13195786625434866,
      "grad_norm": 0.0,
      "learning_rate": 4.3404522613065324e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2731
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.4375,
      "epoch": 0.13200618477000386,
      "grad_norm": 0.0,
      "learning_rate": 4.3402106687282565e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2732
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 98.9375,
      "epoch": 0.13205450328565907,
      "grad_norm": 16.9029204991276,
      "learning_rate": 4.33996907614998e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2733
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.6875,
      "epoch": 0.13210282180131427,
      "grad_norm": 0.0,
      "learning_rate": 4.339727483571705e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2734
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.25,
      "epoch": 0.13215114031696945,
      "grad_norm": 0.0,
      "learning_rate": 4.3394858909934284e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2735
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.78125,
      "epoch": 0.13219945883262466,
      "grad_norm": 0.0,
      "learning_rate": 4.3392442984151525e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2736
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.625,
      "epoch": 0.13224777734827986,
      "grad_norm": 0.0,
      "learning_rate": 4.3390027058368767e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2737
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.3125,
      "epoch": 0.13229609586393506,
      "grad_norm": 0.0,
      "learning_rate": 4.3387611132586e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2738
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.46875,
      "epoch": 0.13234441437959027,
      "grad_norm": 0.0,
      "learning_rate": 4.3385195206803244e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2739
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.15625,
      "epoch": 0.13239273289524545,
      "grad_norm": 0.0,
      "learning_rate": 4.3382779281020485e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 2740
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.28125,
      "epoch": 0.13244105141090065,
      "grad_norm": 0.0,
      "learning_rate": 4.3380363355237727e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2741
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 100.84375,
      "epoch": 0.13248936992655586,
      "grad_norm": 15.62954683745674,
      "learning_rate": 4.3377947429454963e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2742
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.03125,
      "epoch": 0.13253768844221106,
      "grad_norm": 0.0,
      "learning_rate": 4.3375531503672204e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2743
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.0625,
      "epoch": 0.13258600695786626,
      "grad_norm": 0.0,
      "learning_rate": 4.3373115577889445e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2744
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.125,
      "epoch": 0.13263432547352144,
      "grad_norm": 0.0,
      "learning_rate": 4.3370699652106687e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2745
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.90625,
      "epoch": 0.13268264398917665,
      "grad_norm": 0.0,
      "learning_rate": 4.3368283726323923e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2746
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.71875,
      "epoch": 0.13273096250483185,
      "grad_norm": 0.0,
      "learning_rate": 4.336586780054117e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2747
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.0625,
      "epoch": 0.13277928102048706,
      "grad_norm": 0.0,
      "learning_rate": 4.3363451874758406e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2748
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.0,
      "epoch": 0.13282759953614226,
      "grad_norm": 0.0,
      "learning_rate": 4.336103594897564e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2749
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 100.625,
      "epoch": 0.13287591805179744,
      "grad_norm": 13.974379960561794,
      "learning_rate": 4.335862002319289e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2750
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.5625,
      "epoch": 0.13292423656745264,
      "grad_norm": 0.0,
      "learning_rate": 4.3356204097410124e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2751
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.9375,
      "epoch": 0.13297255508310785,
      "grad_norm": 0.0,
      "learning_rate": 4.3353788171627366e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2752
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.1875,
      "epoch": 0.13302087359876305,
      "grad_norm": 0.0,
      "learning_rate": 4.33513722458446e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2753
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.125,
      "epoch": 0.13306919211441826,
      "grad_norm": 0.0,
      "learning_rate": 4.334895632006185e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2754
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.46875,
      "epoch": 0.13311751063007343,
      "grad_norm": 0.0,
      "learning_rate": 4.3346540394279085e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2755
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 108.4375,
      "epoch": 0.13316582914572864,
      "grad_norm": 0.0,
      "learning_rate": 4.3344124468496326e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2756
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.0,
      "epoch": 0.13321414766138384,
      "grad_norm": 0.0,
      "learning_rate": 4.3341708542713567e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2757
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.5625,
      "epoch": 0.13326246617703905,
      "grad_norm": 0.0,
      "learning_rate": 4.3339292616930803e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2758
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.65625,
      "epoch": 0.13331078469269425,
      "grad_norm": 0.0,
      "learning_rate": 4.3336876691148045e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2759
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.625,
      "epoch": 0.13335910320834943,
      "grad_norm": 0.0,
      "learning_rate": 4.3334460765365286e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2760
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.5625,
      "epoch": 0.13340742172400463,
      "grad_norm": 0.0,
      "learning_rate": 4.333204483958253e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2761
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.75,
      "epoch": 0.13345574023965984,
      "grad_norm": 0.0,
      "learning_rate": 4.3329628913799764e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2762
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 119.46875,
      "epoch": 0.13350405875531504,
      "grad_norm": 0.0,
      "learning_rate": 4.3327212988017005e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 2763
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.3125,
      "epoch": 0.13355237727097025,
      "grad_norm": 0.0,
      "learning_rate": 4.3324797062234246e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2764
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.875,
      "epoch": 0.13360069578662542,
      "grad_norm": 0.0,
      "learning_rate": 4.332238113645149e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2765
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.8125,
      "epoch": 0.13364901430228063,
      "grad_norm": 0.0,
      "learning_rate": 4.3319965210668724e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2766
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 114.46875,
      "epoch": 0.13369733281793583,
      "grad_norm": 0.0,
      "learning_rate": 4.331754928488597e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2767
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.78125,
      "epoch": 0.13374565133359104,
      "grad_norm": 0.0,
      "learning_rate": 4.3315133359103206e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 2768
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.1875,
      "epoch": 0.13379396984924624,
      "grad_norm": 0.0,
      "learning_rate": 4.331271743332044e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2769
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.125,
      "epoch": 0.13384228836490142,
      "grad_norm": 0.0,
      "learning_rate": 4.331030150753769e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2770
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.75,
      "epoch": 0.13389060688055662,
      "grad_norm": 0.0,
      "learning_rate": 4.3307885581754925e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2771
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.875,
      "epoch": 0.13393892539621183,
      "grad_norm": 0.0,
      "learning_rate": 4.3305469655972167e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2772
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.125,
      "epoch": 0.13398724391186703,
      "grad_norm": 0.0,
      "learning_rate": 4.3303053730189403e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2773
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.6875,
      "epoch": 0.13403556242752224,
      "grad_norm": 0.0,
      "learning_rate": 4.330063780440665e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2774
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.34375,
      "epoch": 0.1340838809431774,
      "grad_norm": 0.0,
      "learning_rate": 4.3298221878623885e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 2775
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.75,
      "epoch": 0.13413219945883262,
      "grad_norm": 0.0,
      "learning_rate": 4.3295805952841127e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2776
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.0625,
      "epoch": 0.13418051797448782,
      "grad_norm": 0.0,
      "learning_rate": 4.329339002705837e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2777
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.1875,
      "epoch": 0.13422883649014303,
      "grad_norm": 0.0,
      "learning_rate": 4.329097410127561e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2778
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.875,
      "epoch": 0.13427715500579823,
      "grad_norm": 9.886809309622299,
      "learning_rate": 4.3288558175492846e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2779
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.6875,
      "epoch": 0.1343254735214534,
      "grad_norm": 0.0,
      "learning_rate": 4.3286142249710087e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2780
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.46875,
      "epoch": 0.1343737920371086,
      "grad_norm": 0.0,
      "learning_rate": 4.328372632392733e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2781
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.5625,
      "epoch": 0.13442211055276382,
      "grad_norm": 0.0,
      "learning_rate": 4.3281310398144564e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2782
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.6875,
      "epoch": 0.13447042906841902,
      "grad_norm": 0.0,
      "learning_rate": 4.3278894472361806e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2783
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.96875,
      "epoch": 0.13451874758407423,
      "grad_norm": 0.0,
      "learning_rate": 4.3276478546579047e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2784
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.84375,
      "epoch": 0.1345670660997294,
      "grad_norm": 0.0,
      "learning_rate": 4.327406262079629e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2785
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 120.3125,
      "epoch": 0.1346153846153846,
      "grad_norm": 7.533207488000892,
      "learning_rate": 4.3271646695013524e-07,
      "loss": 0.0,
      "reward": 1.453125,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 2786
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 116.125,
      "epoch": 0.1346637031310398,
      "grad_norm": 4.582045966042472,
      "learning_rate": 4.326923076923077e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2787
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 115.1875,
      "epoch": 0.13471202164669502,
      "grad_norm": 19.664431797075025,
      "learning_rate": 4.3266814843448007e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.2461063265800476,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2788
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.53125,
      "epoch": 0.13476034016235022,
      "grad_norm": 0.0,
      "learning_rate": 4.3264398917665243e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2789
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.625,
      "epoch": 0.1348086586780054,
      "grad_norm": 0.0,
      "learning_rate": 4.326198299188249e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2790
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.8125,
      "epoch": 0.1348569771936606,
      "grad_norm": 0.0,
      "learning_rate": 4.3259567066099726e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2791
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.25,
      "epoch": 0.1349052957093158,
      "grad_norm": 0.0,
      "learning_rate": 4.325715114031697e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2792
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.96875,
      "epoch": 0.134953614224971,
      "grad_norm": 0.0,
      "learning_rate": 4.3254735214534203e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2793
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.375,
      "epoch": 0.13500193274062622,
      "grad_norm": 0.0,
      "learning_rate": 4.325231928875145e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2794
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.625,
      "epoch": 0.1350502512562814,
      "grad_norm": 0.0,
      "learning_rate": 4.3249903362968686e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2795
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 103.40625,
      "epoch": 0.1350985697719366,
      "grad_norm": 5.620804199875888,
      "learning_rate": 4.324748743718593e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2796
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.40625,
      "epoch": 0.1351468882875918,
      "grad_norm": 0.0,
      "learning_rate": 4.324507151140317e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 2797
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.6875,
      "epoch": 0.135195206803247,
      "grad_norm": 0.0,
      "learning_rate": 4.324265558562041e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2798
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 99.5625,
      "epoch": 0.1352435253189022,
      "grad_norm": 8.216073363282208,
      "learning_rate": 4.3240239659837646e-07,
      "loss": -0.0,
      "reward": 1.625,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2799
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.4375,
      "epoch": 0.1352918438345574,
      "grad_norm": 0.0,
      "learning_rate": 4.323782373405489e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2800
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.15625,
      "epoch": 0.1353401623502126,
      "grad_norm": 0.0,
      "learning_rate": 4.323540780827213e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2801
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.25,
      "epoch": 0.1353884808658678,
      "grad_norm": 0.0,
      "learning_rate": 4.3232991882489365e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2802
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.8125,
      "epoch": 0.135436799381523,
      "grad_norm": 0.0,
      "learning_rate": 4.3230575956706606e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2803
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.90625,
      "epoch": 0.1354851178971782,
      "grad_norm": 0.0,
      "learning_rate": 4.322816003092385e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2804
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.5,
      "epoch": 0.13553343641283339,
      "grad_norm": 4.941636152686688,
      "learning_rate": 4.322574410514109e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2805
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.96875,
      "epoch": 0.1355817549284886,
      "grad_norm": 0.0,
      "learning_rate": 4.3223328179358325e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2806
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.15625,
      "epoch": 0.1356300734441438,
      "grad_norm": 0.0,
      "learning_rate": 4.322091225357557e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2807
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.8125,
      "epoch": 0.135678391959799,
      "grad_norm": 0.0,
      "learning_rate": 4.321849632779281e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2808
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.71875,
      "epoch": 0.1357267104754542,
      "grad_norm": 0.0,
      "learning_rate": 4.3216080402010044e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2809
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.65625,
      "epoch": 0.13577502899110938,
      "grad_norm": 0.0,
      "learning_rate": 4.321366447622729e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2810
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.625,
      "epoch": 0.13582334750676459,
      "grad_norm": 0.0,
      "learning_rate": 4.3211248550444527e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2811
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.0625,
      "epoch": 0.1358716660224198,
      "grad_norm": 0.0,
      "learning_rate": 4.320883262466177e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2812
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.4375,
      "epoch": 0.135919984538075,
      "grad_norm": 0.0,
      "learning_rate": 4.3206416698879004e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2813
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.15625,
      "epoch": 0.1359683030537302,
      "grad_norm": 0.0,
      "learning_rate": 4.320400077309625e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2814
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.65625,
      "epoch": 0.13601662156938538,
      "grad_norm": 0.0,
      "learning_rate": 4.3201584847313487e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2815
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.5,
      "epoch": 0.13606494008504058,
      "grad_norm": 0.0,
      "learning_rate": 4.319916892153073e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 2816
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.125,
      "epoch": 0.13611325860069579,
      "grad_norm": 0.0,
      "learning_rate": 4.319675299574797e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2817
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.96875,
      "epoch": 0.136161577116351,
      "grad_norm": 0.0,
      "learning_rate": 4.319433706996521e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2818
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.375,
      "epoch": 0.1362098956320062,
      "grad_norm": 0.0,
      "learning_rate": 4.3191921144182447e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2819
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 92.90625,
      "epoch": 0.13625821414766137,
      "grad_norm": 0.0,
      "learning_rate": 4.318950521839969e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 2820
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.625,
      "epoch": 0.13630653266331658,
      "grad_norm": 0.0,
      "learning_rate": 4.318708929261693e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2821
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 67.46875,
      "epoch": 0.13635485117897178,
      "grad_norm": 43.19043194800258,
      "learning_rate": 4.3184673366834166e-07,
      "loss": -0.0,
      "reward": 2.3125,
      "reward_std": 0.3029785752296448,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2822
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.8125,
      "epoch": 0.13640316969462699,
      "grad_norm": 0.0,
      "learning_rate": 4.3182257441051407e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2823
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.125,
      "epoch": 0.1364514882102822,
      "grad_norm": 0.0,
      "learning_rate": 4.317984151526865e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2824
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.84375,
      "epoch": 0.13649980672593737,
      "grad_norm": 0.0,
      "learning_rate": 4.317742558948589e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2825
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 95.5,
      "epoch": 0.13654812524159257,
      "grad_norm": 0.0,
      "learning_rate": 4.3175009663703126e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 2826
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.9375,
      "epoch": 0.13659644375724778,
      "grad_norm": 0.0,
      "learning_rate": 4.3172593737920373e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2827
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.875,
      "epoch": 0.13664476227290298,
      "grad_norm": 0.0,
      "learning_rate": 4.317017781213761e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2828
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.40625,
      "epoch": 0.13669308078855819,
      "grad_norm": 0.0,
      "learning_rate": 4.316776188635485e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2829
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.75,
      "epoch": 0.13674139930421336,
      "grad_norm": 0.0,
      "learning_rate": 4.316534596057209e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2830
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.21875,
      "epoch": 0.13678971781986857,
      "grad_norm": 0.0,
      "learning_rate": 4.316293003478933e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2831
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.6875,
      "epoch": 0.13683803633552377,
      "grad_norm": 0.0,
      "learning_rate": 4.316051410900657e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2832
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.625,
      "epoch": 0.13688635485117898,
      "grad_norm": 0.0,
      "learning_rate": 4.3158098183223805e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2833
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.90625,
      "epoch": 0.13693467336683418,
      "grad_norm": 0.0,
      "learning_rate": 4.315568225744105e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2834
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.3125,
      "epoch": 0.13698299188248936,
      "grad_norm": 9.303735031513527,
      "learning_rate": 4.315326633165829e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2835
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.40625,
      "epoch": 0.13703131039814456,
      "grad_norm": 14.019914527529643,
      "learning_rate": 4.315085040587553e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2836
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.875,
      "epoch": 0.13707962891379977,
      "grad_norm": 0.0,
      "learning_rate": 4.314843448009277e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 2837
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.71875,
      "epoch": 0.13712794742945497,
      "grad_norm": 0.0,
      "learning_rate": 4.314601855431001e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2838
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.3125,
      "epoch": 0.13717626594511018,
      "grad_norm": 0.0,
      "learning_rate": 4.314360262852725e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2839
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.5625,
      "epoch": 0.13722458446076535,
      "grad_norm": 14.89382522425207,
      "learning_rate": 4.3141186702744495e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2840
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.78125,
      "epoch": 0.13727290297642056,
      "grad_norm": 0.0,
      "learning_rate": 4.313877077696173e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2841
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.625,
      "epoch": 0.13732122149207576,
      "grad_norm": 0.0,
      "learning_rate": 4.3136354851178967e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2842
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.90625,
      "epoch": 0.13736954000773097,
      "grad_norm": 0.0,
      "learning_rate": 4.313393892539621e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2843
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.96875,
      "epoch": 0.13741785852338617,
      "grad_norm": 0.0,
      "learning_rate": 4.313152299961345e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2844
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 75.78125,
      "epoch": 0.13746617703904135,
      "grad_norm": 13.782178266852394,
      "learning_rate": 4.312910707383069e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2845
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.625,
      "epoch": 0.13751449555469655,
      "grad_norm": 0.0,
      "learning_rate": 4.3126691148047927e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2846
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.53125,
      "epoch": 0.13756281407035176,
      "grad_norm": 0.0,
      "learning_rate": 4.3124275222265174e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2847
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.3125,
      "epoch": 0.13761113258600696,
      "grad_norm": 0.0,
      "learning_rate": 4.312185929648241e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2848
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.4375,
      "epoch": 0.13765945110166217,
      "grad_norm": 0.0,
      "learning_rate": 4.311944337069965e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2849
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.25,
      "epoch": 0.13770776961731734,
      "grad_norm": 0.0,
      "learning_rate": 4.311702744491689e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2850
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.46875,
      "epoch": 0.13775608813297255,
      "grad_norm": 0.0,
      "learning_rate": 4.311461151913413e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2851
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.84375,
      "epoch": 0.13780440664862775,
      "grad_norm": 0.0,
      "learning_rate": 4.311219559335137e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2852
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.875,
      "epoch": 0.13785272516428296,
      "grad_norm": 0.0,
      "learning_rate": 4.3109779667568606e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2853
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.03125,
      "epoch": 0.13790104367993816,
      "grad_norm": 0.0,
      "learning_rate": 4.310736374178585e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2854
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.09375,
      "epoch": 0.13794936219559334,
      "grad_norm": 0.0,
      "learning_rate": 4.310494781600309e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2855
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 82.3125,
      "epoch": 0.13799768071124854,
      "grad_norm": 20.092732302193507,
      "learning_rate": 4.310253189022033e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2856
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.65625,
      "epoch": 0.13804599922690375,
      "grad_norm": 0.0,
      "learning_rate": 4.310011596443757e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2857
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.59375,
      "epoch": 0.13809431774255895,
      "grad_norm": 0.0,
      "learning_rate": 4.309770003865481e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2858
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.28125,
      "epoch": 0.13814263625821416,
      "grad_norm": 0.0,
      "learning_rate": 4.309528411287205e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2859
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.375,
      "epoch": 0.13819095477386933,
      "grad_norm": 10.94876736869538,
      "learning_rate": 4.3092868187089295e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2860
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 96.8125,
      "epoch": 0.13823927328952454,
      "grad_norm": 7.903299942360239,
      "learning_rate": 4.309045226130653e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2861
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.53125,
      "epoch": 0.13828759180517974,
      "grad_norm": 0.0,
      "learning_rate": 4.308803633552377e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2862
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.03125,
      "epoch": 0.13833591032083495,
      "grad_norm": 0.0,
      "learning_rate": 4.308562040974101e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2863
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 78.84375,
      "epoch": 0.13838422883649015,
      "grad_norm": 26.13498275592019,
      "learning_rate": 4.308320448395825e-07,
      "loss": 0.0,
      "reward": 1.125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.09375,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 2864
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.46875,
      "epoch": 0.13843254735214533,
      "grad_norm": 0.0,
      "learning_rate": 4.308078855817549e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2865
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.5625,
      "epoch": 0.13848086586780053,
      "grad_norm": 0.0,
      "learning_rate": 4.307837263239273e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2866
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 91.125,
      "epoch": 0.13852918438345574,
      "grad_norm": 13.93615902722337,
      "learning_rate": 4.3075956706609974e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2867
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.96875,
      "epoch": 0.13857750289911094,
      "grad_norm": 0.0,
      "learning_rate": 4.307354078082721e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2868
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 98.5625,
      "epoch": 0.13862582141476615,
      "grad_norm": 59.84620175393293,
      "learning_rate": 4.307112485504445e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2869
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.75,
      "epoch": 0.13867413993042133,
      "grad_norm": 0.0,
      "learning_rate": 4.3068708929261693e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2870
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 92.65625,
      "epoch": 0.13872245844607653,
      "grad_norm": 8.585348934880663,
      "learning_rate": 4.306629300347893e-07,
      "loss": 0.0,
      "reward": 1.28125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2871
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.21875,
      "epoch": 0.13877077696173173,
      "grad_norm": 0.0,
      "learning_rate": 4.306387707769617e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2872
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 92.375,
      "epoch": 0.13881909547738694,
      "grad_norm": 10.16573149521658,
      "learning_rate": 4.3061461151913407e-07,
      "loss": -0.0,
      "reward": 1.171875,
      "reward_std": 0.188242569565773,
      "rewards/accuracy_reward": 0.125,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 2873
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.03125,
      "epoch": 0.13886741399304214,
      "grad_norm": 0.0,
      "learning_rate": 4.3059045226130653e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2874
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.1875,
      "epoch": 0.13891573250869732,
      "grad_norm": 0.0,
      "learning_rate": 4.305662930034789e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2561880052089691,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2875
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.84375,
      "epoch": 0.13896405102435253,
      "grad_norm": 0.0,
      "learning_rate": 4.305421337456513e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2876
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.40625,
      "epoch": 0.13901236954000773,
      "grad_norm": 38.176656023074514,
      "learning_rate": 4.305179744878237e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.20633062720298767,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2877
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.15625,
      "epoch": 0.13906068805566293,
      "grad_norm": 0.0,
      "learning_rate": 4.3049381522999613e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2878
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.375,
      "epoch": 0.13910900657131814,
      "grad_norm": 0.0,
      "learning_rate": 4.304696559721685e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2879
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.96875,
      "epoch": 0.13915732508697332,
      "grad_norm": 0.0,
      "learning_rate": 4.3044549671434096e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2880
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 67.875,
      "epoch": 0.13920564360262852,
      "grad_norm": 11.750933775670354,
      "learning_rate": 4.304213374565133e-07,
      "loss": -0.0,
      "reward": 2.328125,
      "reward_std": 0.2993341088294983,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2881
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.03125,
      "epoch": 0.13925396211828373,
      "grad_norm": 0.0,
      "learning_rate": 4.303971781986857e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2882
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.15625,
      "epoch": 0.13930228063393893,
      "grad_norm": 0.0,
      "learning_rate": 4.303730189408581e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2883
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 70.53125,
      "epoch": 0.13935059914959413,
      "grad_norm": 6.281734795966004,
      "learning_rate": 4.303488596830305e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2884
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.9375,
      "epoch": 0.1393989176652493,
      "grad_norm": 0.0,
      "learning_rate": 4.303247004252029e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2885
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.75,
      "epoch": 0.13944723618090452,
      "grad_norm": 34.81202810962672,
      "learning_rate": 4.303005411673753e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.2982703447341919,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2886
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.28125,
      "epoch": 0.13949555469655972,
      "grad_norm": 0.0,
      "learning_rate": 4.3027638190954775e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2887
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.40625,
      "epoch": 0.13954387321221493,
      "grad_norm": 0.0,
      "learning_rate": 4.302522226517201e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2888
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.0625,
      "epoch": 0.13959219172787013,
      "grad_norm": 0.0,
      "learning_rate": 4.302280633938925e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 2889
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.09375,
      "epoch": 0.1396405102435253,
      "grad_norm": 0.0,
      "learning_rate": 4.3020390413606494e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 2890
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 85.40625,
      "epoch": 0.1396888287591805,
      "grad_norm": 0.0,
      "learning_rate": 4.3017974487823735e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 2891
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.90625,
      "epoch": 0.13973714727483572,
      "grad_norm": 0.0,
      "learning_rate": 4.301555856204097e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2892
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.0,
      "epoch": 0.13978546579049092,
      "grad_norm": 0.0,
      "learning_rate": 4.301314263625821e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2893
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.375,
      "epoch": 0.13983378430614612,
      "grad_norm": 0.0,
      "learning_rate": 4.3010726710475454e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2894
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.65625,
      "epoch": 0.1398821028218013,
      "grad_norm": 0.0,
      "learning_rate": 4.300831078469269e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2895
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.96875,
      "epoch": 0.1399304213374565,
      "grad_norm": 0.0,
      "learning_rate": 4.300589485890993e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2896
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.5,
      "epoch": 0.1399787398531117,
      "grad_norm": 0.0,
      "learning_rate": 4.3003478933127173e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2897
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.65625,
      "epoch": 0.14002705836876692,
      "grad_norm": 0.0,
      "learning_rate": 4.3001063007344414e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2898
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.53125,
      "epoch": 0.14007537688442212,
      "grad_norm": 0.0,
      "learning_rate": 4.299864708156165e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2899
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.40625,
      "epoch": 0.1401236954000773,
      "grad_norm": 0.0,
      "learning_rate": 4.2996231155778897e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2900
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.28125,
      "epoch": 0.1401720139157325,
      "grad_norm": 26.71356361585304,
      "learning_rate": 4.2993815229996133e-07,
      "loss": -0.0,
      "reward": 2.140625,
      "reward_std": 0.3266732692718506,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2901
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.15625,
      "epoch": 0.1402203324313877,
      "grad_norm": 0.0,
      "learning_rate": 4.299139930421337e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2902
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 90.875,
      "epoch": 0.1402686509470429,
      "grad_norm": 0.0,
      "learning_rate": 4.298898337843061e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 2903
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.0625,
      "epoch": 0.14031696946269812,
      "grad_norm": 0.0,
      "learning_rate": 4.298656745264785e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 2904
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.375,
      "epoch": 0.1403652879783533,
      "grad_norm": 0.0,
      "learning_rate": 4.2984151526865093e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2905
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 83.53125,
      "epoch": 0.1404136064940085,
      "grad_norm": 72.64582604301884,
      "learning_rate": 4.298173560108233e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.43064606189727783,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2906
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 68.65625,
      "epoch": 0.1404619250096637,
      "grad_norm": 26.781349371155788,
      "learning_rate": 4.2979319675299576e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.2951723635196686,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2907
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.9375,
      "epoch": 0.1405102435253189,
      "grad_norm": 0.0,
      "learning_rate": 4.297690374951681e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 2908
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.625,
      "epoch": 0.1405585620409741,
      "grad_norm": 0.0,
      "learning_rate": 4.2974487823734053e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2909
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.65625,
      "epoch": 0.1406068805566293,
      "grad_norm": 75.80639405855835,
      "learning_rate": 4.2972071897951295e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2910
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.46875,
      "epoch": 0.1406551990722845,
      "grad_norm": 0.0,
      "learning_rate": 4.2969655972168536e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2911
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.8125,
      "epoch": 0.1407035175879397,
      "grad_norm": 0.0,
      "learning_rate": 4.296724004638577e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2912
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.21875,
      "epoch": 0.1407518361035949,
      "grad_norm": 0.0,
      "learning_rate": 4.296482412060301e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2913
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.4375,
      "epoch": 0.1408001546192501,
      "grad_norm": 0.0,
      "learning_rate": 4.2962408194820255e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2914
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.40625,
      "epoch": 0.14084847313490528,
      "grad_norm": 0.0,
      "learning_rate": 4.295999226903749e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2915
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.5625,
      "epoch": 0.1408967916505605,
      "grad_norm": 0.0,
      "learning_rate": 4.295757634325473e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2916
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.09375,
      "epoch": 0.1409451101662157,
      "grad_norm": 0.0,
      "learning_rate": 4.2955160417471974e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 2917
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.5,
      "epoch": 0.1409934286818709,
      "grad_norm": 0.0,
      "learning_rate": 4.2952744491689215e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2918
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.46875,
      "epoch": 0.1410417471975261,
      "grad_norm": 60.83555666732339,
      "learning_rate": 4.295032856590645e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.2573546767234802,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2919
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.78125,
      "epoch": 0.14109006571318128,
      "grad_norm": 22.09241615475881,
      "learning_rate": 4.29479126401237e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2920
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.625,
      "epoch": 0.14113838422883648,
      "grad_norm": 0.0,
      "learning_rate": 4.2945496714340934e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2921
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.71875,
      "epoch": 0.1411867027444917,
      "grad_norm": 0.0,
      "learning_rate": 4.2943080788558175e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2922
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.125,
      "epoch": 0.1412350212601469,
      "grad_norm": 0.0,
      "learning_rate": 4.294066486277541e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2923
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.78125,
      "epoch": 0.1412833397758021,
      "grad_norm": 0.0,
      "learning_rate": 4.293824893699265e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2924
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.71875,
      "epoch": 0.14133165829145727,
      "grad_norm": 0.0,
      "learning_rate": 4.2935833011209894e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2925
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.71875,
      "epoch": 0.14137997680711248,
      "grad_norm": 0.0,
      "learning_rate": 4.293341708542713e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2926
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.90625,
      "epoch": 0.14142829532276768,
      "grad_norm": 27.94812563864133,
      "learning_rate": 4.2931001159644377e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2927
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.34375,
      "epoch": 0.1414766138384229,
      "grad_norm": 0.0,
      "learning_rate": 4.2928585233861613e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2928
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.25,
      "epoch": 0.1415249323540781,
      "grad_norm": 0.0,
      "learning_rate": 4.2926169308078854e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2929
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 87.40625,
      "epoch": 0.14157325086973327,
      "grad_norm": 16.648972800698917,
      "learning_rate": 4.2923753382296095e-07,
      "loss": -0.0,
      "reward": 1.546875,
      "reward_std": 0.188242569565773,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2930
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 80.78125,
      "epoch": 0.14162156938538847,
      "grad_norm": 16.791942440301693,
      "learning_rate": 4.2921337456513337e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2931
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.0625,
      "epoch": 0.14166988790104368,
      "grad_norm": 0.0,
      "learning_rate": 4.2918921530730573e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2932
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.0,
      "epoch": 0.14171820641669888,
      "grad_norm": 0.0,
      "learning_rate": 4.291650560494781e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2933
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.0625,
      "epoch": 0.1417665249323541,
      "grad_norm": 0.0,
      "learning_rate": 4.2914089679165056e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2934
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.625,
      "epoch": 0.14181484344800926,
      "grad_norm": 0.0,
      "learning_rate": 4.291167375338229e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2935
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.71875,
      "epoch": 0.14186316196366447,
      "grad_norm": 0.0,
      "learning_rate": 4.2909257827599533e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2936
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.71875,
      "epoch": 0.14191148047931967,
      "grad_norm": 0.0,
      "learning_rate": 4.2906841901816774e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 2937
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.65625,
      "epoch": 0.14195979899497488,
      "grad_norm": 0.0,
      "learning_rate": 4.2904425976034016e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2938
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.59375,
      "epoch": 0.14200811751063008,
      "grad_norm": 0.0,
      "learning_rate": 4.290201005025125e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2939
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.0625,
      "epoch": 0.14205643602628526,
      "grad_norm": 0.0,
      "learning_rate": 4.28995941244685e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2940
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.90625,
      "epoch": 0.14210475454194046,
      "grad_norm": 0.0,
      "learning_rate": 4.2897178198685735e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2941
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.03125,
      "epoch": 0.14215307305759567,
      "grad_norm": 0.0,
      "learning_rate": 4.2894762272902976e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2942
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 83.1875,
      "epoch": 0.14220139157325087,
      "grad_norm": 22.57769763067431,
      "learning_rate": 4.289234634712021e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.3114554286003113,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2943
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.53125,
      "epoch": 0.14224971008890608,
      "grad_norm": 0.0,
      "learning_rate": 4.2889930421337453e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2944
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 93.53125,
      "epoch": 0.14229802860456126,
      "grad_norm": 15.71450663155099,
      "learning_rate": 4.2887514495554695e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.2852528691291809,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2945
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 94.03125,
      "epoch": 0.14234634712021646,
      "grad_norm": 11.093121260314653,
      "learning_rate": 4.288509856977193e-07,
      "loss": -0.0,
      "reward": 2.359375,
      "reward_std": 0.27883464097976685,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2946
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.71875,
      "epoch": 0.14239466563587166,
      "grad_norm": 0.0,
      "learning_rate": 4.288268264398918e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 2947
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.3125,
      "epoch": 0.14244298415152687,
      "grad_norm": 10.759019239332476,
      "learning_rate": 4.2880266718206414e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2948
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.78125,
      "epoch": 0.14249130266718207,
      "grad_norm": 0.0,
      "learning_rate": 4.2877850792423655e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2949
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.46875,
      "epoch": 0.14253962118283725,
      "grad_norm": 0.0,
      "learning_rate": 4.2875434866640896e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2950
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 93.875,
      "epoch": 0.14258793969849246,
      "grad_norm": 7.99080353647406,
      "learning_rate": 4.287301894085814e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.23421898484230042,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2951
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.71875,
      "epoch": 0.14263625821414766,
      "grad_norm": 0.0,
      "learning_rate": 4.2870603015075374e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2952
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.625,
      "epoch": 0.14268457672980286,
      "grad_norm": 0.0,
      "learning_rate": 4.286818708929261e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2953
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.5625,
      "epoch": 0.14273289524545807,
      "grad_norm": 0.0,
      "learning_rate": 4.2865771163509856e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2954
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.46875,
      "epoch": 0.14278121376111325,
      "grad_norm": 0.0,
      "learning_rate": 4.286335523772709e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2955
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.4375,
      "epoch": 0.14282953227676845,
      "grad_norm": 0.0,
      "learning_rate": 4.2860939311944334e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2956
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.125,
      "epoch": 0.14287785079242366,
      "grad_norm": 0.0,
      "learning_rate": 4.2858523386161575e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 2957
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.03125,
      "epoch": 0.14292616930807886,
      "grad_norm": 0.0,
      "learning_rate": 4.2856107460378817e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2958
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.5,
      "epoch": 0.14297448782373406,
      "grad_norm": 0.0,
      "learning_rate": 4.285369153459605e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2959
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.1875,
      "epoch": 0.14302280633938924,
      "grad_norm": 0.0,
      "learning_rate": 4.28512756088133e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2960
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.5625,
      "epoch": 0.14307112485504445,
      "grad_norm": 0.0,
      "learning_rate": 4.2848859683030535e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2961
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.6875,
      "epoch": 0.14311944337069965,
      "grad_norm": 0.0,
      "learning_rate": 4.2846443757247777e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2962
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.625,
      "epoch": 0.14316776188635486,
      "grad_norm": 11.489103098395265,
      "learning_rate": 4.2844027831465013e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.3072218894958496,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 2963
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.09375,
      "epoch": 0.14321608040201006,
      "grad_norm": 18.980384329201115,
      "learning_rate": 4.2841611905682254e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2964
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.3125,
      "epoch": 0.14326439891766524,
      "grad_norm": 7.841395334101728,
      "learning_rate": 4.2839195979899495e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2965
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.34375,
      "epoch": 0.14331271743332044,
      "grad_norm": 0.0,
      "learning_rate": 4.283678005411673e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2966
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.875,
      "epoch": 0.14336103594897565,
      "grad_norm": 0.0,
      "learning_rate": 4.283436412833398e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2967
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 80.75,
      "epoch": 0.14340935446463085,
      "grad_norm": 83.9761755191548,
      "learning_rate": 4.2831948202551214e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 2968
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.1875,
      "epoch": 0.14345767298028606,
      "grad_norm": 0.0,
      "learning_rate": 4.2829532276768456e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2969
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.34375,
      "epoch": 0.14350599149594123,
      "grad_norm": 0.0,
      "learning_rate": 4.2827116350985697e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 2970
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.875,
      "epoch": 0.14355431001159644,
      "grad_norm": 0.0,
      "learning_rate": 4.282470042520294e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2971
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.9375,
      "epoch": 0.14360262852725164,
      "grad_norm": 0.0,
      "learning_rate": 4.2822284499420174e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2972
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.3125,
      "epoch": 0.14365094704290685,
      "grad_norm": 0.0,
      "learning_rate": 4.2819868573637416e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 2973
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 84.46875,
      "epoch": 0.14369926555856205,
      "grad_norm": 17.880506340589374,
      "learning_rate": 4.2817452647854657e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 2974
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.84375,
      "epoch": 0.14374758407421723,
      "grad_norm": 0.0,
      "learning_rate": 4.2815036722071893e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2975
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 70.75,
      "epoch": 0.14379590258987243,
      "grad_norm": 9.707799514981403,
      "learning_rate": 4.2812620796289135e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.26726123690605164,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2976
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.4375,
      "epoch": 0.14384422110552764,
      "grad_norm": 0.0,
      "learning_rate": 4.2810204870506376e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 2977
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.78125,
      "epoch": 0.14389253962118284,
      "grad_norm": 0.0,
      "learning_rate": 4.2807788944723617e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2978
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.71875,
      "epoch": 0.14394085813683805,
      "grad_norm": 0.0,
      "learning_rate": 4.2805373018940853e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2979
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 76.875,
      "epoch": 0.14398917665249322,
      "grad_norm": 7.952675444579415,
      "learning_rate": 4.28029570931581e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 2980
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.59375,
      "epoch": 0.14403749516814843,
      "grad_norm": 8.454447556475865,
      "learning_rate": 4.2800541167375336e-07,
      "loss": -0.0,
      "reward": 2.40625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 2981
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.21875,
      "epoch": 0.14408581368380363,
      "grad_norm": 0.0,
      "learning_rate": 4.279812524159258e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 2982
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.125,
      "epoch": 0.14413413219945884,
      "grad_norm": 0.0,
      "learning_rate": 4.2795709315809814e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2983
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.1875,
      "epoch": 0.14418245071511404,
      "grad_norm": 0.0,
      "learning_rate": 4.279329339002706e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 2984
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.8125,
      "epoch": 0.14423076923076922,
      "grad_norm": 0.0,
      "learning_rate": 4.2790877464244296e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2985
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.0625,
      "epoch": 0.14427908774642442,
      "grad_norm": 0.0,
      "learning_rate": 4.278846153846153e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2986
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.0625,
      "epoch": 0.14432740626207963,
      "grad_norm": 0.0,
      "learning_rate": 4.278604561267878e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2987
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.9375,
      "epoch": 0.14437572477773483,
      "grad_norm": 0.0,
      "learning_rate": 4.2783629686896015e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2988
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.84375,
      "epoch": 0.14442404329339004,
      "grad_norm": 53.045394962533486,
      "learning_rate": 4.2781213761113256e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.29958203434944153,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 2989
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 80.65625,
      "epoch": 0.1444723618090452,
      "grad_norm": 13.223052498729375,
      "learning_rate": 4.27787978353305e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.36295419931411743,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 2990
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 71.0,
      "epoch": 0.14452068032470042,
      "grad_norm": 0.0,
      "learning_rate": 4.277638190954774e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 2991
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 64.75,
      "epoch": 0.14456899884035562,
      "grad_norm": 28.517268837956298,
      "learning_rate": 4.2773965983764975e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2651395797729492,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 2992
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 69.09375,
      "epoch": 0.14461731735601083,
      "grad_norm": 13.949374391322605,
      "learning_rate": 4.2771550057982217e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.19266104698181152,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 2993
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 68.75,
      "epoch": 0.14466563587166603,
      "grad_norm": 13.393675029870156,
      "learning_rate": 4.276913413219946e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2608962655067444,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 2994
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.8125,
      "epoch": 0.1447139543873212,
      "grad_norm": 0.0,
      "learning_rate": 4.2766718206416694e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 2995
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.78125,
      "epoch": 0.1447622729029764,
      "grad_norm": 0.0,
      "learning_rate": 4.2764302280633935e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 2996
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.53125,
      "epoch": 0.14481059141863162,
      "grad_norm": 0.0,
      "learning_rate": 4.2761886354851177e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 2997
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 70.84375,
      "epoch": 0.14485890993428682,
      "grad_norm": 4.486963671587694,
      "learning_rate": 4.275947042906842e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 2998
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.46875,
      "epoch": 0.14490722844994203,
      "grad_norm": 26.038548179556983,
      "learning_rate": 4.2757054503285654e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 2999
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.96875,
      "epoch": 0.1449555469655972,
      "grad_norm": 0.0,
      "learning_rate": 4.27546385775029e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 3000
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.75,
      "epoch": 0.1450038654812524,
      "grad_norm": 0.0,
      "learning_rate": 4.2752222651720137e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3001
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.09375,
      "epoch": 0.1450521839969076,
      "grad_norm": 0.0,
      "learning_rate": 4.274980672593738e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3002
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.875,
      "epoch": 0.14510050251256282,
      "grad_norm": 0.0,
      "learning_rate": 4.2747390800154614e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3003
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.125,
      "epoch": 0.14514882102821802,
      "grad_norm": 0.0,
      "learning_rate": 4.274497487437186e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3004
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.21875,
      "epoch": 0.1451971395438732,
      "grad_norm": 0.0,
      "learning_rate": 4.2742558948589097e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3005
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.3125,
      "epoch": 0.1452454580595284,
      "grad_norm": 13.463010306266083,
      "learning_rate": 4.2740143022806333e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3006
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.5625,
      "epoch": 0.1452937765751836,
      "grad_norm": 20.539153253696984,
      "learning_rate": 4.273772709702358e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.27841314673423767,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3007
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.375,
      "epoch": 0.1453420950908388,
      "grad_norm": 0.0,
      "learning_rate": 4.2735311171240816e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3008
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.125,
      "epoch": 0.14539041360649402,
      "grad_norm": 0.0,
      "learning_rate": 4.2732895245458057e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3009
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.59375,
      "epoch": 0.1454387321221492,
      "grad_norm": 0.0,
      "learning_rate": 4.27304793196753e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3010
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.1875,
      "epoch": 0.1454870506378044,
      "grad_norm": 0.0,
      "learning_rate": 4.272806339389254e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3011
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.21875,
      "epoch": 0.1455353691534596,
      "grad_norm": 0.0,
      "learning_rate": 4.2725647468109776e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3012
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.90625,
      "epoch": 0.1455836876691148,
      "grad_norm": 0.0,
      "learning_rate": 4.272323154232702e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 3013
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.25,
      "epoch": 0.14563200618477,
      "grad_norm": 0.0,
      "learning_rate": 4.272081561654426e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3014
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.53125,
      "epoch": 0.1456803247004252,
      "grad_norm": 0.0,
      "learning_rate": 4.2718399690761495e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3015
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.75,
      "epoch": 0.1457286432160804,
      "grad_norm": 0.0,
      "learning_rate": 4.2715983764978736e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3016
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.34375,
      "epoch": 0.1457769617317356,
      "grad_norm": 0.0,
      "learning_rate": 4.271356783919598e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 3017
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.0625,
      "epoch": 0.1458252802473908,
      "grad_norm": 0.0,
      "learning_rate": 4.271115191341322e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3018
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 69.625,
      "epoch": 0.145873598763046,
      "grad_norm": 21.027040773182573,
      "learning_rate": 4.2708735987630455e-07,
      "loss": 0.0,
      "reward": 1.390625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 3019
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.96875,
      "epoch": 0.14592191727870119,
      "grad_norm": 0.0,
      "learning_rate": 4.27063200618477e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3020
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.75,
      "epoch": 0.1459702357943564,
      "grad_norm": 0.0,
      "learning_rate": 4.270390413606494e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3021
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.34375,
      "epoch": 0.1460185543100116,
      "grad_norm": 0.0,
      "learning_rate": 4.270148821028218e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3022
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 71.625,
      "epoch": 0.1460668728256668,
      "grad_norm": 9.404281661099143,
      "learning_rate": 4.2699072284499415e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3023
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.0,
      "epoch": 0.146115191341322,
      "grad_norm": 0.0,
      "learning_rate": 4.269665635871666e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3024
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.5625,
      "epoch": 0.14616350985697718,
      "grad_norm": 0.0,
      "learning_rate": 4.26942404329339e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3025
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 76.1875,
      "epoch": 0.14621182837263239,
      "grad_norm": 16.59111446481932,
      "learning_rate": 4.2691824507151134e-07,
      "loss": -0.0,
      "reward": 1.90625,
      "reward_std": 0.23243674635887146,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3026
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.96875,
      "epoch": 0.1462601468882876,
      "grad_norm": 0.0,
      "learning_rate": 4.268940858136838e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3027
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.65625,
      "epoch": 0.1463084654039428,
      "grad_norm": 0.0,
      "learning_rate": 4.2686992655585617e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3028
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 69.6875,
      "epoch": 0.146356783919598,
      "grad_norm": 36.81898233426351,
      "learning_rate": 4.268457672980286e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.3014557659626007,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3029
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.28125,
      "epoch": 0.14640510243525318,
      "grad_norm": 0.0,
      "learning_rate": 4.26821608040201e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 3030
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.75,
      "epoch": 0.14645342095090838,
      "grad_norm": 0.0,
      "learning_rate": 4.267974487823734e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3031
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.84375,
      "epoch": 0.14650173946656359,
      "grad_norm": 0.0,
      "learning_rate": 4.2677328952454577e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3032
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.375,
      "epoch": 0.1465500579822188,
      "grad_norm": 0.0,
      "learning_rate": 4.267491302667182e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3033
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.21875,
      "epoch": 0.146598376497874,
      "grad_norm": 5.25907513914757,
      "learning_rate": 4.267249710088906e-07,
      "loss": -0.0,
      "reward": 2.390625,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3034
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.28125,
      "epoch": 0.14664669501352917,
      "grad_norm": 0.0,
      "learning_rate": 4.26700811751063e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3035
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.0,
      "epoch": 0.14669501352918438,
      "grad_norm": 0.0,
      "learning_rate": 4.2667665249323537e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3036
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.9375,
      "epoch": 0.14674333204483958,
      "grad_norm": 0.0,
      "learning_rate": 4.266524932354078e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3037
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.21875,
      "epoch": 0.14679165056049479,
      "grad_norm": 0.0,
      "learning_rate": 4.266283339775802e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3038
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.71875,
      "epoch": 0.14683996907615,
      "grad_norm": 0.0,
      "learning_rate": 4.2660417471975256e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3039
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.71875,
      "epoch": 0.14688828759180517,
      "grad_norm": 0.0,
      "learning_rate": 4.26580015461925e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3040
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.875,
      "epoch": 0.14693660610746037,
      "grad_norm": 0.0,
      "learning_rate": 4.265558562040974e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3041
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.46875,
      "epoch": 0.14698492462311558,
      "grad_norm": 0.0,
      "learning_rate": 4.265316969462698e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3042
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.90625,
      "epoch": 0.14703324313877078,
      "grad_norm": 0.0,
      "learning_rate": 4.2650753768844216e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3043
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 89.15625,
      "epoch": 0.14708156165442599,
      "grad_norm": 13.909739691157737,
      "learning_rate": 4.264833784306146e-07,
      "loss": -0.0,
      "reward": 1.453125,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3044
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.8125,
      "epoch": 0.14712988017008116,
      "grad_norm": 5.78604325112041,
      "learning_rate": 4.26459219172787e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3045
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 71.46875,
      "epoch": 0.14717819868573637,
      "grad_norm": 15.298265876231024,
      "learning_rate": 4.2643505991495935e-07,
      "loss": -0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3046
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.0625,
      "epoch": 0.14722651720139157,
      "grad_norm": 0.0,
      "learning_rate": 4.264109006571318e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3047
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.78125,
      "epoch": 0.14727483571704678,
      "grad_norm": 0.0,
      "learning_rate": 4.263867413993042e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3048
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 60.5625,
      "epoch": 0.14732315423270198,
      "grad_norm": 23.427234083291022,
      "learning_rate": 4.263625821414766e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3049
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.625,
      "epoch": 0.14737147274835716,
      "grad_norm": 0.0,
      "learning_rate": 4.26338422883649e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3050
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.5625,
      "epoch": 0.14741979126401236,
      "grad_norm": 0.0,
      "learning_rate": 4.263142636258214e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3051
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 69.25,
      "epoch": 0.14746810977966757,
      "grad_norm": 0.0,
      "learning_rate": 4.262901043679938e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3052
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.46875,
      "epoch": 0.14751642829532277,
      "grad_norm": 0.0,
      "learning_rate": 4.262659451101662e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3053
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.40625,
      "epoch": 0.14756474681097798,
      "grad_norm": 0.0,
      "learning_rate": 4.262417858523386e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3054
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.4375,
      "epoch": 0.14761306532663315,
      "grad_norm": 0.0,
      "learning_rate": 4.26217626594511e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3055
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.59375,
      "epoch": 0.14766138384228836,
      "grad_norm": 35.24786045138425,
      "learning_rate": 4.261934673366834e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3056
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.53125,
      "epoch": 0.14770970235794356,
      "grad_norm": 0.0,
      "learning_rate": 4.261693080788558e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3057
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.375,
      "epoch": 0.14775802087359877,
      "grad_norm": 0.0,
      "learning_rate": 4.261451488210282e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3058
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.59375,
      "epoch": 0.14780633938925397,
      "grad_norm": 0.0,
      "learning_rate": 4.2612098956320057e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3059
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.59375,
      "epoch": 0.14785465790490915,
      "grad_norm": 0.0,
      "learning_rate": 4.2609683030537303e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3060
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 76.03125,
      "epoch": 0.14790297642056435,
      "grad_norm": 69.75137837338424,
      "learning_rate": 4.260726710475454e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.31663835048675537,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3061
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.09375,
      "epoch": 0.14795129493621956,
      "grad_norm": 0.0,
      "learning_rate": 4.260485117897178e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3062
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.90625,
      "epoch": 0.14799961345187476,
      "grad_norm": 0.0,
      "learning_rate": 4.2602435253189017e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3063
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.25,
      "epoch": 0.14804793196752997,
      "grad_norm": 0.0,
      "learning_rate": 4.2600019327406263e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3064
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.53125,
      "epoch": 0.14809625048318514,
      "grad_norm": 0.0,
      "learning_rate": 4.25976034016235e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3065
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.40625,
      "epoch": 0.14814456899884035,
      "grad_norm": 0.0,
      "learning_rate": 4.2595187475840735e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3066
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.21875,
      "epoch": 0.14819288751449555,
      "grad_norm": 0.0,
      "learning_rate": 4.259277155005798e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3067
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.78125,
      "epoch": 0.14824120603015076,
      "grad_norm": 0.0,
      "learning_rate": 4.259035562427522e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3068
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.9375,
      "epoch": 0.14828952454580596,
      "grad_norm": 0.0,
      "learning_rate": 4.258793969849246e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 3069
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.65625,
      "epoch": 0.14833784306146114,
      "grad_norm": 0.0,
      "learning_rate": 4.25855237727097e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3070
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.375,
      "epoch": 0.14838616157711634,
      "grad_norm": 0.0,
      "learning_rate": 4.258310784692694e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3071
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.4375,
      "epoch": 0.14843448009277155,
      "grad_norm": 0.0,
      "learning_rate": 4.258069192114418e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3072
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.625,
      "epoch": 0.14848279860842675,
      "grad_norm": 0.0,
      "learning_rate": 4.257827599536142e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3073
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.75,
      "epoch": 0.14853111712408196,
      "grad_norm": 0.0,
      "learning_rate": 4.257586006957866e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3074
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.78125,
      "epoch": 0.14857943563973713,
      "grad_norm": 0.0,
      "learning_rate": 4.25734441437959e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3075
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.8125,
      "epoch": 0.14862775415539234,
      "grad_norm": 0.0,
      "learning_rate": 4.257102821801314e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3076
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.03125,
      "epoch": 0.14867607267104754,
      "grad_norm": 0.0,
      "learning_rate": 4.256861229223038e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3077
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.625,
      "epoch": 0.14872439118670275,
      "grad_norm": 0.0,
      "learning_rate": 4.256619636644762e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3078
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.15625,
      "epoch": 0.14877270970235795,
      "grad_norm": 0.0,
      "learning_rate": 4.2563780440664857e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3079
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.21875,
      "epoch": 0.14882102821801313,
      "grad_norm": 0.0,
      "learning_rate": 4.2561364514882104e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3080
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 89.46875,
      "epoch": 0.14886934673366833,
      "grad_norm": 7.086985458281214,
      "learning_rate": 4.255894858909934e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3081
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.0625,
      "epoch": 0.14891766524932354,
      "grad_norm": 9.246091637685497,
      "learning_rate": 4.255653266331658e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3082
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.1875,
      "epoch": 0.14896598376497874,
      "grad_norm": 0.0,
      "learning_rate": 4.255411673753382e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3083
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.96875,
      "epoch": 0.14901430228063395,
      "grad_norm": 0.0,
      "learning_rate": 4.2551700811751064e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3084
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.09375,
      "epoch": 0.14906262079628912,
      "grad_norm": 0.0,
      "learning_rate": 4.25492848859683e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3085
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.375,
      "epoch": 0.14911093931194433,
      "grad_norm": 0.0,
      "learning_rate": 4.254686896018554e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3086
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.0625,
      "epoch": 0.14915925782759953,
      "grad_norm": 11.706436440963234,
      "learning_rate": 4.2544453034402783e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3087
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.53125,
      "epoch": 0.14920757634325474,
      "grad_norm": 0.0,
      "learning_rate": 4.254203710862002e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3088
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.0625,
      "epoch": 0.14925589485890994,
      "grad_norm": 0.0,
      "learning_rate": 4.253962118283726e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3089
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.90625,
      "epoch": 0.14930421337456512,
      "grad_norm": 8.918849159024319,
      "learning_rate": 4.25372052570545e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23421898484230042,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3090
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.71875,
      "epoch": 0.14935253189022032,
      "grad_norm": 0.0,
      "learning_rate": 4.2534789331271743e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3091
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 68.59375,
      "epoch": 0.14940085040587553,
      "grad_norm": 0.0,
      "learning_rate": 4.253237340548898e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3092
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.53125,
      "epoch": 0.14944916892153073,
      "grad_norm": 26.95492390513556,
      "learning_rate": 4.252995747970622e-07,
      "loss": -0.0,
      "reward": 2.09375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3093
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.53125,
      "epoch": 0.14949748743718594,
      "grad_norm": 0.0,
      "learning_rate": 4.252754155392346e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3094
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.15625,
      "epoch": 0.14954580595284112,
      "grad_norm": 0.0,
      "learning_rate": 4.2525125628140703e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3095
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.0,
      "epoch": 0.14959412446849632,
      "grad_norm": 0.0,
      "learning_rate": 4.252270970235794e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3096
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.09375,
      "epoch": 0.14964244298415152,
      "grad_norm": 0.0,
      "learning_rate": 4.2520293776575186e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3097
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.21875,
      "epoch": 0.14969076149980673,
      "grad_norm": 0.0,
      "learning_rate": 4.251787785079242e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3098
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.375,
      "epoch": 0.14973908001546193,
      "grad_norm": 0.0,
      "learning_rate": 4.251546192500966e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3099
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.25,
      "epoch": 0.1497873985311171,
      "grad_norm": 0.0,
      "learning_rate": 4.2513045999226905e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3100
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.9375,
      "epoch": 0.14983571704677232,
      "grad_norm": 0.0,
      "learning_rate": 4.251063007344414e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3101
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.375,
      "epoch": 0.14988403556242752,
      "grad_norm": 62.197063427511445,
      "learning_rate": 4.250821414766138e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.3234682083129883,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3102
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.625,
      "epoch": 0.14993235407808272,
      "grad_norm": 0.0,
      "learning_rate": 4.250579822187862e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3103
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.125,
      "epoch": 0.14998067259373793,
      "grad_norm": 0.0,
      "learning_rate": 4.2503382296095865e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3104
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.125,
      "epoch": 0.1500289911093931,
      "grad_norm": 0.0,
      "learning_rate": 4.25009663703131e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3105
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.09375,
      "epoch": 0.1500773096250483,
      "grad_norm": 8.941455298553233,
      "learning_rate": 4.249855044453034e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3106
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.90625,
      "epoch": 0.15012562814070352,
      "grad_norm": 0.0,
      "learning_rate": 4.2496134518747584e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3107
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.40625,
      "epoch": 0.15017394665635872,
      "grad_norm": 0.0,
      "learning_rate": 4.249371859296482e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3108
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.75,
      "epoch": 0.15022226517201392,
      "grad_norm": 0.0,
      "learning_rate": 4.249130266718206e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3109
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.28125,
      "epoch": 0.1502705836876691,
      "grad_norm": 11.538297633520681,
      "learning_rate": 4.24888867413993e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2951723635196686,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3110
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.3125,
      "epoch": 0.1503189022033243,
      "grad_norm": 0.0,
      "learning_rate": 4.2486470815616544e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3111
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.75,
      "epoch": 0.1503672207189795,
      "grad_norm": 0.0,
      "learning_rate": 4.248405488983378e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3112
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.125,
      "epoch": 0.15041553923463472,
      "grad_norm": 0.0,
      "learning_rate": 4.248163896405102e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3113
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.5625,
      "epoch": 0.15046385775028992,
      "grad_norm": 0.0,
      "learning_rate": 4.2479223038268263e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3114
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.9375,
      "epoch": 0.1505121762659451,
      "grad_norm": 0.0,
      "learning_rate": 4.2476807112485504e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3115
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.90625,
      "epoch": 0.1505604947816003,
      "grad_norm": 0.0,
      "learning_rate": 4.247439118670274e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3116
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.625,
      "epoch": 0.1506088132972555,
      "grad_norm": 67.73377450995117,
      "learning_rate": 4.2471975260919987e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3117
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.5,
      "epoch": 0.1506571318129107,
      "grad_norm": 0.0,
      "learning_rate": 4.2469559335137223e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.22201895713806152,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3118
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.625,
      "epoch": 0.15070545032856592,
      "grad_norm": 0.0,
      "learning_rate": 4.246714340935446e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3119
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.03125,
      "epoch": 0.1507537688442211,
      "grad_norm": 0.0,
      "learning_rate": 4.2464727483571706e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3120
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.90625,
      "epoch": 0.1508020873598763,
      "grad_norm": 0.0,
      "learning_rate": 4.246231155778894e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3121
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.875,
      "epoch": 0.1508504058755315,
      "grad_norm": 16.232797066366423,
      "learning_rate": 4.2459895632006183e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.16807307302951813,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3122
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.3125,
      "epoch": 0.1508987243911867,
      "grad_norm": 0.0,
      "learning_rate": 4.245747970622342e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3123
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.9375,
      "epoch": 0.1509470429068419,
      "grad_norm": 0.0,
      "learning_rate": 4.2455063780440666e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3124
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.875,
      "epoch": 0.1509953614224971,
      "grad_norm": 0.0,
      "learning_rate": 4.24526478546579e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3125
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.0,
      "epoch": 0.1510436799381523,
      "grad_norm": 0.0,
      "learning_rate": 4.2450231928875143e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3126
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.9375,
      "epoch": 0.1510919984538075,
      "grad_norm": 0.0,
      "learning_rate": 4.2447816003092385e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3127
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.625,
      "epoch": 0.1511403169694627,
      "grad_norm": 0.0,
      "learning_rate": 4.2445400077309626e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3128
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.25,
      "epoch": 0.1511886354851179,
      "grad_norm": 0.0,
      "learning_rate": 4.244298415152686e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3129
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.3125,
      "epoch": 0.15123695400077308,
      "grad_norm": 0.0,
      "learning_rate": 4.2440568225744103e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3130
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.6875,
      "epoch": 0.1512852725164283,
      "grad_norm": 0.0,
      "learning_rate": 4.2438152299961345e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3131
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.5,
      "epoch": 0.1513335910320835,
      "grad_norm": 0.0,
      "learning_rate": 4.243573637417858e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3132
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.65625,
      "epoch": 0.1513819095477387,
      "grad_norm": 0.0,
      "learning_rate": 4.243332044839582e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3133
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.5625,
      "epoch": 0.1514302280633939,
      "grad_norm": 0.0,
      "learning_rate": 4.2430904522613063e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3134
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.40625,
      "epoch": 0.15147854657904908,
      "grad_norm": 0.0,
      "learning_rate": 4.2428488596830305e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3135
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.625,
      "epoch": 0.15152686509470428,
      "grad_norm": 0.0,
      "learning_rate": 4.242607267104754e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3136
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.75,
      "epoch": 0.1515751836103595,
      "grad_norm": 0.0,
      "learning_rate": 4.242365674526479e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3137
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.78125,
      "epoch": 0.1516235021260147,
      "grad_norm": 0.0,
      "learning_rate": 4.2421240819482024e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3138
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.46875,
      "epoch": 0.1516718206416699,
      "grad_norm": 0.0,
      "learning_rate": 4.241882489369926e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3139
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.3125,
      "epoch": 0.15172013915732507,
      "grad_norm": 0.0,
      "learning_rate": 4.2416408967916506e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3140
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.71875,
      "epoch": 0.15176845767298028,
      "grad_norm": 0.0,
      "learning_rate": 4.241399304213374e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3141
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.40625,
      "epoch": 0.15181677618863548,
      "grad_norm": 0.0,
      "learning_rate": 4.2411577116350984e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3142
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.3125,
      "epoch": 0.1518650947042907,
      "grad_norm": 0.0,
      "learning_rate": 4.240916119056822e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3143
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.8125,
      "epoch": 0.1519134132199459,
      "grad_norm": 0.0,
      "learning_rate": 4.2406745264785467e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3144
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.65625,
      "epoch": 0.15196173173560107,
      "grad_norm": 0.0,
      "learning_rate": 4.24043293390027e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3145
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 67.53125,
      "epoch": 0.15201005025125627,
      "grad_norm": 8.49569634245118,
      "learning_rate": 4.2401913413219944e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.3383457660675049,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3146
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.4375,
      "epoch": 0.15205836876691148,
      "grad_norm": 0.0,
      "learning_rate": 4.2399497487437185e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3147
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.53125,
      "epoch": 0.15210668728256668,
      "grad_norm": 0.0,
      "learning_rate": 4.2397081561654427e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3148
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.9375,
      "epoch": 0.1521550057982219,
      "grad_norm": 0.0,
      "learning_rate": 4.2394665635871663e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3149
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.40625,
      "epoch": 0.15220332431387706,
      "grad_norm": 0.0,
      "learning_rate": 4.2392249710088904e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3150
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.8125,
      "epoch": 0.15225164282953227,
      "grad_norm": 0.0,
      "learning_rate": 4.2389833784306145e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3151
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.15625,
      "epoch": 0.15229996134518747,
      "grad_norm": 0.0,
      "learning_rate": 4.238741785852338e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3152
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.15625,
      "epoch": 0.15234827986084268,
      "grad_norm": 0.0,
      "learning_rate": 4.2385001932740623e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3153
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.96875,
      "epoch": 0.15239659837649788,
      "grad_norm": 0.0,
      "learning_rate": 4.2382586006957864e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3154
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.40625,
      "epoch": 0.15244491689215306,
      "grad_norm": 0.0,
      "learning_rate": 4.2380170081175106e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3155
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.90625,
      "epoch": 0.15249323540780826,
      "grad_norm": 0.0,
      "learning_rate": 4.237775415539234e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3156
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.15625,
      "epoch": 0.15254155392346347,
      "grad_norm": 0.0,
      "learning_rate": 4.237533822960959e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3157
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.625,
      "epoch": 0.15258987243911867,
      "grad_norm": 0.0,
      "learning_rate": 4.2372922303826824e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3158
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.09375,
      "epoch": 0.15263819095477388,
      "grad_norm": 0.0,
      "learning_rate": 4.237050637804406e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3159
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.625,
      "epoch": 0.15268650947042905,
      "grad_norm": 0.0,
      "learning_rate": 4.2368090452261307e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3160
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.96875,
      "epoch": 0.15273482798608426,
      "grad_norm": 0.0,
      "learning_rate": 4.2365674526478543e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3161
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 80.71875,
      "epoch": 0.15278314650173946,
      "grad_norm": 19.852488976022492,
      "learning_rate": 4.2363258600695785e-07,
      "loss": -0.0,
      "reward": 2.21875,
      "reward_std": 0.3107999563217163,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3162
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.59375,
      "epoch": 0.15283146501739467,
      "grad_norm": 0.0,
      "learning_rate": 4.236084267491302e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3163
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.8125,
      "epoch": 0.15287978353304987,
      "grad_norm": 0.0,
      "learning_rate": 4.2358426749130267e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3164
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.15625,
      "epoch": 0.15292810204870505,
      "grad_norm": 0.0,
      "learning_rate": 4.2356010823347503e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3165
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.21875,
      "epoch": 0.15297642056436025,
      "grad_norm": 0.0,
      "learning_rate": 4.2353594897564745e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3166
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.625,
      "epoch": 0.15302473908001546,
      "grad_norm": 0.0,
      "learning_rate": 4.2351178971781986e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 3167
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.28125,
      "epoch": 0.15307305759567066,
      "grad_norm": 0.0,
      "learning_rate": 4.234876304599923e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3168
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.125,
      "epoch": 0.15312137611132587,
      "grad_norm": 0.0,
      "learning_rate": 4.2346347120216464e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3169
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.46875,
      "epoch": 0.15316969462698105,
      "grad_norm": 0.0,
      "learning_rate": 4.2343931194433705e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3170
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.1875,
      "epoch": 0.15321801314263625,
      "grad_norm": 0.0,
      "learning_rate": 4.2341515268650946e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3171
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.1875,
      "epoch": 0.15326633165829145,
      "grad_norm": 0.0,
      "learning_rate": 4.233909934286818e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3172
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.59375,
      "epoch": 0.15331465017394666,
      "grad_norm": 0.0,
      "learning_rate": 4.2336683417085424e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3173
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.1875,
      "epoch": 0.15336296868960186,
      "grad_norm": 0.0,
      "learning_rate": 4.2334267491302665e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3174
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.75,
      "epoch": 0.15341128720525704,
      "grad_norm": 0.0,
      "learning_rate": 4.2331851565519906e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 3175
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.1875,
      "epoch": 0.15345960572091225,
      "grad_norm": 0.0,
      "learning_rate": 4.232943563973714e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3176
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.71875,
      "epoch": 0.15350792423656745,
      "grad_norm": 0.0,
      "learning_rate": 4.232701971395439e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3177
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.75,
      "epoch": 0.15355624275222265,
      "grad_norm": 0.0,
      "learning_rate": 4.2324603788171625e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3178
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.6875,
      "epoch": 0.15360456126787786,
      "grad_norm": 0.0,
      "learning_rate": 4.2322187862388867e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3179
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.3125,
      "epoch": 0.15365287978353304,
      "grad_norm": 0.0,
      "learning_rate": 4.231977193660611e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3180
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.90625,
      "epoch": 0.15370119829918824,
      "grad_norm": 0.0,
      "learning_rate": 4.2317356010823344e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3181
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.65625,
      "epoch": 0.15374951681484345,
      "grad_norm": 0.0,
      "learning_rate": 4.2314940085040585e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3182
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.875,
      "epoch": 0.15379783533049865,
      "grad_norm": 0.0,
      "learning_rate": 4.231252415925782e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3183
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.09375,
      "epoch": 0.15384615384615385,
      "grad_norm": 0.0,
      "learning_rate": 4.231010823347507e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3184
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.375,
      "epoch": 0.15389447236180903,
      "grad_norm": 0.0,
      "learning_rate": 4.2307692307692304e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3185
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.46875,
      "epoch": 0.15394279087746424,
      "grad_norm": 0.0,
      "learning_rate": 4.2305276381909546e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3186
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.125,
      "epoch": 0.15399110939311944,
      "grad_norm": 17.01314186552196,
      "learning_rate": 4.2302860456126787e-07,
      "loss": -0.0,
      "reward": 2.296875,
      "reward_std": 0.24464011192321777,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3187
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.78125,
      "epoch": 0.15403942790877465,
      "grad_norm": 0.0,
      "learning_rate": 4.230044453034403e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3188
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.5625,
      "epoch": 0.15408774642442985,
      "grad_norm": 0.0,
      "learning_rate": 4.2298028604561264e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3189
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.5625,
      "epoch": 0.15413606494008503,
      "grad_norm": 0.0,
      "learning_rate": 4.229561267877851e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3190
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.09375,
      "epoch": 0.15418438345574023,
      "grad_norm": 0.0,
      "learning_rate": 4.2293196752995747e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3191
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.46875,
      "epoch": 0.15423270197139544,
      "grad_norm": 0.0,
      "learning_rate": 4.2290780827212983e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3192
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.65625,
      "epoch": 0.15428102048705064,
      "grad_norm": 0.0,
      "learning_rate": 4.2288364901430224e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3193
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.03125,
      "epoch": 0.15432933900270585,
      "grad_norm": 0.0,
      "learning_rate": 4.2285948975647466e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3194
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.5,
      "epoch": 0.15437765751836102,
      "grad_norm": 0.0,
      "learning_rate": 4.2283533049864707e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3195
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.28125,
      "epoch": 0.15442597603401623,
      "grad_norm": 0.0,
      "learning_rate": 4.2281117124081943e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3196
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.21875,
      "epoch": 0.15447429454967143,
      "grad_norm": 0.0,
      "learning_rate": 4.227870119829919e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3197
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.15625,
      "epoch": 0.15452261306532664,
      "grad_norm": 0.0,
      "learning_rate": 4.2276285272516426e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3198
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.40625,
      "epoch": 0.15457093158098184,
      "grad_norm": 0.0,
      "learning_rate": 4.2273869346733667e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3199
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.0,
      "epoch": 0.15461925009663702,
      "grad_norm": 0.0,
      "learning_rate": 4.227145342095091e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3200
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.8125,
      "epoch": 0.15466756861229222,
      "grad_norm": 0.0,
      "learning_rate": 4.2269037495168145e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3201
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.625,
      "epoch": 0.15471588712794743,
      "grad_norm": 0.0,
      "learning_rate": 4.2266621569385386e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3202
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.6875,
      "epoch": 0.15476420564360263,
      "grad_norm": 0.0,
      "learning_rate": 4.226420564360262e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3203
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.96875,
      "epoch": 0.15481252415925784,
      "grad_norm": 0.0,
      "learning_rate": 4.226178971781987e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3204
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 90.125,
      "epoch": 0.154860842674913,
      "grad_norm": 0.0,
      "learning_rate": 4.2259373792037105e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 3205
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.71875,
      "epoch": 0.15490916119056822,
      "grad_norm": 0.0,
      "learning_rate": 4.2256957866254346e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3206
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.90625,
      "epoch": 0.15495747970622342,
      "grad_norm": 0.0,
      "learning_rate": 4.225454194047159e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3207
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.3125,
      "epoch": 0.15500579822187863,
      "grad_norm": 0.0,
      "learning_rate": 4.225212601468883e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3208
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.46875,
      "epoch": 0.15505411673753383,
      "grad_norm": 0.0,
      "learning_rate": 4.2249710088906065e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3209
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.09375,
      "epoch": 0.155102435253189,
      "grad_norm": 0.0,
      "learning_rate": 4.224729416312331e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3210
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.375,
      "epoch": 0.1551507537688442,
      "grad_norm": 0.0,
      "learning_rate": 4.224487823734055e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 3211
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.09375,
      "epoch": 0.15519907228449942,
      "grad_norm": 0.0,
      "learning_rate": 4.2242462311557784e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3212
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.1875,
      "epoch": 0.15524739080015462,
      "grad_norm": 0.0,
      "learning_rate": 4.2240046385775025e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3213
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.8125,
      "epoch": 0.15529570931580983,
      "grad_norm": 0.0,
      "learning_rate": 4.2237630459992267e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3214
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.1875,
      "epoch": 0.155344027831465,
      "grad_norm": 0.0,
      "learning_rate": 4.223521453420951e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 3215
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.90625,
      "epoch": 0.1553923463471202,
      "grad_norm": 0.0,
      "learning_rate": 4.2232798608426744e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3216
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.1875,
      "epoch": 0.1554406648627754,
      "grad_norm": 0.0,
      "learning_rate": 4.223038268264399e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3217
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.34375,
      "epoch": 0.15548898337843062,
      "grad_norm": 0.0,
      "learning_rate": 4.2227966756861227e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3218
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.59375,
      "epoch": 0.15553730189408582,
      "grad_norm": 0.0,
      "learning_rate": 4.222555083107847e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3219
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.3125,
      "epoch": 0.155585620409741,
      "grad_norm": 0.0,
      "learning_rate": 4.222313490529571e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3220
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.03125,
      "epoch": 0.1556339389253962,
      "grad_norm": 13.785635877103857,
      "learning_rate": 4.2220718979512946e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.3039700984954834,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3221
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.4375,
      "epoch": 0.1556822574410514,
      "grad_norm": 0.0,
      "learning_rate": 4.2218303053730187e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3222
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.4375,
      "epoch": 0.1557305759567066,
      "grad_norm": 0.0,
      "learning_rate": 4.2215887127947423e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3223
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.84375,
      "epoch": 0.15577889447236182,
      "grad_norm": 0.0,
      "learning_rate": 4.221347120216467e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3224
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.53125,
      "epoch": 0.155827212988017,
      "grad_norm": 0.0,
      "learning_rate": 4.2211055276381906e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3225
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.4375,
      "epoch": 0.1558755315036722,
      "grad_norm": 0.0,
      "learning_rate": 4.2208639350599147e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3226
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.0625,
      "epoch": 0.1559238500193274,
      "grad_norm": 0.0,
      "learning_rate": 4.220622342481639e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3227
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 89.46875,
      "epoch": 0.1559721685349826,
      "grad_norm": 25.896930503687226,
      "learning_rate": 4.220380749903363e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.24464011192321777,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3228
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.5,
      "epoch": 0.1560204870506378,
      "grad_norm": 0.0,
      "learning_rate": 4.2201391573250866e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3229
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.375,
      "epoch": 0.156068805566293,
      "grad_norm": 0.0,
      "learning_rate": 4.219897564746811e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3230
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.71875,
      "epoch": 0.1561171240819482,
      "grad_norm": 0.0,
      "learning_rate": 4.219655972168535e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3231
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.25,
      "epoch": 0.1561654425976034,
      "grad_norm": 0.0,
      "learning_rate": 4.2194143795902585e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3232
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.78125,
      "epoch": 0.1562137611132586,
      "grad_norm": 0.0,
      "learning_rate": 4.2191727870119826e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3233
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 96.125,
      "epoch": 0.1562620796289138,
      "grad_norm": 24.910597998645155,
      "learning_rate": 4.218931194433707e-07,
      "loss": -0.0,
      "reward": 1.5,
      "reward_std": 0.2529362142086029,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3234
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.65625,
      "epoch": 0.156310398144569,
      "grad_norm": 0.0,
      "learning_rate": 4.218689601855431e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3235
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 82.0,
      "epoch": 0.1563587166602242,
      "grad_norm": 6.306761215818737,
      "learning_rate": 4.2184480092771545e-07,
      "loss": 0.0,
      "reward": 1.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.03125,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 3236
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.9375,
      "epoch": 0.1564070351758794,
      "grad_norm": 0.0,
      "learning_rate": 4.218206416698879e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3237
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.65625,
      "epoch": 0.1564553536915346,
      "grad_norm": 0.0,
      "learning_rate": 4.217964824120603e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3238
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 101.59375,
      "epoch": 0.1565036722071898,
      "grad_norm": 7.536879087248743,
      "learning_rate": 4.217723231542327e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3239
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 94.4375,
      "epoch": 0.156551990722845,
      "grad_norm": 13.683470026863494,
      "learning_rate": 4.217481638964051e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3240
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.59375,
      "epoch": 0.15660030923850018,
      "grad_norm": 0.0,
      "learning_rate": 4.217240046385775e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3241
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.6875,
      "epoch": 0.1566486277541554,
      "grad_norm": 0.0,
      "learning_rate": 4.216998453807499e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3242
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.875,
      "epoch": 0.1566969462698106,
      "grad_norm": 0.0,
      "learning_rate": 4.2167568612292224e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3243
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.21875,
      "epoch": 0.1567452647854658,
      "grad_norm": 0.0,
      "learning_rate": 4.216515268650947e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3244
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.0625,
      "epoch": 0.156793583301121,
      "grad_norm": 0.0,
      "learning_rate": 4.2162736760726707e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3245
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.6875,
      "epoch": 0.15684190181677618,
      "grad_norm": 0.0,
      "learning_rate": 4.216032083494395e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3246
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.09375,
      "epoch": 0.15689022033243138,
      "grad_norm": 0.0,
      "learning_rate": 4.215790490916119e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3247
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.0,
      "epoch": 0.1569385388480866,
      "grad_norm": 0.0,
      "learning_rate": 4.215548898337843e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 3248
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.53125,
      "epoch": 0.1569868573637418,
      "grad_norm": 0.0,
      "learning_rate": 4.2153073057595667e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3249
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.84375,
      "epoch": 0.157035175879397,
      "grad_norm": 0.0,
      "learning_rate": 4.2150657131812913e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3250
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.84375,
      "epoch": 0.15708349439505218,
      "grad_norm": 0.0,
      "learning_rate": 4.214824120603015e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3251
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.21875,
      "epoch": 0.15713181291070738,
      "grad_norm": 0.0,
      "learning_rate": 4.2145825280247385e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3252
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.5625,
      "epoch": 0.15718013142636258,
      "grad_norm": 0.0,
      "learning_rate": 4.2143409354464627e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3253
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.71875,
      "epoch": 0.1572284499420178,
      "grad_norm": 0.0,
      "learning_rate": 4.214099342868187e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3254
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.90625,
      "epoch": 0.157276768457673,
      "grad_norm": 0.0,
      "learning_rate": 4.213857750289911e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3255
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.375,
      "epoch": 0.15732508697332817,
      "grad_norm": 0.0,
      "learning_rate": 4.2136161577116346e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3256
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 80.5625,
      "epoch": 0.15737340548898338,
      "grad_norm": 10.530608377216229,
      "learning_rate": 4.213374565133359e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3257
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.53125,
      "epoch": 0.15742172400463858,
      "grad_norm": 0.0,
      "learning_rate": 4.213132972555083e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3258
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.09375,
      "epoch": 0.15747004252029378,
      "grad_norm": 0.0,
      "learning_rate": 4.212891379976807e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3259
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.8125,
      "epoch": 0.157518361035949,
      "grad_norm": 28.087234692879072,
      "learning_rate": 4.212649787398531e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.2961486577987671,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3260
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.9375,
      "epoch": 0.15756667955160417,
      "grad_norm": 0.0,
      "learning_rate": 4.212408194820255e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3261
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.53125,
      "epoch": 0.15761499806725937,
      "grad_norm": 0.0,
      "learning_rate": 4.212166602241979e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 3262
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.6875,
      "epoch": 0.15766331658291458,
      "grad_norm": 0.0,
      "learning_rate": 4.2119250096637025e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3263
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.90625,
      "epoch": 0.15771163509856978,
      "grad_norm": 0.0,
      "learning_rate": 4.211683417085427e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3264
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.6875,
      "epoch": 0.15775995361422498,
      "grad_norm": 0.0,
      "learning_rate": 4.2114418245071507e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3265
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.96875,
      "epoch": 0.15780827212988016,
      "grad_norm": 0.0,
      "learning_rate": 4.211200231928875e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3266
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.25,
      "epoch": 0.15785659064553537,
      "grad_norm": 0.0,
      "learning_rate": 4.210958639350599e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3267
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 87.1875,
      "epoch": 0.15790490916119057,
      "grad_norm": 11.53342414474811,
      "learning_rate": 4.210717046772323e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3268
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.84375,
      "epoch": 0.15795322767684578,
      "grad_norm": 0.0,
      "learning_rate": 4.210475454194047e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3269
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.0625,
      "epoch": 0.15800154619250098,
      "grad_norm": 13.11087121473765,
      "learning_rate": 4.2102338616157714e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.22835706174373627,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3270
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.84375,
      "epoch": 0.15804986470815616,
      "grad_norm": 0.0,
      "learning_rate": 4.209992269037495e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3271
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.75,
      "epoch": 0.15809818322381136,
      "grad_norm": 0.0,
      "learning_rate": 4.2097506764592186e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3272
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.71875,
      "epoch": 0.15814650173946657,
      "grad_norm": 0.0,
      "learning_rate": 4.209509083880943e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3273
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.375,
      "epoch": 0.15819482025512177,
      "grad_norm": 0.0,
      "learning_rate": 4.209267491302667e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3274
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 79.90625,
      "epoch": 0.15824313877077698,
      "grad_norm": 8.791242719698346,
      "learning_rate": 4.209025898724391e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.23243674635887146,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3275
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.15625,
      "epoch": 0.15829145728643215,
      "grad_norm": 0.0,
      "learning_rate": 4.2087843061461146e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 3276
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.8125,
      "epoch": 0.15833977580208736,
      "grad_norm": 61.22491653563075,
      "learning_rate": 4.2085427135678393e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.25282126665115356,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3277
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.6875,
      "epoch": 0.15838809431774256,
      "grad_norm": 0.0,
      "learning_rate": 4.208301120989563e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3278
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.40625,
      "epoch": 0.15843641283339777,
      "grad_norm": 10.031109626065142,
      "learning_rate": 4.208059528411287e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.2993341088294983,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3279
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.53125,
      "epoch": 0.15848473134905297,
      "grad_norm": 12.01710804678983,
      "learning_rate": 4.207817935833011e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2883424758911133,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3280
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.71875,
      "epoch": 0.15853304986470815,
      "grad_norm": 0.0,
      "learning_rate": 4.2075763432547353e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3281
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.15625,
      "epoch": 0.15858136838036335,
      "grad_norm": 19.885687873664057,
      "learning_rate": 4.207334750676459e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3282
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.90625,
      "epoch": 0.15862968689601856,
      "grad_norm": 0.0,
      "learning_rate": 4.2070931580981825e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3283
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.125,
      "epoch": 0.15867800541167376,
      "grad_norm": 0.0,
      "learning_rate": 4.206851565519907e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3284
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 85.28125,
      "epoch": 0.15872632392732897,
      "grad_norm": 9.862277413854391,
      "learning_rate": 4.206609972941631e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3285
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.0,
      "epoch": 0.15877464244298414,
      "grad_norm": 0.0,
      "learning_rate": 4.206368380363355e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3286
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.90625,
      "epoch": 0.15882296095863935,
      "grad_norm": 0.0,
      "learning_rate": 4.206126787785079e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3287
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.21875,
      "epoch": 0.15887127947429455,
      "grad_norm": 0.0,
      "learning_rate": 4.205885195206803e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3288
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.3125,
      "epoch": 0.15891959798994976,
      "grad_norm": 20.063661699629073,
      "learning_rate": 4.205643602628527e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3289
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.0,
      "epoch": 0.15896791650560496,
      "grad_norm": 0.0,
      "learning_rate": 4.2054020100502515e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3290
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.125,
      "epoch": 0.15901623502126014,
      "grad_norm": 0.0,
      "learning_rate": 4.205160417471975e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3291
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.625,
      "epoch": 0.15906455353691534,
      "grad_norm": 0.0,
      "learning_rate": 4.204918824893699e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3292
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 85.53125,
      "epoch": 0.15911287205257055,
      "grad_norm": 9.470303453806991,
      "learning_rate": 4.204677232315423e-07,
      "loss": -0.0,
      "reward": 2.28125,
      "reward_std": 0.22152717411518097,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3293
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.84375,
      "epoch": 0.15916119056822575,
      "grad_norm": 0.0,
      "learning_rate": 4.204435639737147e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3294
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.96875,
      "epoch": 0.15920950908388096,
      "grad_norm": 12.330276632942626,
      "learning_rate": 4.204194047158871e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3295
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.25,
      "epoch": 0.15925782759953613,
      "grad_norm": 0.0,
      "learning_rate": 4.2039524545805947e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3296
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.4375,
      "epoch": 0.15930614611519134,
      "grad_norm": 0.0,
      "learning_rate": 4.2037108620023194e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3297
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.59375,
      "epoch": 0.15935446463084654,
      "grad_norm": 32.172429309219176,
      "learning_rate": 4.203469269424043e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3298
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 89.84375,
      "epoch": 0.15940278314650175,
      "grad_norm": 7.251893348493563,
      "learning_rate": 4.203227676845767e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3299
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.6875,
      "epoch": 0.15945110166215695,
      "grad_norm": 0.0,
      "learning_rate": 4.202986084267491e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3300
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.8125,
      "epoch": 0.15949942017781213,
      "grad_norm": 0.0,
      "learning_rate": 4.2027444916892154e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3301
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.40625,
      "epoch": 0.15954773869346733,
      "grad_norm": 0.0,
      "learning_rate": 4.202502899110939e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3302
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.96875,
      "epoch": 0.15959605720912254,
      "grad_norm": 0.0,
      "learning_rate": 4.2022613065326626e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3303
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.59375,
      "epoch": 0.15964437572477774,
      "grad_norm": 0.0,
      "learning_rate": 4.2020197139543873e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3304
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 94.5625,
      "epoch": 0.15969269424043295,
      "grad_norm": 49.172531226520576,
      "learning_rate": 4.201778121376111e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.25069957971572876,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3305
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.1875,
      "epoch": 0.15974101275608812,
      "grad_norm": 0.0,
      "learning_rate": 4.201536528797835e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3306
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.28125,
      "epoch": 0.15978933127174333,
      "grad_norm": 0.0,
      "learning_rate": 4.201294936219559e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3307
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.59375,
      "epoch": 0.15983764978739853,
      "grad_norm": 0.0,
      "learning_rate": 4.2010533436412833e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3308
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.0,
      "epoch": 0.15988596830305374,
      "grad_norm": 0.0,
      "learning_rate": 4.200811751063007e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 3309
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.5,
      "epoch": 0.15993428681870894,
      "grad_norm": 0.0,
      "learning_rate": 4.2005701584847316e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3310
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.875,
      "epoch": 0.15998260533436412,
      "grad_norm": 0.0,
      "learning_rate": 4.200328565906455e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3311
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.40625,
      "epoch": 0.16003092385001932,
      "grad_norm": 0.0,
      "learning_rate": 4.2000869733281793e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3312
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.09375,
      "epoch": 0.16007924236567453,
      "grad_norm": 0.0,
      "learning_rate": 4.199845380749903e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3313
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.09375,
      "epoch": 0.16012756088132973,
      "grad_norm": 0.0,
      "learning_rate": 4.199603788171627e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3314
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.21875,
      "epoch": 0.16017587939698494,
      "grad_norm": 0.0,
      "learning_rate": 4.199362195593351e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3315
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.59375,
      "epoch": 0.16022419791264012,
      "grad_norm": 0.0,
      "learning_rate": 4.199120603015075e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3316
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 114.9375,
      "epoch": 0.16027251642829532,
      "grad_norm": 65.93479996286224,
      "learning_rate": 4.1988790104367995e-07,
      "loss": -0.0,
      "reward": 1.625,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3317
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.5,
      "epoch": 0.16032083494395052,
      "grad_norm": 0.0,
      "learning_rate": 4.198637417858523e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3318
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.84375,
      "epoch": 0.16036915345960573,
      "grad_norm": 0.0,
      "learning_rate": 4.198395825280247e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3319
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.5625,
      "epoch": 0.16041747197526093,
      "grad_norm": 0.0,
      "learning_rate": 4.1981542327019713e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3320
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.09375,
      "epoch": 0.1604657904909161,
      "grad_norm": 0.0,
      "learning_rate": 4.1979126401236955e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3321
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 104.40625,
      "epoch": 0.16051410900657131,
      "grad_norm": 0.0,
      "learning_rate": 4.197671047545419e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 3322
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.03125,
      "epoch": 0.16056242752222652,
      "grad_norm": 0.0,
      "learning_rate": 4.1974294549671427e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3323
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.3125,
      "epoch": 0.16061074603788172,
      "grad_norm": 0.0,
      "learning_rate": 4.1971878623888674e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3324
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.59375,
      "epoch": 0.16065906455353693,
      "grad_norm": 0.0,
      "learning_rate": 4.196946269810591e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3325
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.8125,
      "epoch": 0.1607073830691921,
      "grad_norm": 0.0,
      "learning_rate": 4.196704677232315e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3326
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.03125,
      "epoch": 0.1607557015848473,
      "grad_norm": 0.0,
      "learning_rate": 4.196463084654039e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3327
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.8125,
      "epoch": 0.16080402010050251,
      "grad_norm": 0.0,
      "learning_rate": 4.1962214920757634e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3328
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.46875,
      "epoch": 0.16085233861615772,
      "grad_norm": 0.0,
      "learning_rate": 4.195979899497487e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3329
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.375,
      "epoch": 0.16090065713181292,
      "grad_norm": 0.0,
      "learning_rate": 4.1957383069192116e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3330
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.6875,
      "epoch": 0.1609489756474681,
      "grad_norm": 0.0,
      "learning_rate": 4.195496714340935e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3331
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.59375,
      "epoch": 0.1609972941631233,
      "grad_norm": 0.0,
      "learning_rate": 4.1952551217626594e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3332
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.84375,
      "epoch": 0.1610456126787785,
      "grad_norm": 0.0,
      "learning_rate": 4.195013529184383e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3333
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.84375,
      "epoch": 0.16109393119443371,
      "grad_norm": 0.0,
      "learning_rate": 4.194771936606107e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3334
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.9375,
      "epoch": 0.16114224971008892,
      "grad_norm": 0.0,
      "learning_rate": 4.1945303440278313e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3335
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 97.15625,
      "epoch": 0.1611905682257441,
      "grad_norm": 12.499274774919959,
      "learning_rate": 4.194288751449555e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.19999711215496063,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3336
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.46875,
      "epoch": 0.1612388867413993,
      "grad_norm": 0.0,
      "learning_rate": 4.1940471588712795e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3337
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 91.96875,
      "epoch": 0.1612872052570545,
      "grad_norm": 15.736546647426373,
      "learning_rate": 4.193805566293003e-07,
      "loss": -0.0,
      "reward": 2.28125,
      "reward_std": 0.28883427381515503,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3338
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 94.9375,
      "epoch": 0.1613355237727097,
      "grad_norm": 9.986944944244428,
      "learning_rate": 4.1935639737147273e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.23384949564933777,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3339
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.84375,
      "epoch": 0.16138384228836491,
      "grad_norm": 0.0,
      "learning_rate": 4.1933223811364514e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3340
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.0625,
      "epoch": 0.1614321608040201,
      "grad_norm": 0.0,
      "learning_rate": 4.1930807885581756e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3341
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.1875,
      "epoch": 0.1614804793196753,
      "grad_norm": 0.0,
      "learning_rate": 4.192839195979899e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3342
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.84375,
      "epoch": 0.1615287978353305,
      "grad_norm": 0.0,
      "learning_rate": 4.1925976034016233e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3343
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.40625,
      "epoch": 0.1615771163509857,
      "grad_norm": 0.0,
      "learning_rate": 4.1923560108233474e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3344
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.3125,
      "epoch": 0.1616254348666409,
      "grad_norm": 0.0,
      "learning_rate": 4.192114418245071e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3345
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.71875,
      "epoch": 0.1616737533822961,
      "grad_norm": 0.0,
      "learning_rate": 4.191872825666795e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3346
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.28125,
      "epoch": 0.1617220718979513,
      "grad_norm": 12.527102375740101,
      "learning_rate": 4.1916312330885193e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.30537301301956177,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3347
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.1875,
      "epoch": 0.1617703904136065,
      "grad_norm": 0.0,
      "learning_rate": 4.1913896405102435e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3348
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.15625,
      "epoch": 0.1618187089292617,
      "grad_norm": 0.0,
      "learning_rate": 4.191148047931967e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3349
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.21875,
      "epoch": 0.1618670274449169,
      "grad_norm": 0.0,
      "learning_rate": 4.1909064553536917e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3350
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.40625,
      "epoch": 0.16191534596057208,
      "grad_norm": 0.0,
      "learning_rate": 4.1906648627754153e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3351
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.0,
      "epoch": 0.1619636644762273,
      "grad_norm": 0.0,
      "learning_rate": 4.1904232701971395e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3352
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.03125,
      "epoch": 0.1620119829918825,
      "grad_norm": 0.0,
      "learning_rate": 4.190181677618863e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3353
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.875,
      "epoch": 0.1620603015075377,
      "grad_norm": 0.0,
      "learning_rate": 4.189940085040588e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3354
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.1875,
      "epoch": 0.1621086200231929,
      "grad_norm": 0.0,
      "learning_rate": 4.1896984924623113e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3355
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.125,
      "epoch": 0.16215693853884808,
      "grad_norm": 0.0,
      "learning_rate": 4.189456899884035e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3356
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.28125,
      "epoch": 0.16220525705450328,
      "grad_norm": 0.0,
      "learning_rate": 4.1892153073057596e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3357
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.9375,
      "epoch": 0.1622535755701585,
      "grad_norm": 0.0,
      "learning_rate": 4.188973714727483e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3358
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 89.09375,
      "epoch": 0.1623018940858137,
      "grad_norm": 0.0,
      "learning_rate": 4.1887321221492074e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 3359
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.25,
      "epoch": 0.1623502126014689,
      "grad_norm": 0.0,
      "learning_rate": 4.1884905295709315e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3360
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.09375,
      "epoch": 0.16239853111712407,
      "grad_norm": 0.0,
      "learning_rate": 4.1882489369926556e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3361
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.53125,
      "epoch": 0.16244684963277928,
      "grad_norm": 0.0,
      "learning_rate": 4.188007344414379e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3362
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.65625,
      "epoch": 0.16249516814843448,
      "grad_norm": 0.0,
      "learning_rate": 4.1877657518361034e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3363
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.96875,
      "epoch": 0.1625434866640897,
      "grad_norm": 0.0,
      "learning_rate": 4.1875241592578275e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.19727617502212524,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3364
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.4375,
      "epoch": 0.1625918051797449,
      "grad_norm": 0.0,
      "learning_rate": 4.187282566679551e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3365
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 91.34375,
      "epoch": 0.16264012369540007,
      "grad_norm": 5.613715173462165,
      "learning_rate": 4.187040974101275e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3366
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.21875,
      "epoch": 0.16268844221105527,
      "grad_norm": 0.0,
      "learning_rate": 4.1867993815229994e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3367
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.0,
      "epoch": 0.16273676072671048,
      "grad_norm": 0.0,
      "learning_rate": 4.1865577889447235e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3368
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.65625,
      "epoch": 0.16278507924236568,
      "grad_norm": 0.0,
      "learning_rate": 4.186316196366447e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3369
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.125,
      "epoch": 0.1628333977580209,
      "grad_norm": 0.0,
      "learning_rate": 4.186074603788172e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3370
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.46875,
      "epoch": 0.16288171627367606,
      "grad_norm": 0.0,
      "learning_rate": 4.1858330112098954e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3371
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.9375,
      "epoch": 0.16293003478933127,
      "grad_norm": 0.0,
      "learning_rate": 4.1855914186316195e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3372
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.15625,
      "epoch": 0.16297835330498647,
      "grad_norm": 0.0,
      "learning_rate": 4.185349826053343e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3373
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.53125,
      "epoch": 0.16302667182064168,
      "grad_norm": 0.0,
      "learning_rate": 4.185108233475068e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3374
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.03125,
      "epoch": 0.16307499033629688,
      "grad_norm": 0.0,
      "learning_rate": 4.1848666408967914e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3375
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.0625,
      "epoch": 0.16312330885195206,
      "grad_norm": 0.0,
      "learning_rate": 4.184625048318515e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3376
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.40625,
      "epoch": 0.16317162736760726,
      "grad_norm": 0.0,
      "learning_rate": 4.1843834557402397e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3377
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.78125,
      "epoch": 0.16321994588326247,
      "grad_norm": 0.0,
      "learning_rate": 4.1841418631619633e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3378
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.5,
      "epoch": 0.16326826439891767,
      "grad_norm": 0.0,
      "learning_rate": 4.1839002705836874e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3379
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.90625,
      "epoch": 0.16331658291457288,
      "grad_norm": 0.0,
      "learning_rate": 4.1836586780054116e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 3380
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.34375,
      "epoch": 0.16336490143022805,
      "grad_norm": 0.0,
      "learning_rate": 4.1834170854271357e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3381
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.65625,
      "epoch": 0.16341321994588326,
      "grad_norm": 0.0,
      "learning_rate": 4.1831754928488593e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3382
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.78125,
      "epoch": 0.16346153846153846,
      "grad_norm": 0.0,
      "learning_rate": 4.1829339002705835e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3383
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.4375,
      "epoch": 0.16350985697719367,
      "grad_norm": 0.0,
      "learning_rate": 4.1826923076923076e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3384
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.8125,
      "epoch": 0.16355817549284887,
      "grad_norm": 0.0,
      "learning_rate": 4.1824507151140317e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3385
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.625,
      "epoch": 0.16360649400850405,
      "grad_norm": 0.0,
      "learning_rate": 4.1822091225357553e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3386
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.53125,
      "epoch": 0.16365481252415925,
      "grad_norm": 0.0,
      "learning_rate": 4.1819675299574795e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3387
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.0,
      "epoch": 0.16370313103981446,
      "grad_norm": 0.0,
      "learning_rate": 4.1817259373792036e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3388
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.15625,
      "epoch": 0.16375144955546966,
      "grad_norm": 33.80501267336822,
      "learning_rate": 4.181484344800927e-07,
      "loss": -0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3389
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 87.375,
      "epoch": 0.16379976807112487,
      "grad_norm": 19.248213100022646,
      "learning_rate": 4.181242752222652e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3390
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.34375,
      "epoch": 0.16384808658678005,
      "grad_norm": 0.0,
      "learning_rate": 4.1810011596443755e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3391
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.875,
      "epoch": 0.16389640510243525,
      "grad_norm": 0.0,
      "learning_rate": 4.1807595670660996e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3392
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.75,
      "epoch": 0.16394472361809045,
      "grad_norm": 0.0,
      "learning_rate": 4.180517974487823e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3393
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.78125,
      "epoch": 0.16399304213374566,
      "grad_norm": 0.0,
      "learning_rate": 4.180276381909548e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3394
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.9375,
      "epoch": 0.16404136064940086,
      "grad_norm": 0.0,
      "learning_rate": 4.1800347893312715e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3395
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.8125,
      "epoch": 0.16408967916505604,
      "grad_norm": 6.845041952800681,
      "learning_rate": 4.179793196752995e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3396
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.59375,
      "epoch": 0.16413799768071125,
      "grad_norm": 0.0,
      "learning_rate": 4.17955160417472e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3397
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.96875,
      "epoch": 0.16418631619636645,
      "grad_norm": 20.22531831081939,
      "learning_rate": 4.1793100115964434e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.26409149169921875,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3398
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.25,
      "epoch": 0.16423463471202165,
      "grad_norm": 0.0,
      "learning_rate": 4.1790684190181675e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 3399
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.90625,
      "epoch": 0.16428295322767686,
      "grad_norm": 0.0,
      "learning_rate": 4.1788268264398917e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3400
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 85.9375,
      "epoch": 0.16433127174333204,
      "grad_norm": 19.22700584613731,
      "learning_rate": 4.178585233861616e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3401
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.625,
      "epoch": 0.16437959025898724,
      "grad_norm": 0.0,
      "learning_rate": 4.1783436412833394e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3402
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.6875,
      "epoch": 0.16442790877464245,
      "grad_norm": 0.0,
      "learning_rate": 4.1781020487050635e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3403
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.8125,
      "epoch": 0.16447622729029765,
      "grad_norm": 0.0,
      "learning_rate": 4.1778604561267877e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3404
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.0625,
      "epoch": 0.16452454580595285,
      "grad_norm": 0.0,
      "learning_rate": 4.177618863548512e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3405
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.125,
      "epoch": 0.16457286432160803,
      "grad_norm": 0.0,
      "learning_rate": 4.1773772709702354e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3406
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 89.59375,
      "epoch": 0.16462118283726324,
      "grad_norm": 0.0,
      "learning_rate": 4.1771356783919596e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 3407
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.25,
      "epoch": 0.16466950135291844,
      "grad_norm": 0.0,
      "learning_rate": 4.1768940858136837e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3408
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.40625,
      "epoch": 0.16471781986857364,
      "grad_norm": 0.0,
      "learning_rate": 4.1766524932354073e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 3409
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.4375,
      "epoch": 0.16476613838422885,
      "grad_norm": 0.0,
      "learning_rate": 4.176410900657132e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3410
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.71875,
      "epoch": 0.16481445689988403,
      "grad_norm": 0.0,
      "learning_rate": 4.1761693080788556e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3411
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.625,
      "epoch": 0.16486277541553923,
      "grad_norm": 0.0,
      "learning_rate": 4.1759277155005797e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3412
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.125,
      "epoch": 0.16491109393119444,
      "grad_norm": 0.0,
      "learning_rate": 4.1756861229223033e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3413
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.6875,
      "epoch": 0.16495941244684964,
      "grad_norm": 10.053727873894859,
      "learning_rate": 4.175444530344028e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3414
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.34375,
      "epoch": 0.16500773096250484,
      "grad_norm": 0.0,
      "learning_rate": 4.1752029377657516e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3415
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 85.71875,
      "epoch": 0.16505604947816002,
      "grad_norm": 17.933368327636835,
      "learning_rate": 4.174961345187475e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.27841314673423767,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3416
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.375,
      "epoch": 0.16510436799381523,
      "grad_norm": 0.0,
      "learning_rate": 4.1747197526092e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3417
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 94.125,
      "epoch": 0.16515268650947043,
      "grad_norm": 11.215564420659245,
      "learning_rate": 4.1744781600309235e-07,
      "loss": 0.0,
      "reward": 1.140625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.09375,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 3418
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.9375,
      "epoch": 0.16520100502512564,
      "grad_norm": 0.0,
      "learning_rate": 4.1742365674526476e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3419
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 95.59375,
      "epoch": 0.16524932354078084,
      "grad_norm": 16.961631791276268,
      "learning_rate": 4.173994974874372e-07,
      "loss": 0.0,
      "reward": 1.53125,
      "reward_std": 0.24067938327789307,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3420
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.40625,
      "epoch": 0.16529764205643602,
      "grad_norm": 0.0,
      "learning_rate": 4.173753382296096e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3421
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.40625,
      "epoch": 0.16534596057209122,
      "grad_norm": 0.0,
      "learning_rate": 4.1735117897178195e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3422
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.5625,
      "epoch": 0.16539427908774643,
      "grad_norm": 0.0,
      "learning_rate": 4.1732701971395436e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3423
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.5625,
      "epoch": 0.16544259760340163,
      "grad_norm": 0.0,
      "learning_rate": 4.173028604561268e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 3424
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.03125,
      "epoch": 0.16549091611905684,
      "grad_norm": 0.0,
      "learning_rate": 4.172787011982992e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3425
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.25,
      "epoch": 0.165539234634712,
      "grad_norm": 0.0,
      "learning_rate": 4.1725454194047155e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3426
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 84.125,
      "epoch": 0.16558755315036722,
      "grad_norm": 24.526478205778453,
      "learning_rate": 4.1723038268264396e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.23634064197540283,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3427
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.09375,
      "epoch": 0.16563587166602242,
      "grad_norm": 0.0,
      "learning_rate": 4.172062234248164e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3428
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.125,
      "epoch": 0.16568419018167763,
      "grad_norm": 0.0,
      "learning_rate": 4.1718206416698874e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3429
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.5625,
      "epoch": 0.16573250869733283,
      "grad_norm": 0.0,
      "learning_rate": 4.171579049091612e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3430
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.78125,
      "epoch": 0.165780827212988,
      "grad_norm": 21.12992571241681,
      "learning_rate": 4.1713374565133356e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3431
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.4375,
      "epoch": 0.1658291457286432,
      "grad_norm": 0.0,
      "learning_rate": 4.17109586393506e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3432
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.4375,
      "epoch": 0.16587746424429842,
      "grad_norm": 0.0,
      "learning_rate": 4.1708542713567834e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3433
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.1875,
      "epoch": 0.16592578275995362,
      "grad_norm": 0.0,
      "learning_rate": 4.170612678778508e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3434
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.40625,
      "epoch": 0.16597410127560883,
      "grad_norm": 0.0,
      "learning_rate": 4.1703710862002317e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3435
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.40625,
      "epoch": 0.166022419791264,
      "grad_norm": 0.0,
      "learning_rate": 4.170129493621956e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3436
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 87.0,
      "epoch": 0.1660707383069192,
      "grad_norm": 10.660083458299741,
      "learning_rate": 4.16988790104368e-07,
      "loss": -0.0,
      "reward": 1.546875,
      "reward_std": 0.2608962655067444,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3437
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.15625,
      "epoch": 0.1661190568225744,
      "grad_norm": 0.0,
      "learning_rate": 4.1696463084654035e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3438
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.875,
      "epoch": 0.16616737533822962,
      "grad_norm": 0.0,
      "learning_rate": 4.1694047158871277e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3439
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.0625,
      "epoch": 0.16621569385388482,
      "grad_norm": 0.0,
      "learning_rate": 4.169163123308852e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3440
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.875,
      "epoch": 0.16626401236954,
      "grad_norm": 0.0,
      "learning_rate": 4.168921530730576e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3441
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.78125,
      "epoch": 0.1663123308851952,
      "grad_norm": 0.0,
      "learning_rate": 4.1686799381522996e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3442
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.03125,
      "epoch": 0.1663606494008504,
      "grad_norm": 0.0,
      "learning_rate": 4.1684383455740237e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3443
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.96875,
      "epoch": 0.1664089679165056,
      "grad_norm": 0.0,
      "learning_rate": 4.168196752995748e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3444
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.625,
      "epoch": 0.16645728643216082,
      "grad_norm": 7.354926631391436,
      "learning_rate": 4.167955160417472e-07,
      "loss": 0.0,
      "reward": 2.1875,
      "reward_std": 0.3653584420681,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3445
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.4375,
      "epoch": 0.166505604947816,
      "grad_norm": 0.0,
      "learning_rate": 4.1677135678391956e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3446
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.4375,
      "epoch": 0.1665539234634712,
      "grad_norm": 0.0,
      "learning_rate": 4.16747197526092e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3447
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 85.625,
      "epoch": 0.1666022419791264,
      "grad_norm": 68.11163271698632,
      "learning_rate": 4.167230382682644e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3448
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.875,
      "epoch": 0.1666505604947816,
      "grad_norm": 0.0,
      "learning_rate": 4.1669887901043675e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3449
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.28125,
      "epoch": 0.1666988790104368,
      "grad_norm": 0.0,
      "learning_rate": 4.166747197526092e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3450
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.28125,
      "epoch": 0.166747197526092,
      "grad_norm": 0.0,
      "learning_rate": 4.1665056049478157e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3451
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.375,
      "epoch": 0.1667955160417472,
      "grad_norm": 0.0,
      "learning_rate": 4.16626401236954e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3452
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.78125,
      "epoch": 0.1668438345574024,
      "grad_norm": 0.0,
      "learning_rate": 4.1660224197912635e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3453
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.9375,
      "epoch": 0.1668921530730576,
      "grad_norm": 0.0,
      "learning_rate": 4.165780827212988e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3454
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.125,
      "epoch": 0.1669404715887128,
      "grad_norm": 0.0,
      "learning_rate": 4.165539234634712e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3455
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.65625,
      "epoch": 0.16698879010436798,
      "grad_norm": 0.0,
      "learning_rate": 4.165297642056436e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3456
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 80.3125,
      "epoch": 0.1670371086200232,
      "grad_norm": 12.465752915743112,
      "learning_rate": 4.16505604947816e-07,
      "loss": -0.0,
      "reward": 1.6875,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3457
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.53125,
      "epoch": 0.1670854271356784,
      "grad_norm": 0.0,
      "learning_rate": 4.1648144568998836e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3458
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.75,
      "epoch": 0.1671337456513336,
      "grad_norm": 0.0,
      "learning_rate": 4.164572864321608e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 3459
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.0625,
      "epoch": 0.1671820641669888,
      "grad_norm": 0.0,
      "learning_rate": 4.164331271743332e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3460
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.71875,
      "epoch": 0.16723038268264398,
      "grad_norm": 0.0,
      "learning_rate": 4.164089679165056e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3461
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.3125,
      "epoch": 0.16727870119829918,
      "grad_norm": 0.0,
      "learning_rate": 4.1638480865867796e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3462
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.9375,
      "epoch": 0.1673270197139544,
      "grad_norm": 0.0,
      "learning_rate": 4.163606494008504e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3463
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.96875,
      "epoch": 0.1673753382296096,
      "grad_norm": 0.0,
      "learning_rate": 4.163364901430228e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3464
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.0,
      "epoch": 0.1674236567452648,
      "grad_norm": 0.0,
      "learning_rate": 4.163123308851952e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3465
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.25,
      "epoch": 0.16747197526091998,
      "grad_norm": 0.0,
      "learning_rate": 4.1628817162736757e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3466
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.40625,
      "epoch": 0.16752029377657518,
      "grad_norm": 0.0,
      "learning_rate": 4.1626401236954003e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3467
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.6875,
      "epoch": 0.16756861229223038,
      "grad_norm": 0.0,
      "learning_rate": 4.162398531117124e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3468
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.90625,
      "epoch": 0.1676169308078856,
      "grad_norm": 0.0,
      "learning_rate": 4.1621569385388475e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3469
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.1875,
      "epoch": 0.1676652493235408,
      "grad_norm": 0.0,
      "learning_rate": 4.161915345960572e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3470
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.5625,
      "epoch": 0.16771356783919597,
      "grad_norm": 0.0,
      "learning_rate": 4.161673753382296e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3471
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.09375,
      "epoch": 0.16776188635485118,
      "grad_norm": 0.0,
      "learning_rate": 4.16143216080402e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3472
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.0,
      "epoch": 0.16781020487050638,
      "grad_norm": 18.681233711736798,
      "learning_rate": 4.1611905682257435e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.23685520887374878,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3473
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.3125,
      "epoch": 0.16785852338616158,
      "grad_norm": 9.20686329949192,
      "learning_rate": 4.160948975647468e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3474
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 85.9375,
      "epoch": 0.1679068419018168,
      "grad_norm": 27.650555097073926,
      "learning_rate": 4.160707383069192e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.3845272362232208,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3475
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.40625,
      "epoch": 0.16795516041747197,
      "grad_norm": 0.0,
      "learning_rate": 4.160465790490916e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3476
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.8125,
      "epoch": 0.16800347893312717,
      "grad_norm": 0.0,
      "learning_rate": 4.16022419791264e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3477
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 81.3125,
      "epoch": 0.16805179744878238,
      "grad_norm": 0.0,
      "learning_rate": 4.1599826053343637e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 3478
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.46875,
      "epoch": 0.16810011596443758,
      "grad_norm": 0.0,
      "learning_rate": 4.159741012756088e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3479
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.40625,
      "epoch": 0.16814843448009278,
      "grad_norm": 0.0,
      "learning_rate": 4.159499420177812e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3480
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.875,
      "epoch": 0.16819675299574796,
      "grad_norm": 0.0,
      "learning_rate": 4.159257827599536e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3481
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.125,
      "epoch": 0.16824507151140317,
      "grad_norm": 0.0,
      "learning_rate": 4.1590162350212597e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 3482
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.40625,
      "epoch": 0.16829339002705837,
      "grad_norm": 0.0,
      "learning_rate": 4.158774642442984e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3483
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.78125,
      "epoch": 0.16834170854271358,
      "grad_norm": 0.0,
      "learning_rate": 4.158533049864708e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3484
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.96875,
      "epoch": 0.16839002705836878,
      "grad_norm": 0.0,
      "learning_rate": 4.158291457286432e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3485
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.71875,
      "epoch": 0.16843834557402396,
      "grad_norm": 0.0,
      "learning_rate": 4.1580498647081557e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3486
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.875,
      "epoch": 0.16848666408967916,
      "grad_norm": 0.0,
      "learning_rate": 4.1578082721298804e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3487
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.96875,
      "epoch": 0.16853498260533437,
      "grad_norm": 0.0,
      "learning_rate": 4.157566679551604e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3488
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.375,
      "epoch": 0.16858330112098957,
      "grad_norm": 0.0,
      "learning_rate": 4.1573250869733276e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3489
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.9375,
      "epoch": 0.16863161963664478,
      "grad_norm": 0.0,
      "learning_rate": 4.1570834943950523e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3490
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.75,
      "epoch": 0.16867993815229995,
      "grad_norm": 0.0,
      "learning_rate": 4.156841901816776e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3491
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.0,
      "epoch": 0.16872825666795516,
      "grad_norm": 0.0,
      "learning_rate": 4.1566003092385e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3492
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.75,
      "epoch": 0.16877657518361036,
      "grad_norm": 0.0,
      "learning_rate": 4.1563587166602236e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880052089691,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3493
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.3125,
      "epoch": 0.16882489369926557,
      "grad_norm": 0.0,
      "learning_rate": 4.1561171240819483e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3494
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.5,
      "epoch": 0.16887321221492077,
      "grad_norm": 0.0,
      "learning_rate": 4.155875531503672e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3495
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.21875,
      "epoch": 0.16892153073057595,
      "grad_norm": 0.0,
      "learning_rate": 4.155633938925396e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3496
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 83.9375,
      "epoch": 0.16896984924623115,
      "grad_norm": 15.361128633130162,
      "learning_rate": 4.15539234634712e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.3446694612503052,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3497
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.625,
      "epoch": 0.16901816776188636,
      "grad_norm": 0.0,
      "learning_rate": 4.1551507537688443e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3498
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.25,
      "epoch": 0.16906648627754156,
      "grad_norm": 0.0,
      "learning_rate": 4.154909161190568e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3499
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.09375,
      "epoch": 0.16911480479319677,
      "grad_norm": 0.0,
      "learning_rate": 4.154667568612292e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3500
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.03125,
      "epoch": 0.16916312330885194,
      "grad_norm": 0.0,
      "learning_rate": 4.154425976034016e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3501
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.53125,
      "epoch": 0.16921144182450715,
      "grad_norm": 0.0,
      "learning_rate": 4.15418438345574e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 3502
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.15625,
      "epoch": 0.16925976034016235,
      "grad_norm": 0.0,
      "learning_rate": 4.153942790877464e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 3503
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.25,
      "epoch": 0.16930807885581756,
      "grad_norm": 0.0,
      "learning_rate": 4.153701198299188e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3504
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.375,
      "epoch": 0.16935639737147276,
      "grad_norm": 0.0,
      "learning_rate": 4.153459605720912e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3505
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.09375,
      "epoch": 0.16940471588712794,
      "grad_norm": 0.0,
      "learning_rate": 4.153218013142636e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3506
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.84375,
      "epoch": 0.16945303440278314,
      "grad_norm": 0.0,
      "learning_rate": 4.1529764205643605e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3507
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.0,
      "epoch": 0.16950135291843835,
      "grad_norm": 0.0,
      "learning_rate": 4.152734827986084e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3508
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.75,
      "epoch": 0.16954967143409355,
      "grad_norm": 0.0,
      "learning_rate": 4.1524932354078077e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3509
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.75,
      "epoch": 0.16959798994974876,
      "grad_norm": 0.0,
      "learning_rate": 4.1522516428295324e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3510
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.40625,
      "epoch": 0.16964630846540393,
      "grad_norm": 0.0,
      "learning_rate": 4.152010050251256e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3511
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.0,
      "epoch": 0.16969462698105914,
      "grad_norm": 29.872386601129058,
      "learning_rate": 4.15176845767298e-07,
      "loss": -0.0,
      "reward": 2.28125,
      "reward_std": 0.25282126665115356,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3512
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.53125,
      "epoch": 0.16974294549671434,
      "grad_norm": 0.0,
      "learning_rate": 4.1515268650947037e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3513
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.03125,
      "epoch": 0.16979126401236955,
      "grad_norm": 19.13875415172574,
      "learning_rate": 4.1512852725164284e-07,
      "loss": -0.0,
      "reward": 2.421875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3514
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.1875,
      "epoch": 0.16983958252802475,
      "grad_norm": 0.0,
      "learning_rate": 4.151043679938152e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3515
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.40625,
      "epoch": 0.16988790104367993,
      "grad_norm": 0.0,
      "learning_rate": 4.150802087359876e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3516
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.96875,
      "epoch": 0.16993621955933513,
      "grad_norm": 0.0,
      "learning_rate": 4.1505604947816e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3517
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.375,
      "epoch": 0.16998453807499034,
      "grad_norm": 0.0,
      "learning_rate": 4.1503189022033244e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3518
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.3125,
      "epoch": 0.17003285659064554,
      "grad_norm": 0.0,
      "learning_rate": 4.150077309625048e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3519
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.8125,
      "epoch": 0.17008117510630075,
      "grad_norm": 0.0,
      "learning_rate": 4.149835717046772e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3520
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.21875,
      "epoch": 0.17012949362195592,
      "grad_norm": 0.0,
      "learning_rate": 4.1495941244684963e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3521
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.125,
      "epoch": 0.17017781213761113,
      "grad_norm": 0.0,
      "learning_rate": 4.14935253189022e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3522
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.90625,
      "epoch": 0.17022613065326633,
      "grad_norm": 12.591004613949144,
      "learning_rate": 4.149110939311944e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.18857255578041077,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3523
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 81.0,
      "epoch": 0.17027444916892154,
      "grad_norm": 42.63999598058067,
      "learning_rate": 4.148869346733668e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3524
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.09375,
      "epoch": 0.17032276768457674,
      "grad_norm": 0.0,
      "learning_rate": 4.1486277541553923e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3525
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.75,
      "epoch": 0.17037108620023192,
      "grad_norm": 0.0,
      "learning_rate": 4.148386161577116e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3526
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.40625,
      "epoch": 0.17041940471588712,
      "grad_norm": 0.0,
      "learning_rate": 4.1481445689988406e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3527
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.46875,
      "epoch": 0.17046772323154233,
      "grad_norm": 0.0,
      "learning_rate": 4.147902976420564e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3528
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.78125,
      "epoch": 0.17051604174719753,
      "grad_norm": 0.0,
      "learning_rate": 4.147661383842288e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3529
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.5625,
      "epoch": 0.17056436026285274,
      "grad_norm": 0.0,
      "learning_rate": 4.1474197912640124e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3530
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.875,
      "epoch": 0.17061267877850791,
      "grad_norm": 0.0,
      "learning_rate": 4.147178198685736e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3531
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.46875,
      "epoch": 0.17066099729416312,
      "grad_norm": 0.0,
      "learning_rate": 4.14693660610746e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3532
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.75,
      "epoch": 0.17070931580981832,
      "grad_norm": 0.0,
      "learning_rate": 4.146695013529184e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3533
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.5,
      "epoch": 0.17075763432547353,
      "grad_norm": 0.0,
      "learning_rate": 4.1464534209509084e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3534
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.78125,
      "epoch": 0.17080595284112873,
      "grad_norm": 0.0,
      "learning_rate": 4.146211828372632e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3535
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.71875,
      "epoch": 0.1708542713567839,
      "grad_norm": 0.0,
      "learning_rate": 4.145970235794356e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3536
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.6875,
      "epoch": 0.17090258987243911,
      "grad_norm": 12.129504553959224,
      "learning_rate": 4.1457286432160803e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3537
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.0,
      "epoch": 0.17095090838809432,
      "grad_norm": 0.0,
      "learning_rate": 4.1454870506378045e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3538
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.59375,
      "epoch": 0.17099922690374952,
      "grad_norm": 0.0,
      "learning_rate": 4.145245458059528e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3539
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.5625,
      "epoch": 0.17104754541940473,
      "grad_norm": 0.0,
      "learning_rate": 4.145003865481252e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3540
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.71875,
      "epoch": 0.1710958639350599,
      "grad_norm": 0.0,
      "learning_rate": 4.1447622729029763e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3541
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.375,
      "epoch": 0.1711441824507151,
      "grad_norm": 0.0,
      "learning_rate": 4.1445206803247e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3542
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.25,
      "epoch": 0.17119250096637031,
      "grad_norm": 0.0,
      "learning_rate": 4.144279087746424e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3543
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.875,
      "epoch": 0.17124081948202552,
      "grad_norm": 0.0,
      "learning_rate": 4.144037495168148e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3544
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.46875,
      "epoch": 0.17128913799768072,
      "grad_norm": 0.0,
      "learning_rate": 4.1437959025898724e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3545
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.65625,
      "epoch": 0.1713374565133359,
      "grad_norm": 0.0,
      "learning_rate": 4.143554310011596e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3546
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.96875,
      "epoch": 0.1713857750289911,
      "grad_norm": 0.0,
      "learning_rate": 4.1433127174333206e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3547
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.15625,
      "epoch": 0.1714340935446463,
      "grad_norm": 0.0,
      "learning_rate": 4.143071124855044e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3548
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.5,
      "epoch": 0.17148241206030151,
      "grad_norm": 0.0,
      "learning_rate": 4.1428295322767684e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 3549
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.28125,
      "epoch": 0.17153073057595672,
      "grad_norm": 0.0,
      "learning_rate": 4.1425879396984925e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3550
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.9375,
      "epoch": 0.1715790490916119,
      "grad_norm": 0.0,
      "learning_rate": 4.142346347120216e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3551
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.125,
      "epoch": 0.1716273676072671,
      "grad_norm": 0.0,
      "learning_rate": 4.14210475454194e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3552
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.5,
      "epoch": 0.1716756861229223,
      "grad_norm": 0.0,
      "learning_rate": 4.141863161963664e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3553
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.25,
      "epoch": 0.1717240046385775,
      "grad_norm": 0.0,
      "learning_rate": 4.1416215693853885e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3554
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.1875,
      "epoch": 0.17177232315423271,
      "grad_norm": 0.0,
      "learning_rate": 4.141379976807112e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3555
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.8125,
      "epoch": 0.1718206416698879,
      "grad_norm": 0.0,
      "learning_rate": 4.1411383842288363e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3556
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.09375,
      "epoch": 0.1718689601855431,
      "grad_norm": 0.0,
      "learning_rate": 4.1408967916505604e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3557
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.15625,
      "epoch": 0.1719172787011983,
      "grad_norm": 0.0,
      "learning_rate": 4.1406551990722845e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3558
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.4375,
      "epoch": 0.1719655972168535,
      "grad_norm": 0.0,
      "learning_rate": 4.140413606494008e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3559
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.34375,
      "epoch": 0.1720139157325087,
      "grad_norm": 0.0,
      "learning_rate": 4.140172013915733e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3560
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.15625,
      "epoch": 0.1720622342481639,
      "grad_norm": 0.0,
      "learning_rate": 4.1399304213374564e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3561
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 83.84375,
      "epoch": 0.1721105527638191,
      "grad_norm": 0.0,
      "learning_rate": 4.13968882875918e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 3562
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 71.0625,
      "epoch": 0.1721588712794743,
      "grad_norm": 10.744619798788902,
      "learning_rate": 4.139447236180904e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3563
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.21875,
      "epoch": 0.1722071897951295,
      "grad_norm": 0.0,
      "learning_rate": 4.1392056436026283e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3564
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.25,
      "epoch": 0.1722555083107847,
      "grad_norm": 0.0,
      "learning_rate": 4.1389640510243524e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3565
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.84375,
      "epoch": 0.17230382682643988,
      "grad_norm": 0.0,
      "learning_rate": 4.138722458446076e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3566
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.84375,
      "epoch": 0.1723521453420951,
      "grad_norm": 0.0,
      "learning_rate": 4.1384808658678007e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3567
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.21875,
      "epoch": 0.1724004638577503,
      "grad_norm": 0.0,
      "learning_rate": 4.1382392732895243e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 3568
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.3125,
      "epoch": 0.1724487823734055,
      "grad_norm": 0.0,
      "learning_rate": 4.1379976807112485e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3569
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.125,
      "epoch": 0.1724971008890607,
      "grad_norm": 0.0,
      "learning_rate": 4.1377560881329726e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3570
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.375,
      "epoch": 0.17254541940471588,
      "grad_norm": 0.0,
      "learning_rate": 4.137514495554696e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3571
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.375,
      "epoch": 0.17259373792037108,
      "grad_norm": 0.0,
      "learning_rate": 4.1372729029764203e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3572
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.625,
      "epoch": 0.1726420564360263,
      "grad_norm": 0.0,
      "learning_rate": 4.137031310398144e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3573
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.3125,
      "epoch": 0.1726903749516815,
      "grad_norm": 0.0,
      "learning_rate": 4.1367897178198686e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3574
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.53125,
      "epoch": 0.1727386934673367,
      "grad_norm": 0.0,
      "learning_rate": 4.136548125241592e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3575
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.0625,
      "epoch": 0.17278701198299187,
      "grad_norm": 0.0,
      "learning_rate": 4.1363065326633163e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3576
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.25,
      "epoch": 0.17283533049864708,
      "grad_norm": 0.0,
      "learning_rate": 4.1360649400850405e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3577
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.5,
      "epoch": 0.17288364901430228,
      "grad_norm": 0.0,
      "learning_rate": 4.1358233475067646e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3578
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.90625,
      "epoch": 0.1729319675299575,
      "grad_norm": 0.0,
      "learning_rate": 4.135581754928488e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3579
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.03125,
      "epoch": 0.1729802860456127,
      "grad_norm": 0.0,
      "learning_rate": 4.135340162350213e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3580
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.0,
      "epoch": 0.17302860456126787,
      "grad_norm": 0.0,
      "learning_rate": 4.1350985697719365e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3581
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.3125,
      "epoch": 0.17307692307692307,
      "grad_norm": 0.0,
      "learning_rate": 4.13485697719366e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.260896235704422,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3582
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.84375,
      "epoch": 0.17312524159257828,
      "grad_norm": 0.0,
      "learning_rate": 4.134615384615384e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3583
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.78125,
      "epoch": 0.17317356010823348,
      "grad_norm": 0.0,
      "learning_rate": 4.1343737920371084e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3584
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.5625,
      "epoch": 0.1732218786238887,
      "grad_norm": 0.0,
      "learning_rate": 4.1341321994588325e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3585
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.6875,
      "epoch": 0.17327019713954386,
      "grad_norm": 0.0,
      "learning_rate": 4.133890606880556e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3586
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 72.3125,
      "epoch": 0.17331851565519907,
      "grad_norm": 0.0,
      "learning_rate": 4.133649014302281e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3587
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 89.375,
      "epoch": 0.17336683417085427,
      "grad_norm": 9.577468387396875,
      "learning_rate": 4.1334074217240044e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3588
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.1875,
      "epoch": 0.17341515268650948,
      "grad_norm": 0.0,
      "learning_rate": 4.1331658291457285e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3589
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.46875,
      "epoch": 0.17346347120216468,
      "grad_norm": 0.0,
      "learning_rate": 4.1329242365674527e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3590
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.34375,
      "epoch": 0.17351178971781986,
      "grad_norm": 0.0,
      "learning_rate": 4.132682643989177e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3591
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.6875,
      "epoch": 0.17356010823347506,
      "grad_norm": 0.0,
      "learning_rate": 4.1324410514109004e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3592
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 76.0,
      "epoch": 0.17360842674913027,
      "grad_norm": 31.325927516751,
      "learning_rate": 4.132199458832624e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.24359199404716492,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3593
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.6875,
      "epoch": 0.17365674526478547,
      "grad_norm": 0.0,
      "learning_rate": 4.1319578662543487e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3594
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.15625,
      "epoch": 0.17370506378044068,
      "grad_norm": 0.0,
      "learning_rate": 4.1317162736760723e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3595
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.25,
      "epoch": 0.17375338229609585,
      "grad_norm": 0.0,
      "learning_rate": 4.1314746810977964e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3596
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.53125,
      "epoch": 0.17380170081175106,
      "grad_norm": 0.0,
      "learning_rate": 4.1312330885195206e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3597
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.34375,
      "epoch": 0.17385001932740626,
      "grad_norm": 0.0,
      "learning_rate": 4.1309914959412447e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3598
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.53125,
      "epoch": 0.17389833784306147,
      "grad_norm": 0.0,
      "learning_rate": 4.1307499033629683e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3599
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.9375,
      "epoch": 0.17394665635871667,
      "grad_norm": 0.0,
      "learning_rate": 4.130508310784693e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3600
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.65625,
      "epoch": 0.17399497487437185,
      "grad_norm": 0.0,
      "learning_rate": 4.1302667182064166e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3601
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.84375,
      "epoch": 0.17404329339002705,
      "grad_norm": 0.0,
      "learning_rate": 4.13002512562814e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 3602
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 80.125,
      "epoch": 0.17409161190568226,
      "grad_norm": 11.321728324256116,
      "learning_rate": 4.1297835330498643e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.23827512562274933,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3603
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.09375,
      "epoch": 0.17413993042133746,
      "grad_norm": 0.0,
      "learning_rate": 4.1295419404715885e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3604
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 70.5,
      "epoch": 0.17418824893699267,
      "grad_norm": 17.5910207360874,
      "learning_rate": 4.1293003478933126e-07,
      "loss": -0.0,
      "reward": 1.625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3605
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.34375,
      "epoch": 0.17423656745264784,
      "grad_norm": 0.0,
      "learning_rate": 4.129058755315036e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3606
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.625,
      "epoch": 0.17428488596830305,
      "grad_norm": 0.0,
      "learning_rate": 4.128817162736761e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3607
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.625,
      "epoch": 0.17433320448395825,
      "grad_norm": 0.0,
      "learning_rate": 4.1285755701584845e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3608
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.1875,
      "epoch": 0.17438152299961346,
      "grad_norm": 0.0,
      "learning_rate": 4.1283339775802086e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3609
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.75,
      "epoch": 0.17442984151526866,
      "grad_norm": 0.0,
      "learning_rate": 4.128092385001933e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3610
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.40625,
      "epoch": 0.17447816003092384,
      "grad_norm": 0.0,
      "learning_rate": 4.127850792423657e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3611
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.65625,
      "epoch": 0.17452647854657904,
      "grad_norm": 0.0,
      "learning_rate": 4.1276091998453805e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3612
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.90625,
      "epoch": 0.17457479706223425,
      "grad_norm": 0.0,
      "learning_rate": 4.127367607267104e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3613
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.40625,
      "epoch": 0.17462311557788945,
      "grad_norm": 0.0,
      "learning_rate": 4.127126014688829e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3614
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.875,
      "epoch": 0.17467143409354466,
      "grad_norm": 0.0,
      "learning_rate": 4.1268844221105524e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3615
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.78125,
      "epoch": 0.17471975260919984,
      "grad_norm": 20.98296379257055,
      "learning_rate": 4.1266428295322765e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3616
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.21875,
      "epoch": 0.17476807112485504,
      "grad_norm": 0.0,
      "learning_rate": 4.1264012369540006e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3617
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.8125,
      "epoch": 0.17481638964051024,
      "grad_norm": 0.0,
      "learning_rate": 4.126159644375725e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3618
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.15625,
      "epoch": 0.17486470815616545,
      "grad_norm": 0.0,
      "learning_rate": 4.1259180517974484e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3619
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.3125,
      "epoch": 0.17491302667182065,
      "grad_norm": 0.0,
      "learning_rate": 4.125676459219173e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3620
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.96875,
      "epoch": 0.17496134518747583,
      "grad_norm": 0.0,
      "learning_rate": 4.1254348666408967e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3621
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 77.15625,
      "epoch": 0.17500966370313104,
      "grad_norm": 7.424845645354312,
      "learning_rate": 4.1251932740626203e-07,
      "loss": 0.0,
      "reward": 1.390625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 3622
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.75,
      "epoch": 0.17505798221878624,
      "grad_norm": 0.0,
      "learning_rate": 4.1249516814843444e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 3623
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 76.59375,
      "epoch": 0.17510630073444144,
      "grad_norm": 11.963638133112292,
      "learning_rate": 4.1247100889060685e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3624
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.0,
      "epoch": 0.17515461925009665,
      "grad_norm": 0.0,
      "learning_rate": 4.1244684963277927e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3625
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.21875,
      "epoch": 0.17520293776575183,
      "grad_norm": 0.0,
      "learning_rate": 4.1242269037495163e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3626
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.53125,
      "epoch": 0.17525125628140703,
      "grad_norm": 0.0,
      "learning_rate": 4.123985311171241e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3627
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.53125,
      "epoch": 0.17529957479706224,
      "grad_norm": 0.0,
      "learning_rate": 4.1237437185929646e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3628
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 70.3125,
      "epoch": 0.17534789331271744,
      "grad_norm": 12.697167168189644,
      "learning_rate": 4.1235021260146887e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.28460073471069336,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3629
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.96875,
      "epoch": 0.17539621182837264,
      "grad_norm": 10.80248513679243,
      "learning_rate": 4.123260533436413e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3630
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.03125,
      "epoch": 0.17544453034402782,
      "grad_norm": 0.0,
      "learning_rate": 4.123018940858137e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3631
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.46875,
      "epoch": 0.17549284885968303,
      "grad_norm": 0.0,
      "learning_rate": 4.1227773482798606e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3632
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.53125,
      "epoch": 0.17554116737533823,
      "grad_norm": 0.0,
      "learning_rate": 4.122535755701584e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3633
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.53125,
      "epoch": 0.17558948589099344,
      "grad_norm": 0.0,
      "learning_rate": 4.122294163123309e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3634
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.03125,
      "epoch": 0.17563780440664864,
      "grad_norm": 26.606269414087187,
      "learning_rate": 4.1220525705450324e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3635
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.1875,
      "epoch": 0.17568612292230382,
      "grad_norm": 0.0,
      "learning_rate": 4.1218109779667566e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3636
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.15625,
      "epoch": 0.17573444143795902,
      "grad_norm": 0.0,
      "learning_rate": 4.1215693853884807e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3637
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.03125,
      "epoch": 0.17578275995361423,
      "grad_norm": 0.0,
      "learning_rate": 4.121327792810205e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3638
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 71.21875,
      "epoch": 0.17583107846926943,
      "grad_norm": 11.974510365412934,
      "learning_rate": 4.1210862002319285e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.27564918994903564,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3639
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.6875,
      "epoch": 0.17587939698492464,
      "grad_norm": 0.0,
      "learning_rate": 4.120844607653653e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3640
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.0625,
      "epoch": 0.1759277155005798,
      "grad_norm": 0.0,
      "learning_rate": 4.120603015075377e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3641
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.5,
      "epoch": 0.17597603401623502,
      "grad_norm": 7.041226129652117,
      "learning_rate": 4.120361422497101e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3642
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.25,
      "epoch": 0.17602435253189022,
      "grad_norm": 0.0,
      "learning_rate": 4.1201198299188245e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3643
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.375,
      "epoch": 0.17607267104754543,
      "grad_norm": 0.0,
      "learning_rate": 4.1198782373405486e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3644
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.90625,
      "epoch": 0.17612098956320063,
      "grad_norm": 0.0,
      "learning_rate": 4.119636644762273e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3645
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.4375,
      "epoch": 0.1761693080788558,
      "grad_norm": 0.0,
      "learning_rate": 4.1193950521839964e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3646
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.0625,
      "epoch": 0.176217626594511,
      "grad_norm": 0.0,
      "learning_rate": 4.119153459605721e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3647
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.40625,
      "epoch": 0.17626594511016622,
      "grad_norm": 0.0,
      "learning_rate": 4.1189118670274446e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3648
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.4375,
      "epoch": 0.17631426362582142,
      "grad_norm": 0.0,
      "learning_rate": 4.118670274449169e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3649
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.09375,
      "epoch": 0.17636258214147663,
      "grad_norm": 0.0,
      "learning_rate": 4.118428681870893e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3650
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.46875,
      "epoch": 0.1764109006571318,
      "grad_norm": 0.0,
      "learning_rate": 4.118187089292617e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3651
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.5,
      "epoch": 0.176459219172787,
      "grad_norm": 0.0,
      "learning_rate": 4.1179454967143406e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3652
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.5625,
      "epoch": 0.1765075376884422,
      "grad_norm": 0.0,
      "learning_rate": 4.117703904136064e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3653
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.0,
      "epoch": 0.17655585620409742,
      "grad_norm": 0.0,
      "learning_rate": 4.117462311557789e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3654
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.65625,
      "epoch": 0.17660417471975262,
      "grad_norm": 0.0,
      "learning_rate": 4.1172207189795125e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3655
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.375,
      "epoch": 0.1766524932354078,
      "grad_norm": 0.0,
      "learning_rate": 4.1169791264012367e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3656
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.53125,
      "epoch": 0.176700811751063,
      "grad_norm": 0.0,
      "learning_rate": 4.116737533822961e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 3657
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 76.65625,
      "epoch": 0.1767491302667182,
      "grad_norm": 31.085249313479228,
      "learning_rate": 4.116495941244685e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3658
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.25,
      "epoch": 0.1767974487823734,
      "grad_norm": 0.0,
      "learning_rate": 4.1162543486664085e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3659
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.125,
      "epoch": 0.17684576729802862,
      "grad_norm": 0.0,
      "learning_rate": 4.116012756088133e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3660
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.3125,
      "epoch": 0.1768940858136838,
      "grad_norm": 0.0,
      "learning_rate": 4.115771163509857e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3661
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.125,
      "epoch": 0.176942404329339,
      "grad_norm": 0.0,
      "learning_rate": 4.115529570931581e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3662
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.6875,
      "epoch": 0.1769907228449942,
      "grad_norm": 0.0,
      "learning_rate": 4.1152879783533046e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3663
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.5625,
      "epoch": 0.1770390413606494,
      "grad_norm": 0.0,
      "learning_rate": 4.1150463857750287e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3664
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.5,
      "epoch": 0.1770873598763046,
      "grad_norm": 0.0,
      "learning_rate": 4.114804793196753e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3665
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.0,
      "epoch": 0.1771356783919598,
      "grad_norm": 0.0,
      "learning_rate": 4.1145632006184764e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3666
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.40625,
      "epoch": 0.177183996907615,
      "grad_norm": 0.0,
      "learning_rate": 4.114321608040201e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3667
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.53125,
      "epoch": 0.1772323154232702,
      "grad_norm": 0.0,
      "learning_rate": 4.1140800154619247e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3668
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.625,
      "epoch": 0.1772806339389254,
      "grad_norm": 0.0,
      "learning_rate": 4.113838422883649e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3669
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.71875,
      "epoch": 0.1773289524545806,
      "grad_norm": 0.0,
      "learning_rate": 4.113596830305373e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3670
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.71875,
      "epoch": 0.17737727097023578,
      "grad_norm": 0.0,
      "learning_rate": 4.113355237727097e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3671
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.28125,
      "epoch": 0.177425589485891,
      "grad_norm": 0.0,
      "learning_rate": 4.1131136451488207e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3672
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.84375,
      "epoch": 0.1774739080015462,
      "grad_norm": 0.0,
      "learning_rate": 4.1128720525705443e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3673
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.53125,
      "epoch": 0.1775222265172014,
      "grad_norm": 6.06551651851983,
      "learning_rate": 4.112630459992269e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3674
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.71875,
      "epoch": 0.1775705450328566,
      "grad_norm": 0.0,
      "learning_rate": 4.1123888674139926e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3675
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.25,
      "epoch": 0.17761886354851178,
      "grad_norm": 0.0,
      "learning_rate": 4.112147274835717e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3676
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.875,
      "epoch": 0.17766718206416698,
      "grad_norm": 0.0,
      "learning_rate": 4.111905682257441e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3677
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.9375,
      "epoch": 0.1777155005798222,
      "grad_norm": 0.0,
      "learning_rate": 4.111664089679165e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3678
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.6875,
      "epoch": 0.1777638190954774,
      "grad_norm": 0.0,
      "learning_rate": 4.1114224971008886e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 3679
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.5625,
      "epoch": 0.1778121376111326,
      "grad_norm": 0.0,
      "learning_rate": 4.1111809045226133e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3680
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.3125,
      "epoch": 0.17786045612678777,
      "grad_norm": 0.0,
      "learning_rate": 4.110939311944337e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3681
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.21875,
      "epoch": 0.17790877464244298,
      "grad_norm": 0.0,
      "learning_rate": 4.110697719366061e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3682
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 82.125,
      "epoch": 0.17795709315809818,
      "grad_norm": 27.179676297699796,
      "learning_rate": 4.1104561267877846e-07,
      "loss": -0.0,
      "reward": 1.09375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.0625,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 3683
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.96875,
      "epoch": 0.1780054116737534,
      "grad_norm": 0.0,
      "learning_rate": 4.110214534209509e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3684
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.09375,
      "epoch": 0.1780537301894086,
      "grad_norm": 0.0,
      "learning_rate": 4.109972941631233e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3685
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.65625,
      "epoch": 0.17810204870506377,
      "grad_norm": 33.27995841238892,
      "learning_rate": 4.1097313490529565e-07,
      "loss": -0.0,
      "reward": 1.921875,
      "reward_std": 0.2608962655067444,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3686
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.28125,
      "epoch": 0.17815036722071897,
      "grad_norm": 59.72368335521228,
      "learning_rate": 4.109489756474681e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.29854315519332886,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3687
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.78125,
      "epoch": 0.17819868573637418,
      "grad_norm": 0.0,
      "learning_rate": 4.109248163896405e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3688
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.0,
      "epoch": 0.17824700425202938,
      "grad_norm": 0.0,
      "learning_rate": 4.109006571318129e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 3689
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.1875,
      "epoch": 0.1782953227676846,
      "grad_norm": 0.0,
      "learning_rate": 4.108764978739853e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3690
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.34375,
      "epoch": 0.17834364128333977,
      "grad_norm": 0.0,
      "learning_rate": 4.108523386161577e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3691
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.21875,
      "epoch": 0.17839195979899497,
      "grad_norm": 0.0,
      "learning_rate": 4.108281793583301e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3692
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.5,
      "epoch": 0.17844027831465017,
      "grad_norm": 0.0,
      "learning_rate": 4.108040201005025e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3693
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.34375,
      "epoch": 0.17848859683030538,
      "grad_norm": 0.0,
      "learning_rate": 4.107798608426749e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3694
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.96875,
      "epoch": 0.17853691534596058,
      "grad_norm": 0.0,
      "learning_rate": 4.1075570158484727e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3695
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 74.875,
      "epoch": 0.17858523386161576,
      "grad_norm": 6.223010818745562,
      "learning_rate": 4.107315423270197e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.24359199404716492,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3696
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.1875,
      "epoch": 0.17863355237727097,
      "grad_norm": 0.0,
      "learning_rate": 4.107073830691921e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3697
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.15625,
      "epoch": 0.17868187089292617,
      "grad_norm": 74.57458239911954,
      "learning_rate": 4.106832238113645e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.22017991542816162,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3698
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.25,
      "epoch": 0.17873018940858137,
      "grad_norm": 0.0,
      "learning_rate": 4.1065906455353687e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3699
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.25,
      "epoch": 0.17877850792423658,
      "grad_norm": 0.0,
      "learning_rate": 4.1063490529570934e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3700
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.6875,
      "epoch": 0.17882682643989176,
      "grad_norm": 0.0,
      "learning_rate": 4.106107460378817e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3701
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.71875,
      "epoch": 0.17887514495554696,
      "grad_norm": 0.0,
      "learning_rate": 4.105865867800541e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3702
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.875,
      "epoch": 0.17892346347120217,
      "grad_norm": 0.0,
      "learning_rate": 4.1056242752222647e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3703
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.34375,
      "epoch": 0.17897178198685737,
      "grad_norm": 0.0,
      "learning_rate": 4.1053826826439894e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3704
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.125,
      "epoch": 0.17902010050251257,
      "grad_norm": 0.0,
      "learning_rate": 4.105141090065713e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3705
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.0625,
      "epoch": 0.17906841901816775,
      "grad_norm": 0.0,
      "learning_rate": 4.1048994974874366e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3706
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.4375,
      "epoch": 0.17911673753382296,
      "grad_norm": 23.88198995137247,
      "learning_rate": 4.104657904909161e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.2982703447341919,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3707
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.59375,
      "epoch": 0.17916505604947816,
      "grad_norm": 0.0,
      "learning_rate": 4.104416312330885e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3708
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.25,
      "epoch": 0.17921337456513337,
      "grad_norm": 0.0,
      "learning_rate": 4.104174719752609e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3709
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.5,
      "epoch": 0.17926169308078857,
      "grad_norm": 0.0,
      "learning_rate": 4.103933127174333e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3710
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.9375,
      "epoch": 0.17931001159644375,
      "grad_norm": 22.919894447507524,
      "learning_rate": 4.1036915345960573e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.34352827072143555,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3711
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.34375,
      "epoch": 0.17935833011209895,
      "grad_norm": 0.0,
      "learning_rate": 4.103449942017781e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3712
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.34375,
      "epoch": 0.17940664862775416,
      "grad_norm": 0.0,
      "learning_rate": 4.103208349439505e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3713
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.46875,
      "epoch": 0.17945496714340936,
      "grad_norm": 0.0,
      "learning_rate": 4.102966756861229e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3714
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 83.90625,
      "epoch": 0.17950328565906457,
      "grad_norm": 18.95206450292989,
      "learning_rate": 4.102725164282953e-07,
      "loss": -0.0,
      "reward": 2.140625,
      "reward_std": 0.2961486577987671,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3715
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.5,
      "epoch": 0.17955160417471974,
      "grad_norm": 0.0,
      "learning_rate": 4.102483571704677e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3716
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.6875,
      "epoch": 0.17959992269037495,
      "grad_norm": 0.0,
      "learning_rate": 4.102241979126401e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3717
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.75,
      "epoch": 0.17964824120603015,
      "grad_norm": 0.0,
      "learning_rate": 4.102000386548125e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3718
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.5625,
      "epoch": 0.17969655972168536,
      "grad_norm": 0.0,
      "learning_rate": 4.101758793969849e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3719
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.0,
      "epoch": 0.17974487823734056,
      "grad_norm": 0.0,
      "learning_rate": 4.1015172013915734e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3720
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.8125,
      "epoch": 0.17979319675299574,
      "grad_norm": 0.0,
      "learning_rate": 4.101275608813297e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3721
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.28125,
      "epoch": 0.17984151526865094,
      "grad_norm": 0.0,
      "learning_rate": 4.101034016235021e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3722
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.09375,
      "epoch": 0.17988983378430615,
      "grad_norm": 0.0,
      "learning_rate": 4.100792423656745e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3723
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.125,
      "epoch": 0.17993815229996135,
      "grad_norm": 0.0,
      "learning_rate": 4.1005508310784695e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3724
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.0,
      "epoch": 0.17998647081561656,
      "grad_norm": 0.0,
      "learning_rate": 4.100309238500193e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3725
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.53125,
      "epoch": 0.18003478933127173,
      "grad_norm": 0.0,
      "learning_rate": 4.1000676459219167e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3726
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.21875,
      "epoch": 0.18008310784692694,
      "grad_norm": 0.0,
      "learning_rate": 4.0998260533436413e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3727
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.0625,
      "epoch": 0.18013142636258214,
      "grad_norm": 0.0,
      "learning_rate": 4.099584460765365e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3728
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.9375,
      "epoch": 0.18017974487823735,
      "grad_norm": 0.0,
      "learning_rate": 4.099342868187089e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3729
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.84375,
      "epoch": 0.18022806339389255,
      "grad_norm": 0.0,
      "learning_rate": 4.099101275608813e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3730
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.9375,
      "epoch": 0.18027638190954773,
      "grad_norm": 0.0,
      "learning_rate": 4.0988596830305374e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3731
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.625,
      "epoch": 0.18032470042520293,
      "grad_norm": 0.0,
      "learning_rate": 4.098618090452261e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3732
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.96875,
      "epoch": 0.18037301894085814,
      "grad_norm": 0.0,
      "learning_rate": 4.098376497873985e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3733
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.8125,
      "epoch": 0.18042133745651334,
      "grad_norm": 0.0,
      "learning_rate": 4.098134905295709e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3734
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.375,
      "epoch": 0.18046965597216855,
      "grad_norm": 0.0,
      "learning_rate": 4.097893312717433e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3735
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.34375,
      "epoch": 0.18051797448782372,
      "grad_norm": 0.0,
      "learning_rate": 4.097651720139157e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3736
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.4375,
      "epoch": 0.18056629300347893,
      "grad_norm": 0.0,
      "learning_rate": 4.097410127560881e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3737
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.375,
      "epoch": 0.18061461151913413,
      "grad_norm": 0.0,
      "learning_rate": 4.097168534982605e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3738
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.84375,
      "epoch": 0.18066293003478934,
      "grad_norm": 0.0,
      "learning_rate": 4.096926942404329e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3739
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.0,
      "epoch": 0.18071124855044454,
      "grad_norm": 0.0,
      "learning_rate": 4.0966853498260535e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3740
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.875,
      "epoch": 0.18075956706609972,
      "grad_norm": 0.0,
      "learning_rate": 4.096443757247777e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3741
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.59375,
      "epoch": 0.18080788558175492,
      "grad_norm": 0.0,
      "learning_rate": 4.0962021646695013e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3742
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.28125,
      "epoch": 0.18085620409741013,
      "grad_norm": 0.0,
      "learning_rate": 4.095960572091225e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3743
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.5,
      "epoch": 0.18090452261306533,
      "grad_norm": 0.0,
      "learning_rate": 4.0957189795129495e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3744
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.78125,
      "epoch": 0.18095284112872054,
      "grad_norm": 0.0,
      "learning_rate": 4.095477386934673e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3745
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.6875,
      "epoch": 0.18100115964437571,
      "grad_norm": 0.0,
      "learning_rate": 4.095235794356397e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3746
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.53125,
      "epoch": 0.18104947816003092,
      "grad_norm": 0.0,
      "learning_rate": 4.0949942017781214e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3747
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.03125,
      "epoch": 0.18109779667568612,
      "grad_norm": 0.0,
      "learning_rate": 4.094752609199845e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3748
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.6875,
      "epoch": 0.18114611519134133,
      "grad_norm": 0.0,
      "learning_rate": 4.094511016621569e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3749
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.78125,
      "epoch": 0.18119443370699653,
      "grad_norm": 0.0,
      "learning_rate": 4.0942694240432933e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3750
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.71875,
      "epoch": 0.1812427522226517,
      "grad_norm": 0.0,
      "learning_rate": 4.0940278314650174e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3751
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.40625,
      "epoch": 0.18129107073830691,
      "grad_norm": 0.0,
      "learning_rate": 4.093786238886741e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3752
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.40625,
      "epoch": 0.18133938925396212,
      "grad_norm": 0.0,
      "learning_rate": 4.093544646308465e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3753
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.5625,
      "epoch": 0.18138770776961732,
      "grad_norm": 0.0,
      "learning_rate": 4.0933030537301893e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3754
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.15625,
      "epoch": 0.18143602628527253,
      "grad_norm": 0.0,
      "learning_rate": 4.0930614611519135e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3755
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.53125,
      "epoch": 0.1814843448009277,
      "grad_norm": 0.0,
      "learning_rate": 4.092819868573637e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3756
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.71875,
      "epoch": 0.1815326633165829,
      "grad_norm": 0.0,
      "learning_rate": 4.092578275995361e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 3757
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.0625,
      "epoch": 0.18158098183223811,
      "grad_norm": 0.0,
      "learning_rate": 4.0923366834170853e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 3758
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.09375,
      "epoch": 0.18162930034789332,
      "grad_norm": 0.0,
      "learning_rate": 4.092095090838809e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3759
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.84375,
      "epoch": 0.18167761886354852,
      "grad_norm": 0.0,
      "learning_rate": 4.0918534982605336e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3760
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.40625,
      "epoch": 0.1817259373792037,
      "grad_norm": 0.0,
      "learning_rate": 4.091611905682257e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3761
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.3125,
      "epoch": 0.1817742558948589,
      "grad_norm": 0.0,
      "learning_rate": 4.0913703131039813e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3762
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.0625,
      "epoch": 0.1818225744105141,
      "grad_norm": 0.0,
      "learning_rate": 4.091128720525705e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3763
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.96875,
      "epoch": 0.18187089292616931,
      "grad_norm": 0.0,
      "learning_rate": 4.0908871279474296e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3764
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 75.875,
      "epoch": 0.18191921144182452,
      "grad_norm": 250.62791729445274,
      "learning_rate": 4.090645535369153e-07,
      "loss": 0.0,
      "reward": 1.28125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 3765
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.15625,
      "epoch": 0.1819675299574797,
      "grad_norm": 0.0,
      "learning_rate": 4.090403942790877e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3766
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.75,
      "epoch": 0.1820158484731349,
      "grad_norm": 0.0,
      "learning_rate": 4.0901623502126015e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3767
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.65625,
      "epoch": 0.1820641669887901,
      "grad_norm": 21.84537620400594,
      "learning_rate": 4.089920757634325e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.3029785752296448,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3768
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.59375,
      "epoch": 0.1821124855044453,
      "grad_norm": 0.0,
      "learning_rate": 4.089679165056049e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 3769
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.96875,
      "epoch": 0.18216080402010051,
      "grad_norm": 0.0,
      "learning_rate": 4.0894375724777734e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3770
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.03125,
      "epoch": 0.1822091225357557,
      "grad_norm": 0.0,
      "learning_rate": 4.0891959798994975e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3771
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.84375,
      "epoch": 0.1822574410514109,
      "grad_norm": 0.0,
      "learning_rate": 4.088954387321221e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3772
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.9375,
      "epoch": 0.1823057595670661,
      "grad_norm": 0.0,
      "learning_rate": 4.088712794742945e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3773
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.3125,
      "epoch": 0.1823540780827213,
      "grad_norm": 0.0,
      "learning_rate": 4.0884712021646694e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3774
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.84375,
      "epoch": 0.1824023965983765,
      "grad_norm": 0.0,
      "learning_rate": 4.0882296095863935e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3775
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.6875,
      "epoch": 0.1824507151140317,
      "grad_norm": 0.0,
      "learning_rate": 4.087988017008117e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3776
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.09375,
      "epoch": 0.1824990336296869,
      "grad_norm": 0.0,
      "learning_rate": 4.0877464244298413e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3777
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.5625,
      "epoch": 0.1825473521453421,
      "grad_norm": 0.0,
      "learning_rate": 4.0875048318515654e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3778
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.90625,
      "epoch": 0.1825956706609973,
      "grad_norm": 0.0,
      "learning_rate": 4.087263239273289e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3779
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.03125,
      "epoch": 0.1826439891766525,
      "grad_norm": 0.0,
      "learning_rate": 4.0870216466950137e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3780
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.40625,
      "epoch": 0.18269230769230768,
      "grad_norm": 0.0,
      "learning_rate": 4.0867800541167373e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3781
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.65625,
      "epoch": 0.1827406262079629,
      "grad_norm": 0.0,
      "learning_rate": 4.0865384615384614e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3782
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.59375,
      "epoch": 0.1827889447236181,
      "grad_norm": 0.0,
      "learning_rate": 4.086296868960185e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3783
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.75,
      "epoch": 0.1828372632392733,
      "grad_norm": 0.0,
      "learning_rate": 4.0860552763819097e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 3784
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.9375,
      "epoch": 0.1828855817549285,
      "grad_norm": 0.0,
      "learning_rate": 4.0858136838036333e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3785
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.3125,
      "epoch": 0.18293390027058368,
      "grad_norm": 0.0,
      "learning_rate": 4.085572091225357e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3786
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.0625,
      "epoch": 0.18298221878623888,
      "grad_norm": 0.0,
      "learning_rate": 4.0853304986470816e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3787
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.15625,
      "epoch": 0.1830305373018941,
      "grad_norm": 0.0,
      "learning_rate": 4.085088906068805e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3788
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.03125,
      "epoch": 0.1830788558175493,
      "grad_norm": 0.0,
      "learning_rate": 4.0848473134905293e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3789
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.9375,
      "epoch": 0.1831271743332045,
      "grad_norm": 0.0,
      "learning_rate": 4.0846057209122535e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3790
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.21875,
      "epoch": 0.18317549284885967,
      "grad_norm": 0.0,
      "learning_rate": 4.0843641283339776e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3791
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 87.5,
      "epoch": 0.18322381136451488,
      "grad_norm": 8.046025787475466,
      "learning_rate": 4.084122535755701e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3792
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.4375,
      "epoch": 0.18327212988017008,
      "grad_norm": 0.0,
      "learning_rate": 4.0838809431774253e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3793
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.4375,
      "epoch": 0.1833204483958253,
      "grad_norm": 0.0,
      "learning_rate": 4.0836393505991495e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3794
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.28125,
      "epoch": 0.1833687669114805,
      "grad_norm": 0.0,
      "learning_rate": 4.0833977580208736e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3795
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.40625,
      "epoch": 0.18341708542713567,
      "grad_norm": 0.0,
      "learning_rate": 4.083156165442597e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3796
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.84375,
      "epoch": 0.18346540394279087,
      "grad_norm": 0.0,
      "learning_rate": 4.0829145728643214e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3797
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.28125,
      "epoch": 0.18351372245844608,
      "grad_norm": 0.0,
      "learning_rate": 4.0826729802860455e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3798
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.90625,
      "epoch": 0.18356204097410128,
      "grad_norm": 0.0,
      "learning_rate": 4.082431387707769e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3799
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.96875,
      "epoch": 0.1836103594897565,
      "grad_norm": 10.563888123569832,
      "learning_rate": 4.082189795129494e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.3255898952484131,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3800
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.15625,
      "epoch": 0.18365867800541166,
      "grad_norm": 0.0,
      "learning_rate": 4.0819482025512174e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3801
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.78125,
      "epoch": 0.18370699652106687,
      "grad_norm": 0.0,
      "learning_rate": 4.0817066099729415e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3802
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.4375,
      "epoch": 0.18375531503672207,
      "grad_norm": 0.0,
      "learning_rate": 4.081465017394665e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3803
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 80.65625,
      "epoch": 0.18380363355237728,
      "grad_norm": 6.989630366377989,
      "learning_rate": 4.08122342481639e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.12387890368700027,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3804
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.875,
      "epoch": 0.18385195206803248,
      "grad_norm": 0.0,
      "learning_rate": 4.0809818322381134e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3805
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.15625,
      "epoch": 0.18390027058368766,
      "grad_norm": 0.0,
      "learning_rate": 4.0807402396598375e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3806
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.15625,
      "epoch": 0.18394858909934286,
      "grad_norm": 0.0,
      "learning_rate": 4.0804986470815617e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3807
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.1875,
      "epoch": 0.18399690761499807,
      "grad_norm": 0.0,
      "learning_rate": 4.080257054503285e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3808
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.3125,
      "epoch": 0.18404522613065327,
      "grad_norm": 0.0,
      "learning_rate": 4.0800154619250094e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3809
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.25,
      "epoch": 0.18409354464630848,
      "grad_norm": 12.500419126786511,
      "learning_rate": 4.0797738693467335e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.30509042739868164,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3810
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.78125,
      "epoch": 0.18414186316196365,
      "grad_norm": 0.0,
      "learning_rate": 4.0795322767684577e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3811
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 80.78125,
      "epoch": 0.18419018167761886,
      "grad_norm": 9.134786268390949,
      "learning_rate": 4.0792906841901813e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3812
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.78125,
      "epoch": 0.18423850019327406,
      "grad_norm": 0.0,
      "learning_rate": 4.0790490916119054e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3813
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.875,
      "epoch": 0.18428681870892927,
      "grad_norm": 0.0,
      "learning_rate": 4.0788074990336296e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3814
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.1875,
      "epoch": 0.18433513722458447,
      "grad_norm": 0.0,
      "learning_rate": 4.0785659064553537e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3815
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.375,
      "epoch": 0.18438345574023965,
      "grad_norm": 0.0,
      "learning_rate": 4.0783243138770773e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3816
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.6875,
      "epoch": 0.18443177425589485,
      "grad_norm": 0.0,
      "learning_rate": 4.078082721298802e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3817
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.53125,
      "epoch": 0.18448009277155006,
      "grad_norm": 0.0,
      "learning_rate": 4.0778411287205256e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3818
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.625,
      "epoch": 0.18452841128720526,
      "grad_norm": 0.0,
      "learning_rate": 4.077599536142249e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3819
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.53125,
      "epoch": 0.18457672980286047,
      "grad_norm": 0.0,
      "learning_rate": 4.077357943563974e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3820
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.125,
      "epoch": 0.18462504831851564,
      "grad_norm": 0.0,
      "learning_rate": 4.0771163509856974e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3821
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.34375,
      "epoch": 0.18467336683417085,
      "grad_norm": 0.0,
      "learning_rate": 4.0768747584074216e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3822
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.875,
      "epoch": 0.18472168534982605,
      "grad_norm": 0.0,
      "learning_rate": 4.076633165829145e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3823
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.34375,
      "epoch": 0.18477000386548126,
      "grad_norm": 0.0,
      "learning_rate": 4.07639157325087e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3824
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.9375,
      "epoch": 0.18481832238113646,
      "grad_norm": 0.0,
      "learning_rate": 4.0761499806725935e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3825
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.625,
      "epoch": 0.18486664089679164,
      "grad_norm": 0.0,
      "learning_rate": 4.0759083880943176e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3826
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.40625,
      "epoch": 0.18491495941244684,
      "grad_norm": 0.0,
      "learning_rate": 4.0756667955160417e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3827
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.46875,
      "epoch": 0.18496327792810205,
      "grad_norm": 0.0,
      "learning_rate": 4.0754252029377653e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3828
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.0625,
      "epoch": 0.18501159644375725,
      "grad_norm": 0.0,
      "learning_rate": 4.0751836103594895e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3829
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.4375,
      "epoch": 0.18505991495941246,
      "grad_norm": 0.0,
      "learning_rate": 4.0749420177812136e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3830
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.75,
      "epoch": 0.18510823347506764,
      "grad_norm": 0.0,
      "learning_rate": 4.074700425202938e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3831
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.65625,
      "epoch": 0.18515655199072284,
      "grad_norm": 0.0,
      "learning_rate": 4.0744588326246614e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3832
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.8125,
      "epoch": 0.18520487050637804,
      "grad_norm": 0.0,
      "learning_rate": 4.0742172400463855e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3833
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.25,
      "epoch": 0.18525318902203325,
      "grad_norm": 0.0,
      "learning_rate": 4.0739756474681096e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3834
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.875,
      "epoch": 0.18530150753768845,
      "grad_norm": 0.0,
      "learning_rate": 4.073734054889834e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3835
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.90625,
      "epoch": 0.18534982605334363,
      "grad_norm": 0.0,
      "learning_rate": 4.0734924623115574e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3836
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.4375,
      "epoch": 0.18539814456899883,
      "grad_norm": 0.0,
      "learning_rate": 4.073250869733282e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 3837
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 75.125,
      "epoch": 0.18544646308465404,
      "grad_norm": 45.8745123751646,
      "learning_rate": 4.0730092771550056e-07,
      "loss": -0.0,
      "reward": 2.21875,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3838
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.96875,
      "epoch": 0.18549478160030924,
      "grad_norm": 0.0,
      "learning_rate": 4.072767684576729e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3839
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.03125,
      "epoch": 0.18554310011596445,
      "grad_norm": 0.0,
      "learning_rate": 4.072526091998454e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3840
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.53125,
      "epoch": 0.18559141863161963,
      "grad_norm": 0.0,
      "learning_rate": 4.0722844994201775e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3841
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 71.84375,
      "epoch": 0.18563973714727483,
      "grad_norm": 0.0,
      "learning_rate": 4.0720429068419017e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3842
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.75,
      "epoch": 0.18568805566293003,
      "grad_norm": 0.0,
      "learning_rate": 4.0718013142636253e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3843
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.0625,
      "epoch": 0.18573637417858524,
      "grad_norm": 0.0,
      "learning_rate": 4.07155972168535e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3844
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.28125,
      "epoch": 0.18578469269424044,
      "grad_norm": 0.0,
      "learning_rate": 4.0713181291070735e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3845
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.875,
      "epoch": 0.18583301120989562,
      "grad_norm": 0.0,
      "learning_rate": 4.0710765365287977e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3846
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.0625,
      "epoch": 0.18588132972555083,
      "grad_norm": 14.11968053743672,
      "learning_rate": 4.070834943950522e-07,
      "loss": -0.0,
      "reward": 2.140625,
      "reward_std": 0.308285653591156,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3847
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.625,
      "epoch": 0.18592964824120603,
      "grad_norm": 0.0,
      "learning_rate": 4.070593351372246e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3848
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.59375,
      "epoch": 0.18597796675686123,
      "grad_norm": 0.0,
      "learning_rate": 4.0703517587939696e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3849
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.3125,
      "epoch": 0.18602628527251644,
      "grad_norm": 0.0,
      "learning_rate": 4.0701101662156937e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3850
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.71875,
      "epoch": 0.18607460378817162,
      "grad_norm": 0.0,
      "learning_rate": 4.069868573637418e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3851
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.5625,
      "epoch": 0.18612292230382682,
      "grad_norm": 0.0,
      "learning_rate": 4.0696269810591414e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3852
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.0625,
      "epoch": 0.18617124081948203,
      "grad_norm": 0.0,
      "learning_rate": 4.0693853884808656e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3853
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.6875,
      "epoch": 0.18621955933513723,
      "grad_norm": 0.0,
      "learning_rate": 4.0691437959025897e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 3854
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.0625,
      "epoch": 0.18626787785079243,
      "grad_norm": 0.0,
      "learning_rate": 4.068902203324314e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3855
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.59375,
      "epoch": 0.1863161963664476,
      "grad_norm": 0.0,
      "learning_rate": 4.0686606107460375e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3856
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.75,
      "epoch": 0.18636451488210282,
      "grad_norm": 0.0,
      "learning_rate": 4.068419018167762e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3857
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.90625,
      "epoch": 0.18641283339775802,
      "grad_norm": 0.0,
      "learning_rate": 4.0681774255894857e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 3858
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.40625,
      "epoch": 0.18646115191341323,
      "grad_norm": 0.0,
      "learning_rate": 4.0679358330112093e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 3859
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.78125,
      "epoch": 0.18650947042906843,
      "grad_norm": 0.0,
      "learning_rate": 4.067694240432934e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3860
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.4375,
      "epoch": 0.1865577889447236,
      "grad_norm": 0.0,
      "learning_rate": 4.0674526478546576e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3861
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.34375,
      "epoch": 0.1866061074603788,
      "grad_norm": 0.0,
      "learning_rate": 4.067211055276382e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3862
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.46875,
      "epoch": 0.18665442597603402,
      "grad_norm": 0.0,
      "learning_rate": 4.0669694626981053e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3863
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.59375,
      "epoch": 0.18670274449168922,
      "grad_norm": 0.0,
      "learning_rate": 4.06672787011983e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3864
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.1875,
      "epoch": 0.18675106300734443,
      "grad_norm": 0.0,
      "learning_rate": 4.0664862775415536e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3865
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.625,
      "epoch": 0.1867993815229996,
      "grad_norm": 0.0,
      "learning_rate": 4.066244684963278e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3866
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.5625,
      "epoch": 0.1868477000386548,
      "grad_norm": 0.0,
      "learning_rate": 4.066003092385002e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3867
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.1875,
      "epoch": 0.18689601855431,
      "grad_norm": 0.0,
      "learning_rate": 4.065761499806726e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3868
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.96875,
      "epoch": 0.18694433706996522,
      "grad_norm": 0.0,
      "learning_rate": 4.0655199072284496e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3869
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.09375,
      "epoch": 0.18699265558562042,
      "grad_norm": 0.0,
      "learning_rate": 4.065278314650174e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3870
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 86.375,
      "epoch": 0.1870409741012756,
      "grad_norm": 44.03251365682515,
      "learning_rate": 4.065036722071898e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.37722259759902954,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3871
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.03125,
      "epoch": 0.1870892926169308,
      "grad_norm": 0.0,
      "learning_rate": 4.0647951294936215e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3872
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.21875,
      "epoch": 0.187137611132586,
      "grad_norm": 0.0,
      "learning_rate": 4.0645535369153457e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3873
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 92.5625,
      "epoch": 0.1871859296482412,
      "grad_norm": 5.463674392761304,
      "learning_rate": 4.06431194433707e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3874
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.875,
      "epoch": 0.18723424816389642,
      "grad_norm": 9.945009758160786,
      "learning_rate": 4.064070351758794e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3875
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.71875,
      "epoch": 0.1872825666795516,
      "grad_norm": 0.0,
      "learning_rate": 4.0638287591805175e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3876
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.53125,
      "epoch": 0.1873308851952068,
      "grad_norm": 0.0,
      "learning_rate": 4.063587166602242e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3877
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.78125,
      "epoch": 0.187379203710862,
      "grad_norm": 0.0,
      "learning_rate": 4.063345574023966e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3878
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.78125,
      "epoch": 0.1874275222265172,
      "grad_norm": 0.0,
      "learning_rate": 4.0631039814456894e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 3879
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.375,
      "epoch": 0.1874758407421724,
      "grad_norm": 0.0,
      "learning_rate": 4.062862388867414e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3880
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.09375,
      "epoch": 0.1875241592578276,
      "grad_norm": 0.0,
      "learning_rate": 4.0626207962891377e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3881
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.5625,
      "epoch": 0.1875724777734828,
      "grad_norm": 0.0,
      "learning_rate": 4.062379203710862e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3882
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 93.40625,
      "epoch": 0.187620796289138,
      "grad_norm": 21.86892918638986,
      "learning_rate": 4.0621376111325854e-07,
      "loss": -0.0,
      "reward": 1.515625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3883
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.875,
      "epoch": 0.1876691148047932,
      "grad_norm": 0.0,
      "learning_rate": 4.06189601855431e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3884
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.21875,
      "epoch": 0.1877174333204484,
      "grad_norm": 0.0,
      "learning_rate": 4.0616544259760337e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3885
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.53125,
      "epoch": 0.18776575183610358,
      "grad_norm": 0.0,
      "learning_rate": 4.061412833397758e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3886
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.0625,
      "epoch": 0.1878140703517588,
      "grad_norm": 0.0,
      "learning_rate": 4.061171240819482e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3887
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.40625,
      "epoch": 0.187862388867414,
      "grad_norm": 0.0,
      "learning_rate": 4.060929648241206e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3888
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 85.65625,
      "epoch": 0.1879107073830692,
      "grad_norm": 16.286359412605286,
      "learning_rate": 4.0606880556629297e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3889
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.3125,
      "epoch": 0.1879590258987244,
      "grad_norm": 0.0,
      "learning_rate": 4.060446463084654e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 3890
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 86.03125,
      "epoch": 0.18800734441437958,
      "grad_norm": 24.06875083375243,
      "learning_rate": 4.060204870506378e-07,
      "loss": -0.0,
      "reward": 1.296875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 3891
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.5,
      "epoch": 0.18805566293003478,
      "grad_norm": 9.574347771446911,
      "learning_rate": 4.0599632779281016e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3892
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.53125,
      "epoch": 0.18810398144569,
      "grad_norm": 0.0,
      "learning_rate": 4.0597216853498257e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 3893
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.90625,
      "epoch": 0.1881522999613452,
      "grad_norm": 0.0,
      "learning_rate": 4.05948009277155e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 3894
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.25,
      "epoch": 0.1882006184770004,
      "grad_norm": 9.062403125905373,
      "learning_rate": 4.059238500193274e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.2435920089483261,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3895
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.4375,
      "epoch": 0.18824893699265557,
      "grad_norm": 0.0,
      "learning_rate": 4.0589969076149976e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 3896
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 91.125,
      "epoch": 0.18829725550831078,
      "grad_norm": 38.47405629212547,
      "learning_rate": 4.0587553150367223e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.4697399437427521,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3897
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.34375,
      "epoch": 0.18834557402396598,
      "grad_norm": 0.0,
      "learning_rate": 4.058513722458446e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3898
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.34375,
      "epoch": 0.1883938925396212,
      "grad_norm": 0.0,
      "learning_rate": 4.05827212988017e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3899
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.65625,
      "epoch": 0.1884422110552764,
      "grad_norm": 0.0,
      "learning_rate": 4.058030537301894e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3900
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.0625,
      "epoch": 0.18849052957093157,
      "grad_norm": 0.0,
      "learning_rate": 4.057788944723618e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3901
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.84375,
      "epoch": 0.18853884808658677,
      "grad_norm": 0.0,
      "learning_rate": 4.057547352145342e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3902
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.34375,
      "epoch": 0.18858716660224198,
      "grad_norm": 99.96942914425524,
      "learning_rate": 4.0573057595670655e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3903
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.90625,
      "epoch": 0.18863548511789718,
      "grad_norm": 0.0,
      "learning_rate": 4.05706416698879e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3904
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.03125,
      "epoch": 0.1886838036335524,
      "grad_norm": 0.0,
      "learning_rate": 4.056822574410514e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3905
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.84375,
      "epoch": 0.18873212214920757,
      "grad_norm": 0.0,
      "learning_rate": 4.056580981832238e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3906
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.28125,
      "epoch": 0.18878044066486277,
      "grad_norm": 0.0,
      "learning_rate": 4.056339389253962e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3907
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 92.0,
      "epoch": 0.18882875918051797,
      "grad_norm": 4.753411063647309,
      "learning_rate": 4.056097796675686e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.24831002950668335,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3908
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.34375,
      "epoch": 0.18887707769617318,
      "grad_norm": 0.0,
      "learning_rate": 4.05585620409741e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3909
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.9375,
      "epoch": 0.18892539621182838,
      "grad_norm": 0.0,
      "learning_rate": 4.0556146115191345e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3910
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.46875,
      "epoch": 0.18897371472748356,
      "grad_norm": 0.0,
      "learning_rate": 4.055373018940858e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3911
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 90.8125,
      "epoch": 0.18902203324313877,
      "grad_norm": 0.0,
      "learning_rate": 4.0551314263625817e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 3912
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.9375,
      "epoch": 0.18907035175879397,
      "grad_norm": 22.05136963425983,
      "learning_rate": 4.054889833784306e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3913
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.6875,
      "epoch": 0.18911867027444917,
      "grad_norm": 0.0,
      "learning_rate": 4.05464824120603e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3914
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.9375,
      "epoch": 0.18916698879010438,
      "grad_norm": 0.0,
      "learning_rate": 4.054406648627754e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3915
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.34375,
      "epoch": 0.18921530730575956,
      "grad_norm": 0.0,
      "learning_rate": 4.0541650560494777e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3916
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.53125,
      "epoch": 0.18926362582141476,
      "grad_norm": 0.0,
      "learning_rate": 4.0539234634712024e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3917
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.4375,
      "epoch": 0.18931194433706997,
      "grad_norm": 0.0,
      "learning_rate": 4.053681870892926e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3918
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 98.0625,
      "epoch": 0.18936026285272517,
      "grad_norm": 14.931637076357832,
      "learning_rate": 4.05344027831465e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.3051002025604248,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3919
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.75,
      "epoch": 0.18940858136838037,
      "grad_norm": 0.0,
      "learning_rate": 4.053198685736374e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3920
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.09375,
      "epoch": 0.18945689988403555,
      "grad_norm": 0.0,
      "learning_rate": 4.052957093158098e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3921
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.625,
      "epoch": 0.18950521839969076,
      "grad_norm": 0.0,
      "learning_rate": 4.052715500579822e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3922
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.90625,
      "epoch": 0.18955353691534596,
      "grad_norm": 0.0,
      "learning_rate": 4.0524739080015456e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3923
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.53125,
      "epoch": 0.18960185543100117,
      "grad_norm": 0.0,
      "learning_rate": 4.05223231542327e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3924
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.25,
      "epoch": 0.18965017394665637,
      "grad_norm": 0.0,
      "learning_rate": 4.051990722844994e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3925
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.84375,
      "epoch": 0.18969849246231155,
      "grad_norm": 0.0,
      "learning_rate": 4.051749130266718e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3926
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.5,
      "epoch": 0.18974681097796675,
      "grad_norm": 0.0,
      "learning_rate": 4.051507537688442e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 3927
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 104.09375,
      "epoch": 0.18979512949362196,
      "grad_norm": 5.192979214008399,
      "learning_rate": 4.051265945110166e-07,
      "loss": -0.0,
      "reward": 1.6875,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3928
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 91.21875,
      "epoch": 0.18984344800927716,
      "grad_norm": 12.93755429086488,
      "learning_rate": 4.05102435253189e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.3298332095146179,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3929
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.9375,
      "epoch": 0.18989176652493236,
      "grad_norm": 0.0,
      "learning_rate": 4.0507827599536145e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3930
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.40625,
      "epoch": 0.18994008504058754,
      "grad_norm": 0.0,
      "learning_rate": 4.050541167375338e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3931
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 97.3125,
      "epoch": 0.18998840355624275,
      "grad_norm": 0.0,
      "learning_rate": 4.050299574797062e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3932
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.25,
      "epoch": 0.19003672207189795,
      "grad_norm": 0.0,
      "learning_rate": 4.050057982218786e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3933
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.625,
      "epoch": 0.19008504058755316,
      "grad_norm": 0.0,
      "learning_rate": 4.04981638964051e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3934
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.0625,
      "epoch": 0.19013335910320836,
      "grad_norm": 0.0,
      "learning_rate": 4.049574797062234e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3935
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.75,
      "epoch": 0.19018167761886354,
      "grad_norm": 0.0,
      "learning_rate": 4.049333204483958e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3936
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.9375,
      "epoch": 0.19022999613451874,
      "grad_norm": 0.0,
      "learning_rate": 4.0490916119056824e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3937
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.59375,
      "epoch": 0.19027831465017395,
      "grad_norm": 0.0,
      "learning_rate": 4.048850019327406e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3938
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.875,
      "epoch": 0.19032663316582915,
      "grad_norm": 0.0,
      "learning_rate": 4.04860842674913e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3939
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.5625,
      "epoch": 0.19037495168148436,
      "grad_norm": 0.0,
      "learning_rate": 4.0483668341708543e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3940
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.125,
      "epoch": 0.19042327019713953,
      "grad_norm": 0.0,
      "learning_rate": 4.048125241592578e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3941
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.6875,
      "epoch": 0.19047158871279474,
      "grad_norm": 0.0,
      "learning_rate": 4.047883649014302e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3942
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.0625,
      "epoch": 0.19051990722844994,
      "grad_norm": 0.0,
      "learning_rate": 4.0476420564360257e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 3943
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.875,
      "epoch": 0.19056822574410515,
      "grad_norm": 11.689507364450943,
      "learning_rate": 4.0474004638577503e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3944
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 96.53125,
      "epoch": 0.19061654425976035,
      "grad_norm": 29.517278648880005,
      "learning_rate": 4.047158871279474e-07,
      "loss": 0.0,
      "reward": 2.1875,
      "reward_std": 0.3197515308856964,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3945
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.75,
      "epoch": 0.19066486277541553,
      "grad_norm": 0.0,
      "learning_rate": 4.046917278701198e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3946
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.28125,
      "epoch": 0.19071318129107073,
      "grad_norm": 0.0,
      "learning_rate": 4.046675686122922e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3947
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.78125,
      "epoch": 0.19076149980672594,
      "grad_norm": 0.0,
      "learning_rate": 4.0464340935446463e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3948
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.4375,
      "epoch": 0.19080981832238114,
      "grad_norm": 0.0,
      "learning_rate": 4.04619250096637e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3949
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.40625,
      "epoch": 0.19085813683803635,
      "grad_norm": 0.0,
      "learning_rate": 4.0459509083880946e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 3950
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 86.78125,
      "epoch": 0.19090645535369152,
      "grad_norm": 9.218901498425598,
      "learning_rate": 4.045709315809818e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3951
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.40625,
      "epoch": 0.19095477386934673,
      "grad_norm": 0.0,
      "learning_rate": 4.045467723231542e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3952
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.375,
      "epoch": 0.19100309238500193,
      "grad_norm": 0.0,
      "learning_rate": 4.045226130653266e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3953
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.46875,
      "epoch": 0.19105141090065714,
      "grad_norm": 0.0,
      "learning_rate": 4.04498453807499e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3954
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 101.46875,
      "epoch": 0.19109972941631234,
      "grad_norm": 133.81890242291303,
      "learning_rate": 4.044742945496714e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.376485139131546,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3955
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.5,
      "epoch": 0.19114804793196752,
      "grad_norm": 0.0,
      "learning_rate": 4.044501352918438e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3956
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.4375,
      "epoch": 0.19119636644762272,
      "grad_norm": 0.0,
      "learning_rate": 4.0442597603401625e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3957
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.09375,
      "epoch": 0.19124468496327793,
      "grad_norm": 0.0,
      "learning_rate": 4.044018167761886e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3958
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.03125,
      "epoch": 0.19129300347893313,
      "grad_norm": 0.0,
      "learning_rate": 4.04377657518361e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3959
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.59375,
      "epoch": 0.19134132199458834,
      "grad_norm": 0.0,
      "learning_rate": 4.0435349826053344e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 3960
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.59375,
      "epoch": 0.1913896405102435,
      "grad_norm": 0.0,
      "learning_rate": 4.0432933900270585e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 3961
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.71875,
      "epoch": 0.19143795902589872,
      "grad_norm": 0.0,
      "learning_rate": 4.043051797448782e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 3962
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.65625,
      "epoch": 0.19148627754155392,
      "grad_norm": 0.0,
      "learning_rate": 4.042810204870506e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3963
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.71875,
      "epoch": 0.19153459605720913,
      "grad_norm": 0.0,
      "learning_rate": 4.0425686122922304e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 3964
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.03125,
      "epoch": 0.19158291457286433,
      "grad_norm": 0.0,
      "learning_rate": 4.042327019713954e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3965
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 98.3125,
      "epoch": 0.1916312330885195,
      "grad_norm": 34.706151296969004,
      "learning_rate": 4.042085427135678e-07,
      "loss": -0.0,
      "reward": 1.515625,
      "reward_std": 0.3331887722015381,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3966
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.40625,
      "epoch": 0.1916795516041747,
      "grad_norm": 0.0,
      "learning_rate": 4.0418438345574023e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 3967
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 92.03125,
      "epoch": 0.19172787011982992,
      "grad_norm": 16.149400460746644,
      "learning_rate": 4.0416022419791264e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 3968
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 103.75,
      "epoch": 0.19177618863548512,
      "grad_norm": 7.279976727976142,
      "learning_rate": 4.04136064940085e-07,
      "loss": -0.0,
      "reward": 1.515625,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 3969
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.71875,
      "epoch": 0.19182450715114033,
      "grad_norm": 0.0,
      "learning_rate": 4.0411190568225747e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 3970
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.625,
      "epoch": 0.1918728256667955,
      "grad_norm": 0.0,
      "learning_rate": 4.0408774642442983e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3971
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.28125,
      "epoch": 0.1919211441824507,
      "grad_norm": 0.0,
      "learning_rate": 4.040635871666022e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3972
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 101.28125,
      "epoch": 0.1919694626981059,
      "grad_norm": 5.494954358786813,
      "learning_rate": 4.040394279087746e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3973
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.78125,
      "epoch": 0.19201778121376112,
      "grad_norm": 0.0,
      "learning_rate": 4.04015268650947e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3974
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.3125,
      "epoch": 0.19206609972941632,
      "grad_norm": 0.0,
      "learning_rate": 4.0399110939311943e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 3975
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.875,
      "epoch": 0.1921144182450715,
      "grad_norm": 0.0,
      "learning_rate": 4.039669501352918e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3976
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.28125,
      "epoch": 0.1921627367607267,
      "grad_norm": 0.0,
      "learning_rate": 4.0394279087746426e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3977
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.3125,
      "epoch": 0.1922110552763819,
      "grad_norm": 0.0,
      "learning_rate": 4.039186316196366e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3978
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.03125,
      "epoch": 0.1922593737920371,
      "grad_norm": 0.0,
      "learning_rate": 4.0389447236180903e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3979
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.0625,
      "epoch": 0.19230769230769232,
      "grad_norm": 0.0,
      "learning_rate": 4.0387031310398145e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3980
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.8125,
      "epoch": 0.1923560108233475,
      "grad_norm": 0.0,
      "learning_rate": 4.0384615384615386e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3981
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.5,
      "epoch": 0.1924043293390027,
      "grad_norm": 0.0,
      "learning_rate": 4.038219945883262e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 3982
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.75,
      "epoch": 0.1924526478546579,
      "grad_norm": 0.0,
      "learning_rate": 4.037978353304986e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 3983
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.75,
      "epoch": 0.1925009663703131,
      "grad_norm": 0.0,
      "learning_rate": 4.0377367607267105e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 3984
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.4375,
      "epoch": 0.1925492848859683,
      "grad_norm": 11.76748107670613,
      "learning_rate": 4.037495168148434e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.22049658000469208,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 3985
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.125,
      "epoch": 0.1925976034016235,
      "grad_norm": 0.0,
      "learning_rate": 4.037253575570158e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3986
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.8125,
      "epoch": 0.1926459219172787,
      "grad_norm": 0.0,
      "learning_rate": 4.0370119829918824e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3987
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.21875,
      "epoch": 0.1926942404329339,
      "grad_norm": 0.0,
      "learning_rate": 4.0367703904136065e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 3988
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.84375,
      "epoch": 0.1927425589485891,
      "grad_norm": 0.0,
      "learning_rate": 4.03652879783533e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 3989
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.28125,
      "epoch": 0.1927908774642443,
      "grad_norm": 0.0,
      "learning_rate": 4.036287205257055e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 3990
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.03125,
      "epoch": 0.19283919597989949,
      "grad_norm": 0.0,
      "learning_rate": 4.0360456126787784e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 3991
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 95.65625,
      "epoch": 0.1928875144955547,
      "grad_norm": 4.244213634272018,
      "learning_rate": 4.035804020100502e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3992
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.40625,
      "epoch": 0.1929358330112099,
      "grad_norm": 0.0,
      "learning_rate": 4.035562427522226e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3993
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.46875,
      "epoch": 0.1929841515268651,
      "grad_norm": 0.0,
      "learning_rate": 4.03532083494395e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 3994
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.71875,
      "epoch": 0.1930324700425203,
      "grad_norm": 0.0,
      "learning_rate": 4.0350792423656744e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 3995
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.0625,
      "epoch": 0.19308078855817548,
      "grad_norm": 0.0,
      "learning_rate": 4.034837649787398e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 3996
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.4375,
      "epoch": 0.19312910707383069,
      "grad_norm": 0.0,
      "learning_rate": 4.0345960572091227e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 3997
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.1875,
      "epoch": 0.1931774255894859,
      "grad_norm": 0.0,
      "learning_rate": 4.0343544646308463e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 3998
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.53125,
      "epoch": 0.1932257441051411,
      "grad_norm": 0.0,
      "learning_rate": 4.0341128720525704e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 3999
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.1875,
      "epoch": 0.1932740626207963,
      "grad_norm": 0.0,
      "learning_rate": 4.0338712794742945e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4000
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.0,
      "epoch": 0.19332238113645148,
      "grad_norm": 0.0,
      "learning_rate": 4.0336296868960187e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4001
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.8125,
      "epoch": 0.19337069965210668,
      "grad_norm": 0.0,
      "learning_rate": 4.0333880943177423e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4002
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.5625,
      "epoch": 0.19341901816776189,
      "grad_norm": 0.0,
      "learning_rate": 4.033146501739466e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4003
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.46875,
      "epoch": 0.1934673366834171,
      "grad_norm": 0.0,
      "learning_rate": 4.0329049091611906e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4004
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 107.5625,
      "epoch": 0.1935156551990723,
      "grad_norm": 27.66830958439,
      "learning_rate": 4.032663316582914e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4005
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.5625,
      "epoch": 0.19356397371472747,
      "grad_norm": 0.0,
      "learning_rate": 4.0324217240046383e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4006
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.5,
      "epoch": 0.19361229223038268,
      "grad_norm": 0.0,
      "learning_rate": 4.0321801314263624e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4007
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.625,
      "epoch": 0.19366061074603788,
      "grad_norm": 0.0,
      "learning_rate": 4.0319385388480866e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4008
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.9375,
      "epoch": 0.19370892926169309,
      "grad_norm": 0.0,
      "learning_rate": 4.03169694626981e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4009
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.125,
      "epoch": 0.1937572477773483,
      "grad_norm": 0.0,
      "learning_rate": 4.031455353691535e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4010
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 111.8125,
      "epoch": 0.19380556629300347,
      "grad_norm": 17.860645547990597,
      "learning_rate": 4.0312137611132585e-07,
      "loss": 0.0,
      "reward": 1.484375,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4011
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.65625,
      "epoch": 0.19385388480865867,
      "grad_norm": 0.0,
      "learning_rate": 4.0309721685349826e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4012
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 112.96875,
      "epoch": 0.19390220332431388,
      "grad_norm": 0.0,
      "learning_rate": 4.030730575956706e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4013
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 106.78125,
      "epoch": 0.19395052183996908,
      "grad_norm": 29.458028472462207,
      "learning_rate": 4.0304889833784303e-07,
      "loss": 0.0,
      "reward": 1.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.03125,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 4014
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.71875,
      "epoch": 0.19399884035562429,
      "grad_norm": 0.0,
      "learning_rate": 4.0302473908001545e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4015
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.09375,
      "epoch": 0.19404715887127946,
      "grad_norm": 0.0,
      "learning_rate": 4.030005798221878e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4016
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.1875,
      "epoch": 0.19409547738693467,
      "grad_norm": 0.0,
      "learning_rate": 4.029764205643603e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4017
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.65625,
      "epoch": 0.19414379590258987,
      "grad_norm": 0.0,
      "learning_rate": 4.0295226130653264e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4018
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 105.6875,
      "epoch": 0.19419211441824508,
      "grad_norm": 7.981942650447057,
      "learning_rate": 4.0292810204870505e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4019
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 112.03125,
      "epoch": 0.19424043293390028,
      "grad_norm": 0.0,
      "learning_rate": 4.0290394279087746e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4020
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.59375,
      "epoch": 0.19428875144955546,
      "grad_norm": 0.0,
      "learning_rate": 4.028797835330499e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4021
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.8125,
      "epoch": 0.19433706996521066,
      "grad_norm": 0.0,
      "learning_rate": 4.0285562427522224e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4022
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 112.375,
      "epoch": 0.19438538848086587,
      "grad_norm": 0.0,
      "learning_rate": 4.028314650173946e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4023
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 114.0625,
      "epoch": 0.19443370699652107,
      "grad_norm": 0.0,
      "learning_rate": 4.0280730575956706e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 4024
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 122.90625,
      "epoch": 0.19448202551217628,
      "grad_norm": 0.0,
      "learning_rate": 4.027831465017394e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4025
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.34375,
      "epoch": 0.19453034402783145,
      "grad_norm": 0.0,
      "learning_rate": 4.0275898724391184e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4026
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.375,
      "epoch": 0.19457866254348666,
      "grad_norm": 0.0,
      "learning_rate": 4.0273482798608425e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4027
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.09375,
      "epoch": 0.19462698105914186,
      "grad_norm": 0.0,
      "learning_rate": 4.0271066872825667e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4028
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.125,
      "epoch": 0.19467529957479707,
      "grad_norm": 0.0,
      "learning_rate": 4.02686509470429e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4029
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 128.8125,
      "epoch": 0.19472361809045227,
      "grad_norm": 0.0,
      "learning_rate": 4.026623502126015e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4030
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.53125,
      "epoch": 0.19477193660610745,
      "grad_norm": 0.0,
      "learning_rate": 4.0263819095477385e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4031
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.59375,
      "epoch": 0.19482025512176265,
      "grad_norm": 0.0,
      "learning_rate": 4.0261403169694627e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4032
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 118.21875,
      "epoch": 0.19486857363741786,
      "grad_norm": 17.682479965049488,
      "learning_rate": 4.0258987243911863e-07,
      "loss": -0.0,
      "reward": 1.453125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4033
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 112.28125,
      "epoch": 0.19491689215307306,
      "grad_norm": 0.0,
      "learning_rate": 4.0256571318129104e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4034
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.4375,
      "epoch": 0.19496521066872827,
      "grad_norm": 0.0,
      "learning_rate": 4.0254155392346346e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4035
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.53125,
      "epoch": 0.19501352918438344,
      "grad_norm": 0.0,
      "learning_rate": 4.025173946656358e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4036
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.84375,
      "epoch": 0.19506184770003865,
      "grad_norm": 0.0,
      "learning_rate": 4.024932354078083e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4037
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.40625,
      "epoch": 0.19511016621569385,
      "grad_norm": 0.0,
      "learning_rate": 4.0246907614998064e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4038
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.0625,
      "epoch": 0.19515848473134906,
      "grad_norm": 0.0,
      "learning_rate": 4.0244491689215306e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4039
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.46875,
      "epoch": 0.19520680324700426,
      "grad_norm": 0.0,
      "learning_rate": 4.0242075763432547e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4040
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 108.5625,
      "epoch": 0.19525512176265944,
      "grad_norm": 5.607316512732246,
      "learning_rate": 4.023965983764979e-07,
      "loss": -0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4041
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.84375,
      "epoch": 0.19530344027831464,
      "grad_norm": 0.0,
      "learning_rate": 4.0237243911867024e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4042
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 119.71875,
      "epoch": 0.19535175879396985,
      "grad_norm": 0.0,
      "learning_rate": 4.0234827986084266e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4043
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.0,
      "epoch": 0.19540007730962505,
      "grad_norm": 0.0,
      "learning_rate": 4.0232412060301507e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4044
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.0625,
      "epoch": 0.19544839582528026,
      "grad_norm": 0.0,
      "learning_rate": 4.0229996134518743e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4045
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 116.9375,
      "epoch": 0.19549671434093543,
      "grad_norm": 0.0,
      "learning_rate": 4.0227580208735985e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 4046
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.90625,
      "epoch": 0.19554503285659064,
      "grad_norm": 0.0,
      "learning_rate": 4.0225164282953226e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4047
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 124.3125,
      "epoch": 0.19559335137224584,
      "grad_norm": 3.979329365121494,
      "learning_rate": 4.022274835717047e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4048
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 130.09375,
      "epoch": 0.19564166988790105,
      "grad_norm": 0.0,
      "learning_rate": 4.0220332431387703e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4049
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.6875,
      "epoch": 0.19568998840355625,
      "grad_norm": 0.0,
      "learning_rate": 4.021791650560495e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4050
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.625,
      "epoch": 0.19573830691921143,
      "grad_norm": 0.0,
      "learning_rate": 4.0215500579822186e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4051
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.9375,
      "epoch": 0.19578662543486663,
      "grad_norm": 0.0,
      "learning_rate": 4.021308465403943e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4052
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 104.5625,
      "epoch": 0.19583494395052184,
      "grad_norm": 6.806634080283921,
      "learning_rate": 4.0210668728256664e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4053
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.1875,
      "epoch": 0.19588326246617704,
      "grad_norm": 0.0,
      "learning_rate": 4.020825280247391e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 4054
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.96875,
      "epoch": 0.19593158098183225,
      "grad_norm": 0.0,
      "learning_rate": 4.0205836876691146e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4055
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.3125,
      "epoch": 0.19597989949748743,
      "grad_norm": 0.0,
      "learning_rate": 4.020342095090838e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4056
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.875,
      "epoch": 0.19602821801314263,
      "grad_norm": 0.0,
      "learning_rate": 4.020100502512563e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4057
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.5,
      "epoch": 0.19607653652879783,
      "grad_norm": 0.0,
      "learning_rate": 4.0198589099342865e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4058
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.125,
      "epoch": 0.19612485504445304,
      "grad_norm": 0.0,
      "learning_rate": 4.0196173173560106e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4059
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.125,
      "epoch": 0.19617317356010824,
      "grad_norm": 0.0,
      "learning_rate": 4.019375724777735e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4060
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 108.53125,
      "epoch": 0.19622149207576342,
      "grad_norm": 0.0,
      "learning_rate": 4.019134132199459e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4061
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.90625,
      "epoch": 0.19626981059141863,
      "grad_norm": 0.0,
      "learning_rate": 4.0188925396211825e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4062
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.5,
      "epoch": 0.19631812910707383,
      "grad_norm": 0.0,
      "learning_rate": 4.0186509470429067e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4063
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.875,
      "epoch": 0.19636644762272903,
      "grad_norm": 0.0,
      "learning_rate": 4.018409354464631e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 4064
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.25,
      "epoch": 0.19641476613838424,
      "grad_norm": 0.0,
      "learning_rate": 4.0181677618863544e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4065
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.0625,
      "epoch": 0.19646308465403942,
      "grad_norm": 0.0,
      "learning_rate": 4.0179261693080785e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4066
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 108.90625,
      "epoch": 0.19651140316969462,
      "grad_norm": 0.0,
      "learning_rate": 4.0176845767298027e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 4067
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 119.28125,
      "epoch": 0.19655972168534983,
      "grad_norm": 68.39006818853375,
      "learning_rate": 4.017442984151527e-07,
      "loss": -0.0,
      "reward": 2.265625,
      "reward_std": 0.2461063265800476,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4068
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 109.96875,
      "epoch": 0.19660804020100503,
      "grad_norm": 7.627462664416945,
      "learning_rate": 4.0172013915732504e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4069
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.21875,
      "epoch": 0.19665635871666023,
      "grad_norm": 0.0,
      "learning_rate": 4.016959798994975e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4070
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.59375,
      "epoch": 0.1967046772323154,
      "grad_norm": 0.0,
      "learning_rate": 4.0167182064166987e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4071
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.71875,
      "epoch": 0.19675299574797062,
      "grad_norm": 0.0,
      "learning_rate": 4.016476613838423e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4072
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.3125,
      "epoch": 0.19680131426362582,
      "grad_norm": 0.0,
      "learning_rate": 4.0162350212601464e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4073
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.875,
      "epoch": 0.19684963277928103,
      "grad_norm": 0.0,
      "learning_rate": 4.015993428681871e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4074
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.96875,
      "epoch": 0.19689795129493623,
      "grad_norm": 0.0,
      "learning_rate": 4.0157518361035947e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4075
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.21875,
      "epoch": 0.1969462698105914,
      "grad_norm": 0.0,
      "learning_rate": 4.0155102435253183e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4076
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.375,
      "epoch": 0.1969945883262466,
      "grad_norm": 0.0,
      "learning_rate": 4.015268650947043e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4077
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.25,
      "epoch": 0.19704290684190182,
      "grad_norm": 0.0,
      "learning_rate": 4.0150270583687666e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4078
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.3125,
      "epoch": 0.19709122535755702,
      "grad_norm": 0.0,
      "learning_rate": 4.0147854657904907e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4079
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.78125,
      "epoch": 0.19713954387321223,
      "grad_norm": 0.0,
      "learning_rate": 4.014543873212215e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4080
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.65625,
      "epoch": 0.1971878623888674,
      "grad_norm": 0.0,
      "learning_rate": 4.014302280633939e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4081
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 121.21875,
      "epoch": 0.1972361809045226,
      "grad_norm": 0.0,
      "learning_rate": 4.0140606880556626e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4082
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.8125,
      "epoch": 0.1972844994201778,
      "grad_norm": 0.0,
      "learning_rate": 4.013819095477387e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4083
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.65625,
      "epoch": 0.19733281793583302,
      "grad_norm": 0.0,
      "learning_rate": 4.013577502899111e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4084
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.3125,
      "epoch": 0.19738113645148822,
      "grad_norm": 0.0,
      "learning_rate": 4.0133359103208345e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4085
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 110.9375,
      "epoch": 0.1974294549671434,
      "grad_norm": 4.949759549560794,
      "learning_rate": 4.0130943177425586e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4086
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 106.46875,
      "epoch": 0.1974777734827986,
      "grad_norm": 27.20089856790266,
      "learning_rate": 4.012852725164283e-07,
      "loss": -0.0,
      "reward": 1.46875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4087
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.46875,
      "epoch": 0.1975260919984538,
      "grad_norm": 0.0,
      "learning_rate": 4.012611132586007e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4088
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.5625,
      "epoch": 0.197574410514109,
      "grad_norm": 0.0,
      "learning_rate": 4.0123695400077305e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4089
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.15625,
      "epoch": 0.19762272902976422,
      "grad_norm": 0.0,
      "learning_rate": 4.012127947429455e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4090
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.15625,
      "epoch": 0.1976710475454194,
      "grad_norm": 0.0,
      "learning_rate": 4.011886354851179e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4091
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.65625,
      "epoch": 0.1977193660610746,
      "grad_norm": 0.0,
      "learning_rate": 4.011644762272903e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4092
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.46875,
      "epoch": 0.1977676845767298,
      "grad_norm": 0.0,
      "learning_rate": 4.0114031696946265e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4093
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.4375,
      "epoch": 0.197816003092385,
      "grad_norm": 0.0,
      "learning_rate": 4.011161577116351e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4094
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.1875,
      "epoch": 0.1978643216080402,
      "grad_norm": 0.0,
      "learning_rate": 4.010919984538075e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4095
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.34375,
      "epoch": 0.1979126401236954,
      "grad_norm": 0.0,
      "learning_rate": 4.0106783919597984e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4096
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.125,
      "epoch": 0.1979609586393506,
      "grad_norm": 3.357673070409013,
      "learning_rate": 4.010436799381523e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4097
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.5,
      "epoch": 0.1980092771550058,
      "grad_norm": 0.0,
      "learning_rate": 4.0101952068032467e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4098
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.25,
      "epoch": 0.198057595670661,
      "grad_norm": 0.0,
      "learning_rate": 4.009953614224971e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4099
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.0,
      "epoch": 0.1981059141863162,
      "grad_norm": 0.0,
      "learning_rate": 4.009712021646695e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4100
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.875,
      "epoch": 0.19815423270197138,
      "grad_norm": 0.0,
      "learning_rate": 4.009470429068419e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4101
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 112.90625,
      "epoch": 0.1982025512176266,
      "grad_norm": 12.055033318031708,
      "learning_rate": 4.0092288364901427e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.17598573863506317,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4102
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.25,
      "epoch": 0.1982508697332818,
      "grad_norm": 0.0,
      "learning_rate": 4.008987243911867e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4103
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 112.15625,
      "epoch": 0.198299188248937,
      "grad_norm": 23.59982352422022,
      "learning_rate": 4.008745651333591e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4104
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.40625,
      "epoch": 0.1983475067645922,
      "grad_norm": 0.0,
      "learning_rate": 4.008504058755315e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4105
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.9375,
      "epoch": 0.19839582528024738,
      "grad_norm": 0.0,
      "learning_rate": 4.0082624661770387e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4106
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 110.125,
      "epoch": 0.19844414379590258,
      "grad_norm": 6.62803442125892,
      "learning_rate": 4.008020873598763e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.2962459325790405,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4107
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.78125,
      "epoch": 0.1984924623115578,
      "grad_norm": 0.0,
      "learning_rate": 4.007779281020487e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4108
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.03125,
      "epoch": 0.198540780827213,
      "grad_norm": 0.0,
      "learning_rate": 4.0075376884422106e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4109
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 116.59375,
      "epoch": 0.1985890993428682,
      "grad_norm": 5.891458928984166,
      "learning_rate": 4.007296095863935e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2226182520389557,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4110
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 121.21875,
      "epoch": 0.19863741785852337,
      "grad_norm": 3.6815303624669014,
      "learning_rate": 4.007054503285659e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4111
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 105.46875,
      "epoch": 0.19868573637417858,
      "grad_norm": 0.0,
      "learning_rate": 4.006812910707383e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 4112
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.09375,
      "epoch": 0.19873405488983378,
      "grad_norm": 0.0,
      "learning_rate": 4.0065713181291066e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4113
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.625,
      "epoch": 0.198782373405489,
      "grad_norm": 0.0,
      "learning_rate": 4.006329725550831e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4114
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 119.0,
      "epoch": 0.1988306919211442,
      "grad_norm": 6.604077754840974,
      "learning_rate": 4.006088132972555e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4115
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 108.8125,
      "epoch": 0.19887901043679937,
      "grad_norm": 4.4414663835939425,
      "learning_rate": 4.0058465403942785e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4116
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 115.8125,
      "epoch": 0.19892732895245457,
      "grad_norm": 62.794057290570024,
      "learning_rate": 4.005604947816003e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4117
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.46875,
      "epoch": 0.19897564746810978,
      "grad_norm": 0.0,
      "learning_rate": 4.005363355237727e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4118
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 109.375,
      "epoch": 0.19902396598376498,
      "grad_norm": 24.86688595683192,
      "learning_rate": 4.005121762659451e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4119
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.3125,
      "epoch": 0.1990722844994202,
      "grad_norm": 0.0,
      "learning_rate": 4.004880170081175e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4120
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.5625,
      "epoch": 0.19912060301507536,
      "grad_norm": 0.0,
      "learning_rate": 4.004638577502899e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4121
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.40625,
      "epoch": 0.19916892153073057,
      "grad_norm": 0.0,
      "learning_rate": 4.004396984924623e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4122
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 105.5,
      "epoch": 0.19921724004638577,
      "grad_norm": 5.180163116008707,
      "learning_rate": 4.004155392346347e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.18857255578041077,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4123
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 98.15625,
      "epoch": 0.19926555856204098,
      "grad_norm": 14.251101927797835,
      "learning_rate": 4.003913799768071e-07,
      "loss": -0.0,
      "reward": 2.34375,
      "reward_std": 0.23827512562274933,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4124
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.53125,
      "epoch": 0.19931387707769618,
      "grad_norm": 0.0,
      "learning_rate": 4.003672207189795e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4125
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.65625,
      "epoch": 0.19936219559335136,
      "grad_norm": 0.0,
      "learning_rate": 4.003430614611519e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4126
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 97.25,
      "epoch": 0.19941051410900656,
      "grad_norm": 10.57189623940122,
      "learning_rate": 4.003189022033243e-07,
      "loss": -0.0,
      "reward": 2.296875,
      "reward_std": 0.188242569565773,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4127
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.40625,
      "epoch": 0.19945883262466177,
      "grad_norm": 0.0,
      "learning_rate": 4.002947429454967e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4128
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.90625,
      "epoch": 0.19950715114031697,
      "grad_norm": 0.0,
      "learning_rate": 4.0027058368766907e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 4129
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.59375,
      "epoch": 0.19955546965597218,
      "grad_norm": 0.0,
      "learning_rate": 4.0024642442984153e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4130
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.875,
      "epoch": 0.19960378817162736,
      "grad_norm": 0.0,
      "learning_rate": 4.002222651720139e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4131
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.25,
      "epoch": 0.19965210668728256,
      "grad_norm": 0.0,
      "learning_rate": 4.001981059141863e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4132
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.09375,
      "epoch": 0.19970042520293776,
      "grad_norm": 0.0,
      "learning_rate": 4.0017394665635867e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4133
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.78125,
      "epoch": 0.19974874371859297,
      "grad_norm": 0.0,
      "learning_rate": 4.0014978739853113e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4134
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.46875,
      "epoch": 0.19979706223424817,
      "grad_norm": 0.0,
      "learning_rate": 4.001256281407035e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4135
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.34375,
      "epoch": 0.19984538074990335,
      "grad_norm": 0.0,
      "learning_rate": 4.0010146888287586e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4136
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.3125,
      "epoch": 0.19989369926555856,
      "grad_norm": 0.0,
      "learning_rate": 4.000773096250483e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4137
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.34375,
      "epoch": 0.19994201778121376,
      "grad_norm": 0.0,
      "learning_rate": 4.000531503672207e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4138
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.125,
      "epoch": 0.19999033629686896,
      "grad_norm": 0.0,
      "learning_rate": 4.000289911093931e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4139
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.03125,
      "epoch": 0.20003865481252417,
      "grad_norm": 0.0,
      "learning_rate": 4.000048318515655e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4140
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.9375,
      "epoch": 0.20008697332817935,
      "grad_norm": 0.0,
      "learning_rate": 3.999806725937379e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4141
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.5625,
      "epoch": 0.20013529184383455,
      "grad_norm": 0.0,
      "learning_rate": 3.999565133359103e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4142
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.375,
      "epoch": 0.20018361035948976,
      "grad_norm": 0.0,
      "learning_rate": 3.999323540780827e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4143
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.9375,
      "epoch": 0.20023192887514496,
      "grad_norm": 0.0,
      "learning_rate": 3.999081948202551e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4144
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.3125,
      "epoch": 0.20028024739080016,
      "grad_norm": 0.0,
      "learning_rate": 3.998840355624275e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4145
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.9375,
      "epoch": 0.20032856590645534,
      "grad_norm": 0.0,
      "learning_rate": 3.998598763045999e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4146
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.21875,
      "epoch": 0.20037688442211055,
      "grad_norm": 0.0,
      "learning_rate": 3.998357170467723e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4147
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.28125,
      "epoch": 0.20042520293776575,
      "grad_norm": 0.0,
      "learning_rate": 3.998115577889447e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4148
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.875,
      "epoch": 0.20047352145342096,
      "grad_norm": 0.0,
      "learning_rate": 3.997873985311171e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4149
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.9375,
      "epoch": 0.20052183996907616,
      "grad_norm": 0.0,
      "learning_rate": 3.9976323927328954e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4150
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.375,
      "epoch": 0.20057015848473134,
      "grad_norm": 0.0,
      "learning_rate": 3.997390800154619e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4151
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.9375,
      "epoch": 0.20061847700038654,
      "grad_norm": 0.0,
      "learning_rate": 3.997149207576343e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4152
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.0625,
      "epoch": 0.20066679551604175,
      "grad_norm": 0.0,
      "learning_rate": 3.996907614998067e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4153
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.9375,
      "epoch": 0.20071511403169695,
      "grad_norm": 0.0,
      "learning_rate": 3.9966660224197914e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4154
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.28125,
      "epoch": 0.20076343254735216,
      "grad_norm": 0.0,
      "learning_rate": 3.996424429841515e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4155
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.34375,
      "epoch": 0.20081175106300733,
      "grad_norm": 0.0,
      "learning_rate": 3.996182837263239e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4156
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.0625,
      "epoch": 0.20086006957866254,
      "grad_norm": 0.0,
      "learning_rate": 3.9959412446849633e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4157
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.84375,
      "epoch": 0.20090838809431774,
      "grad_norm": 0.0,
      "learning_rate": 3.995699652106687e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4158
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.84375,
      "epoch": 0.20095670660997295,
      "grad_norm": 0.0,
      "learning_rate": 3.995458059528411e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4159
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.6875,
      "epoch": 0.20100502512562815,
      "grad_norm": 0.0,
      "learning_rate": 3.995216466950135e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4160
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.40625,
      "epoch": 0.20105334364128333,
      "grad_norm": 0.0,
      "learning_rate": 3.9949748743718593e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4161
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.09375,
      "epoch": 0.20110166215693853,
      "grad_norm": 0.0,
      "learning_rate": 3.994733281793583e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4162
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.78125,
      "epoch": 0.20114998067259374,
      "grad_norm": 0.0,
      "learning_rate": 3.994491689215307e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4163
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.4375,
      "epoch": 0.20119829918824894,
      "grad_norm": 0.0,
      "learning_rate": 3.994250096637031e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 4164
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.375,
      "epoch": 0.20124661770390415,
      "grad_norm": 0.0,
      "learning_rate": 3.9940085040587553e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4165
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.3125,
      "epoch": 0.20129493621955932,
      "grad_norm": 0.0,
      "learning_rate": 3.993766911480479e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4166
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.96875,
      "epoch": 0.20134325473521453,
      "grad_norm": 0.0,
      "learning_rate": 3.9935253189022036e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4167
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.0625,
      "epoch": 0.20139157325086973,
      "grad_norm": 0.0,
      "learning_rate": 3.993283726323927e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4168
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.34375,
      "epoch": 0.20143989176652494,
      "grad_norm": 0.0,
      "learning_rate": 3.993042133745651e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4169
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.8125,
      "epoch": 0.20148821028218014,
      "grad_norm": 0.0,
      "learning_rate": 3.9928005411673755e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4170
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.25,
      "epoch": 0.20153652879783532,
      "grad_norm": 0.0,
      "learning_rate": 3.992558948589099e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4171
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 86.75,
      "epoch": 0.20158484731349052,
      "grad_norm": 14.312070570721248,
      "learning_rate": 3.992317356010823e-07,
      "loss": 0.0,
      "reward": 1.265625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4172
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 87.25,
      "epoch": 0.20163316582914573,
      "grad_norm": 7.323525744359981,
      "learning_rate": 3.992075763432547e-07,
      "loss": -0.0,
      "reward": 1.625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4173
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.84375,
      "epoch": 0.20168148434480093,
      "grad_norm": 0.0,
      "learning_rate": 3.9918341708542715e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4174
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.1875,
      "epoch": 0.20172980286045614,
      "grad_norm": 0.0,
      "learning_rate": 3.991592578275995e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4175
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.15625,
      "epoch": 0.2017781213761113,
      "grad_norm": 0.0,
      "learning_rate": 3.991350985697719e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4176
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.90625,
      "epoch": 0.20182643989176652,
      "grad_norm": 0.0,
      "learning_rate": 3.9911093931194434e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4177
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.8125,
      "epoch": 0.20187475840742172,
      "grad_norm": 0.0,
      "learning_rate": 3.990867800541167e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 4178
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.78125,
      "epoch": 0.20192307692307693,
      "grad_norm": 0.0,
      "learning_rate": 3.990626207962891e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4179
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.0,
      "epoch": 0.20197139543873213,
      "grad_norm": 0.0,
      "learning_rate": 3.990384615384615e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4180
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.625,
      "epoch": 0.2020197139543873,
      "grad_norm": 0.0,
      "learning_rate": 3.9901430228063394e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4181
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.625,
      "epoch": 0.2020680324700425,
      "grad_norm": 0.0,
      "learning_rate": 3.989901430228063e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4182
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.6875,
      "epoch": 0.20211635098569772,
      "grad_norm": 0.0,
      "learning_rate": 3.989659837649787e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4183
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.34375,
      "epoch": 0.20216466950135292,
      "grad_norm": 0.0,
      "learning_rate": 3.9894182450715113e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4184
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.125,
      "epoch": 0.20221298801700813,
      "grad_norm": 0.0,
      "learning_rate": 3.9891766524932354e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4185
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.84375,
      "epoch": 0.2022613065326633,
      "grad_norm": 0.0,
      "learning_rate": 3.988935059914959e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4186
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.75,
      "epoch": 0.2023096250483185,
      "grad_norm": 0.0,
      "learning_rate": 3.9886934673366837e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4187
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.84375,
      "epoch": 0.2023579435639737,
      "grad_norm": 0.0,
      "learning_rate": 3.9884518747584073e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4188
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.53125,
      "epoch": 0.20240626207962892,
      "grad_norm": 0.0,
      "learning_rate": 3.988210282180131e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4189
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.34375,
      "epoch": 0.20245458059528412,
      "grad_norm": 0.0,
      "learning_rate": 3.9879686896018556e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4190
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.65625,
      "epoch": 0.2025028991109393,
      "grad_norm": 0.0,
      "learning_rate": 3.987727097023579e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4191
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.59375,
      "epoch": 0.2025512176265945,
      "grad_norm": 0.0,
      "learning_rate": 3.9874855044453033e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4192
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.875,
      "epoch": 0.2025995361422497,
      "grad_norm": 0.0,
      "learning_rate": 3.987243911867027e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4193
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.03125,
      "epoch": 0.2026478546579049,
      "grad_norm": 0.0,
      "learning_rate": 3.9870023192887516e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4194
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.625,
      "epoch": 0.20269617317356012,
      "grad_norm": 0.0,
      "learning_rate": 3.986760726710475e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4195
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 94.875,
      "epoch": 0.2027444916892153,
      "grad_norm": 15.522647929140064,
      "learning_rate": 3.9865191341321993e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.23384949564933777,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4196
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.15625,
      "epoch": 0.2027928102048705,
      "grad_norm": 0.0,
      "learning_rate": 3.9862775415539235e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4197
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.25,
      "epoch": 0.2028411287205257,
      "grad_norm": 0.0,
      "learning_rate": 3.986035948975647e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4198
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.40625,
      "epoch": 0.2028894472361809,
      "grad_norm": 0.0,
      "learning_rate": 3.985794356397371e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4199
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.40625,
      "epoch": 0.2029377657518361,
      "grad_norm": 0.0,
      "learning_rate": 3.9855527638190953e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4200
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.03125,
      "epoch": 0.2029860842674913,
      "grad_norm": 0.0,
      "learning_rate": 3.9853111712408195e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4201
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.71875,
      "epoch": 0.2030344027831465,
      "grad_norm": 0.0,
      "learning_rate": 3.985069578662543e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4202
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.59375,
      "epoch": 0.2030827212988017,
      "grad_norm": 0.0,
      "learning_rate": 3.984827986084267e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4203
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.625,
      "epoch": 0.2031310398144569,
      "grad_norm": 131.2043885796678,
      "learning_rate": 3.9845863935059913e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4204
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.46875,
      "epoch": 0.2031793583301121,
      "grad_norm": 0.0,
      "learning_rate": 3.9843448009277155e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4205
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.4375,
      "epoch": 0.20322767684576729,
      "grad_norm": 0.0,
      "learning_rate": 3.984103208349439e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 4206
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.40625,
      "epoch": 0.2032759953614225,
      "grad_norm": 0.0,
      "learning_rate": 3.983861615771164e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4207
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.75,
      "epoch": 0.2033243138770777,
      "grad_norm": 0.0,
      "learning_rate": 3.9836200231928874e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4208
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.40625,
      "epoch": 0.2033726323927329,
      "grad_norm": 0.0,
      "learning_rate": 3.983378430614611e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4209
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.15625,
      "epoch": 0.2034209509083881,
      "grad_norm": 0.0,
      "learning_rate": 3.9831368380363356e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4210
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.46875,
      "epoch": 0.20346926942404328,
      "grad_norm": 0.0,
      "learning_rate": 3.982895245458059e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4211
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.0625,
      "epoch": 0.20351758793969849,
      "grad_norm": 0.0,
      "learning_rate": 3.9826536528797834e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4212
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.125,
      "epoch": 0.2035659064553537,
      "grad_norm": 0.0,
      "learning_rate": 3.982412060301507e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4213
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.1875,
      "epoch": 0.2036142249710089,
      "grad_norm": 0.0,
      "learning_rate": 3.9821704677232317e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4214
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.71875,
      "epoch": 0.2036625434866641,
      "grad_norm": 0.0,
      "learning_rate": 3.981928875144955e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4215
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.15625,
      "epoch": 0.20371086200231928,
      "grad_norm": 0.0,
      "learning_rate": 3.9816872825666794e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4216
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.34375,
      "epoch": 0.20375918051797448,
      "grad_norm": 0.0,
      "learning_rate": 3.9814456899884035e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4217
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.0,
      "epoch": 0.20380749903362969,
      "grad_norm": 0.0,
      "learning_rate": 3.9812040974101277e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4218
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.375,
      "epoch": 0.2038558175492849,
      "grad_norm": 0.0,
      "learning_rate": 3.9809625048318513e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4219
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.59375,
      "epoch": 0.2039041360649401,
      "grad_norm": 0.0,
      "learning_rate": 3.9807209122535754e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4220
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.5625,
      "epoch": 0.20395245458059527,
      "grad_norm": 18.30654078004325,
      "learning_rate": 3.9804793196752995e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4221
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.625,
      "epoch": 0.20400077309625048,
      "grad_norm": 0.0,
      "learning_rate": 3.980237727097023e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4222
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.03125,
      "epoch": 0.20404909161190568,
      "grad_norm": 0.0,
      "learning_rate": 3.9799961345187473e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4223
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.4375,
      "epoch": 0.20409741012756089,
      "grad_norm": 0.0,
      "learning_rate": 3.9797545419404714e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4224
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.0,
      "epoch": 0.2041457286432161,
      "grad_norm": 0.0,
      "learning_rate": 3.9795129493621956e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4225
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.8125,
      "epoch": 0.20419404715887127,
      "grad_norm": 0.0,
      "learning_rate": 3.979271356783919e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4226
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.625,
      "epoch": 0.20424236567452647,
      "grad_norm": 0.0,
      "learning_rate": 3.979029764205644e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4227
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.375,
      "epoch": 0.20429068419018168,
      "grad_norm": 29.869608447008932,
      "learning_rate": 3.9787881716273674e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4228
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.75,
      "epoch": 0.20433900270583688,
      "grad_norm": 0.0,
      "learning_rate": 3.978546579049091e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4229
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.5,
      "epoch": 0.20438732122149209,
      "grad_norm": 0.0,
      "learning_rate": 3.9783049864708157e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4230
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.15625,
      "epoch": 0.20443563973714726,
      "grad_norm": 0.0,
      "learning_rate": 3.9780633938925393e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4231
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.03125,
      "epoch": 0.20448395825280247,
      "grad_norm": 0.0,
      "learning_rate": 3.9778218013142635e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4232
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.28125,
      "epoch": 0.20453227676845767,
      "grad_norm": 0.0,
      "learning_rate": 3.977580208735987e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4233
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.59375,
      "epoch": 0.20458059528411288,
      "grad_norm": 0.0,
      "learning_rate": 3.9773386161577117e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4234
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.28125,
      "epoch": 0.20462891379976808,
      "grad_norm": 0.0,
      "learning_rate": 3.9770970235794353e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4235
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.59375,
      "epoch": 0.20467723231542326,
      "grad_norm": 0.0,
      "learning_rate": 3.9768554310011595e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4236
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.84375,
      "epoch": 0.20472555083107846,
      "grad_norm": 0.0,
      "learning_rate": 3.9766138384228836e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4237
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.21875,
      "epoch": 0.20477386934673367,
      "grad_norm": 0.0,
      "learning_rate": 3.976372245844608e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4238
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.21875,
      "epoch": 0.20482218786238887,
      "grad_norm": 0.0,
      "learning_rate": 3.9761306532663314e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4239
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.84375,
      "epoch": 0.20487050637804408,
      "grad_norm": 0.0,
      "learning_rate": 3.9758890606880555e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4240
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.125,
      "epoch": 0.20491882489369925,
      "grad_norm": 0.0,
      "learning_rate": 3.9756474681097796e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4241
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.03125,
      "epoch": 0.20496714340935446,
      "grad_norm": 0.0,
      "learning_rate": 3.975405875531503e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4242
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.25,
      "epoch": 0.20501546192500966,
      "grad_norm": 0.0,
      "learning_rate": 3.9751642829532274e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4243
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.46875,
      "epoch": 0.20506378044066487,
      "grad_norm": 0.0,
      "learning_rate": 3.9749226903749515e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4244
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.84375,
      "epoch": 0.20511209895632007,
      "grad_norm": 0.0,
      "learning_rate": 3.9746810977966756e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4245
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.3125,
      "epoch": 0.20516041747197525,
      "grad_norm": 0.0,
      "learning_rate": 3.974439505218399e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4246
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.90625,
      "epoch": 0.20520873598763045,
      "grad_norm": 0.0,
      "learning_rate": 3.974197912640124e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 4247
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 86.375,
      "epoch": 0.20525705450328566,
      "grad_norm": 5.9163631193849255,
      "learning_rate": 3.9739563200618475e-07,
      "loss": -0.0,
      "reward": 2.328125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4248
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.15625,
      "epoch": 0.20530537301894086,
      "grad_norm": 0.0,
      "learning_rate": 3.973714727483571e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4249
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.5,
      "epoch": 0.20535369153459607,
      "grad_norm": 0.0,
      "learning_rate": 3.973473134905296e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4250
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.5,
      "epoch": 0.20540201005025124,
      "grad_norm": 0.0,
      "learning_rate": 3.9732315423270194e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4251
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.0,
      "epoch": 0.20545032856590645,
      "grad_norm": 0.0,
      "learning_rate": 3.9729899497487435e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4252
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.09375,
      "epoch": 0.20549864708156165,
      "grad_norm": 0.0,
      "learning_rate": 3.972748357170467e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4253
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.5,
      "epoch": 0.20554696559721686,
      "grad_norm": 0.0,
      "learning_rate": 3.972506764592192e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4254
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.5,
      "epoch": 0.20559528411287206,
      "grad_norm": 0.0,
      "learning_rate": 3.9722651720139154e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4255
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.3125,
      "epoch": 0.20564360262852724,
      "grad_norm": 0.0,
      "learning_rate": 3.9720235794356396e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4256
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.5625,
      "epoch": 0.20569192114418244,
      "grad_norm": 0.0,
      "learning_rate": 3.9717819868573637e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4257
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.25,
      "epoch": 0.20574023965983765,
      "grad_norm": 0.0,
      "learning_rate": 3.971540394279088e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 4258
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.25,
      "epoch": 0.20578855817549285,
      "grad_norm": 0.0,
      "learning_rate": 3.9712988017008114e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4259
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.28125,
      "epoch": 0.20583687669114806,
      "grad_norm": 0.0,
      "learning_rate": 3.9710572091225356e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4260
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.09375,
      "epoch": 0.20588519520680323,
      "grad_norm": 6.763567608861141,
      "learning_rate": 3.9708156165442597e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4261
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.46875,
      "epoch": 0.20593351372245844,
      "grad_norm": 0.0,
      "learning_rate": 3.9705740239659833e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4262
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.5625,
      "epoch": 0.20598183223811364,
      "grad_norm": 0.0,
      "learning_rate": 3.9703324313877074e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4263
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.125,
      "epoch": 0.20603015075376885,
      "grad_norm": 0.0,
      "learning_rate": 3.9700908388094316e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4264
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.53125,
      "epoch": 0.20607846926942405,
      "grad_norm": 0.0,
      "learning_rate": 3.9698492462311557e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4265
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.1875,
      "epoch": 0.20612678778507923,
      "grad_norm": 0.0,
      "learning_rate": 3.9696076536528793e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4266
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.9375,
      "epoch": 0.20617510630073443,
      "grad_norm": 0.0,
      "learning_rate": 3.969366061074604e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4267
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.34375,
      "epoch": 0.20622342481638964,
      "grad_norm": 0.0,
      "learning_rate": 3.9691244684963276e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4268
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.09375,
      "epoch": 0.20627174333204484,
      "grad_norm": 0.0,
      "learning_rate": 3.968882875918052e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4269
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.84375,
      "epoch": 0.20632006184770005,
      "grad_norm": 0.0,
      "learning_rate": 3.968641283339776e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4270
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.46875,
      "epoch": 0.20636838036335522,
      "grad_norm": 0.0,
      "learning_rate": 3.9683996907614995e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 4271
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.1875,
      "epoch": 0.20641669887901043,
      "grad_norm": 24.24114782140374,
      "learning_rate": 3.9681580981832236e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.3608423173427582,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4272
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.3125,
      "epoch": 0.20646501739466563,
      "grad_norm": 0.0,
      "learning_rate": 3.967916505604947e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4273
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 89.65625,
      "epoch": 0.20651333591032084,
      "grad_norm": 17.860589509023775,
      "learning_rate": 3.967674913026672e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4274
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.84375,
      "epoch": 0.20656165442597604,
      "grad_norm": 5.162004863984396,
      "learning_rate": 3.9674333204483955e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039678275585175,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4275
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.25,
      "epoch": 0.20660997294163122,
      "grad_norm": 0.0,
      "learning_rate": 3.9671917278701196e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4276
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.90625,
      "epoch": 0.20665829145728642,
      "grad_norm": 0.0,
      "learning_rate": 3.966950135291844e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4277
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.5,
      "epoch": 0.20670660997294163,
      "grad_norm": 0.0,
      "learning_rate": 3.966708542713568e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4278
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.78125,
      "epoch": 0.20675492848859683,
      "grad_norm": 0.0,
      "learning_rate": 3.9664669501352915e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4279
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.84375,
      "epoch": 0.20680324700425204,
      "grad_norm": 5.9993095270098875,
      "learning_rate": 3.966225357557016e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4280
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.75,
      "epoch": 0.20685156551990722,
      "grad_norm": 0.0,
      "learning_rate": 3.96598376497874e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4281
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.78125,
      "epoch": 0.20689988403556242,
      "grad_norm": 0.0,
      "learning_rate": 3.9657421724004634e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4282
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.125,
      "epoch": 0.20694820255121762,
      "grad_norm": 0.0,
      "learning_rate": 3.9655005798221875e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4283
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 98.40625,
      "epoch": 0.20699652106687283,
      "grad_norm": 0.0,
      "learning_rate": 3.9652589872439117e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4284
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.65625,
      "epoch": 0.20704483958252803,
      "grad_norm": 0.0,
      "learning_rate": 3.965017394665636e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 4285
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 107.1875,
      "epoch": 0.2070931580981832,
      "grad_norm": 22.987602099493827,
      "learning_rate": 3.9647758020873594e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4286
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.25,
      "epoch": 0.20714147661383842,
      "grad_norm": 0.0,
      "learning_rate": 3.964534209509084e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4287
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.9375,
      "epoch": 0.20718979512949362,
      "grad_norm": 0.0,
      "learning_rate": 3.9642926169308077e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4288
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.53125,
      "epoch": 0.20723811364514882,
      "grad_norm": 0.0,
      "learning_rate": 3.964051024352532e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4289
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 96.25,
      "epoch": 0.20728643216080403,
      "grad_norm": 6.646721191604483,
      "learning_rate": 3.963809431774256e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4290
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.40625,
      "epoch": 0.2073347506764592,
      "grad_norm": 0.0,
      "learning_rate": 3.9635678391959796e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4291
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.4375,
      "epoch": 0.2073830691921144,
      "grad_norm": 0.0,
      "learning_rate": 3.9633262466177037e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4292
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.28125,
      "epoch": 0.20743138770776962,
      "grad_norm": 0.0,
      "learning_rate": 3.9630846540394273e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4293
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.28125,
      "epoch": 0.20747970622342482,
      "grad_norm": 0.0,
      "learning_rate": 3.962843061461152e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4294
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 110.625,
      "epoch": 0.20752802473908002,
      "grad_norm": 0.0,
      "learning_rate": 3.9626014688828756e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 4295
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 111.03125,
      "epoch": 0.2075763432547352,
      "grad_norm": 0.0,
      "learning_rate": 3.9623598763045997e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4296
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.53125,
      "epoch": 0.2076246617703904,
      "grad_norm": 0.0,
      "learning_rate": 3.962118283726324e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4297
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.25,
      "epoch": 0.2076729802860456,
      "grad_norm": 0.0,
      "learning_rate": 3.961876691148048e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4298
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 110.1875,
      "epoch": 0.20772129880170082,
      "grad_norm": 13.566931051460951,
      "learning_rate": 3.9616350985697716e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.323003351688385,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4299
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.03125,
      "epoch": 0.20776961731735602,
      "grad_norm": 0.0,
      "learning_rate": 3.961393505991496e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4300
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.28125,
      "epoch": 0.2078179358330112,
      "grad_norm": 0.0,
      "learning_rate": 3.96115191341322e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4301
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 118.375,
      "epoch": 0.2078662543486664,
      "grad_norm": 0.0,
      "learning_rate": 3.9609103208349435e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4302
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 113.59375,
      "epoch": 0.2079145728643216,
      "grad_norm": 0.0,
      "learning_rate": 3.9606687282566676e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 4303
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.40625,
      "epoch": 0.2079628913799768,
      "grad_norm": 0.0,
      "learning_rate": 3.960427135678392e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4304
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.84375,
      "epoch": 0.20801120989563202,
      "grad_norm": 0.0,
      "learning_rate": 3.960185543100116e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4305
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.34375,
      "epoch": 0.2080595284112872,
      "grad_norm": 0.0,
      "learning_rate": 3.9599439505218395e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4306
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.125,
      "epoch": 0.2081078469269424,
      "grad_norm": 0.0,
      "learning_rate": 3.959702357943564e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4307
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.25,
      "epoch": 0.2081561654425976,
      "grad_norm": 0.0,
      "learning_rate": 3.959460765365288e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4308
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.25,
      "epoch": 0.2082044839582528,
      "grad_norm": 0.0,
      "learning_rate": 3.959219172787012e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4309
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.21875,
      "epoch": 0.208252802473908,
      "grad_norm": 0.0,
      "learning_rate": 3.958977580208736e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4310
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 114.21875,
      "epoch": 0.2083011209895632,
      "grad_norm": 0.0,
      "learning_rate": 3.95873598763046e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4311
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.75,
      "epoch": 0.2083494395052184,
      "grad_norm": 0.0,
      "learning_rate": 3.958494395052184e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4312
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.15625,
      "epoch": 0.2083977580208736,
      "grad_norm": 0.0,
      "learning_rate": 3.9582528024739074e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4313
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.65625,
      "epoch": 0.2084460765365288,
      "grad_norm": 4.751326238115716,
      "learning_rate": 3.958011209895632e-07,
      "loss": -0.0,
      "reward": 1.921875,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4314
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.46875,
      "epoch": 0.208494395052184,
      "grad_norm": 0.0,
      "learning_rate": 3.9577696173173557e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4315
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 96.0625,
      "epoch": 0.20854271356783918,
      "grad_norm": 7.795745993435483,
      "learning_rate": 3.95752802473908e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4316
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.875,
      "epoch": 0.2085910320834944,
      "grad_norm": 0.0,
      "learning_rate": 3.957286432160804e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4317
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.9375,
      "epoch": 0.2086393505991496,
      "grad_norm": 0.0,
      "learning_rate": 3.957044839582528e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4318
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.46875,
      "epoch": 0.2086876691148048,
      "grad_norm": 0.0,
      "learning_rate": 3.9568032470042517e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4319
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.125,
      "epoch": 0.20873598763046,
      "grad_norm": 0.0,
      "learning_rate": 3.9565616544259763e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23568853735923767,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4320
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 109.75,
      "epoch": 0.20878430614611518,
      "grad_norm": 0.0,
      "learning_rate": 3.9563200618477e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4321
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.0,
      "epoch": 0.20883262466177038,
      "grad_norm": 0.0,
      "learning_rate": 3.9560784692694235e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4322
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.96875,
      "epoch": 0.2088809431774256,
      "grad_norm": 0.0,
      "learning_rate": 3.9558368766911477e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4323
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.5625,
      "epoch": 0.2089292616930808,
      "grad_norm": 0.0,
      "learning_rate": 3.955595284112872e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4324
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.0,
      "epoch": 0.208977580208736,
      "grad_norm": 0.0,
      "learning_rate": 3.955353691534596e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4325
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.25,
      "epoch": 0.20902589872439117,
      "grad_norm": 0.0,
      "learning_rate": 3.9551120989563196e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4326
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.78125,
      "epoch": 0.20907421724004638,
      "grad_norm": 0.0,
      "learning_rate": 3.954870506378044e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4327
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.375,
      "epoch": 0.20912253575570158,
      "grad_norm": 0.0,
      "learning_rate": 3.954628913799768e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4328
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 104.21875,
      "epoch": 0.2091708542713568,
      "grad_norm": 0.0,
      "learning_rate": 3.954387321221492e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4329
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.75,
      "epoch": 0.209219172787012,
      "grad_norm": 0.0,
      "learning_rate": 3.954145728643216e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4330
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 108.5,
      "epoch": 0.20926749130266717,
      "grad_norm": 4.887680274067017,
      "learning_rate": 3.95390413606494e-07,
      "loss": -0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 4331
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.46875,
      "epoch": 0.20931580981832237,
      "grad_norm": 0.0,
      "learning_rate": 3.953662543486664e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4332
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 104.5,
      "epoch": 0.20936412833397758,
      "grad_norm": 8.176096133993928,
      "learning_rate": 3.9534209509083875e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4333
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.5,
      "epoch": 0.20941244684963278,
      "grad_norm": 0.0,
      "learning_rate": 3.953179358330112e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4334
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.21875,
      "epoch": 0.209460765365288,
      "grad_norm": 0.0,
      "learning_rate": 3.9529377657518357e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4335
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.0,
      "epoch": 0.20950908388094316,
      "grad_norm": 0.0,
      "learning_rate": 3.95269617317356e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 4336
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.5,
      "epoch": 0.20955740239659837,
      "grad_norm": 0.0,
      "learning_rate": 3.952454580595284e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4337
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 102.75,
      "epoch": 0.20960572091225357,
      "grad_norm": 6.880712399025246,
      "learning_rate": 3.952212988017008e-07,
      "loss": -0.0,
      "reward": 1.671875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4338
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.34375,
      "epoch": 0.20965403942790878,
      "grad_norm": 0.0,
      "learning_rate": 3.951971395438732e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4339
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.25,
      "epoch": 0.20970235794356398,
      "grad_norm": 0.0,
      "learning_rate": 3.9517298028604564e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4340
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 113.21875,
      "epoch": 0.20975067645921916,
      "grad_norm": 6.6237115133615685,
      "learning_rate": 3.95148821028218e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.20785759389400482,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4341
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.21875,
      "epoch": 0.20979899497487436,
      "grad_norm": 0.0,
      "learning_rate": 3.9512466177039036e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4342
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.28125,
      "epoch": 0.20984731349052957,
      "grad_norm": 0.0,
      "learning_rate": 3.951005025125628e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4343
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 104.0625,
      "epoch": 0.20989563200618477,
      "grad_norm": 52.6659368740292,
      "learning_rate": 3.950763432547352e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2848735451698303,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4344
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.8125,
      "epoch": 0.20994395052183998,
      "grad_norm": 0.0,
      "learning_rate": 3.950521839969076e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4345
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.53125,
      "epoch": 0.20999226903749516,
      "grad_norm": 0.0,
      "learning_rate": 3.9502802473907996e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4346
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.21875,
      "epoch": 0.21004058755315036,
      "grad_norm": 0.0,
      "learning_rate": 3.9500386548125243e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4347
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.25,
      "epoch": 0.21008890606880556,
      "grad_norm": 0.0,
      "learning_rate": 3.949797062234248e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4348
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.34375,
      "epoch": 0.21013722458446077,
      "grad_norm": 0.0,
      "learning_rate": 3.949555469655972e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4349
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.8125,
      "epoch": 0.21018554310011597,
      "grad_norm": 0.0,
      "learning_rate": 3.949313877077696e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4350
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.5625,
      "epoch": 0.21023386161577115,
      "grad_norm": 0.0,
      "learning_rate": 3.9490722844994203e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4351
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.375,
      "epoch": 0.21028218013142636,
      "grad_norm": 0.0,
      "learning_rate": 3.948830691921144e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4352
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 117.78125,
      "epoch": 0.21033049864708156,
      "grad_norm": 16.472626921193235,
      "learning_rate": 3.9485890993428675e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.2961388826370239,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4353
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 108.25,
      "epoch": 0.21037881716273676,
      "grad_norm": 0.0,
      "learning_rate": 3.948347506764592e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4354
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.5,
      "epoch": 0.21042713567839197,
      "grad_norm": 0.0,
      "learning_rate": 3.948105914186316e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4355
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 111.28125,
      "epoch": 0.21047545419404715,
      "grad_norm": 0.0,
      "learning_rate": 3.94786432160804e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4356
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.625,
      "epoch": 0.21052377270970235,
      "grad_norm": 0.0,
      "learning_rate": 3.947622729029764e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4357
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.3125,
      "epoch": 0.21057209122535755,
      "grad_norm": 0.0,
      "learning_rate": 3.947381136451488e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4358
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.75,
      "epoch": 0.21062040974101276,
      "grad_norm": 0.0,
      "learning_rate": 3.947139543873212e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4359
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.75,
      "epoch": 0.21066872825666796,
      "grad_norm": 0.0,
      "learning_rate": 3.9468979512949365e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4360
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.0625,
      "epoch": 0.21071704677232314,
      "grad_norm": 0.0,
      "learning_rate": 3.94665635871666e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4361
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.84375,
      "epoch": 0.21076536528797835,
      "grad_norm": 0.0,
      "learning_rate": 3.946414766138384e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4362
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.25,
      "epoch": 0.21081368380363355,
      "grad_norm": 0.0,
      "learning_rate": 3.946173173560108e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4363
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.4375,
      "epoch": 0.21086200231928875,
      "grad_norm": 0.0,
      "learning_rate": 3.945931580981832e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4364
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.4375,
      "epoch": 0.21091032083494396,
      "grad_norm": 0.0,
      "learning_rate": 3.945689988403556e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4365
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.59375,
      "epoch": 0.21095863935059914,
      "grad_norm": 0.0,
      "learning_rate": 3.9454483958252797e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4366
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.15625,
      "epoch": 0.21100695786625434,
      "grad_norm": 0.0,
      "learning_rate": 3.9452068032470044e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4367
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.65625,
      "epoch": 0.21105527638190955,
      "grad_norm": 0.0,
      "learning_rate": 3.944965210668728e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4368
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.5625,
      "epoch": 0.21110359489756475,
      "grad_norm": 0.0,
      "learning_rate": 3.944723618090452e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4369
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.46875,
      "epoch": 0.21115191341321995,
      "grad_norm": 0.0,
      "learning_rate": 3.9444820255121763e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4370
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 116.25,
      "epoch": 0.21120023192887513,
      "grad_norm": 19.74036572923435,
      "learning_rate": 3.9442404329339004e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.23421898484230042,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4371
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 114.90625,
      "epoch": 0.21124855044453034,
      "grad_norm": 0.0,
      "learning_rate": 3.943998840355624e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4372
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.65625,
      "epoch": 0.21129686896018554,
      "grad_norm": 0.0,
      "learning_rate": 3.9437572477773476e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4373
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.65625,
      "epoch": 0.21134518747584075,
      "grad_norm": 0.0,
      "learning_rate": 3.9435156551990723e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4374
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.5625,
      "epoch": 0.21139350599149595,
      "grad_norm": 0.0,
      "learning_rate": 3.943274062620796e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4375
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.53125,
      "epoch": 0.21144182450715113,
      "grad_norm": 0.0,
      "learning_rate": 3.94303247004252e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4376
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.09375,
      "epoch": 0.21149014302280633,
      "grad_norm": 0.0,
      "learning_rate": 3.942790877464244e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4377
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.78125,
      "epoch": 0.21153846153846154,
      "grad_norm": 0.0,
      "learning_rate": 3.9425492848859683e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4378
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 112.84375,
      "epoch": 0.21158678005411674,
      "grad_norm": 0.0,
      "learning_rate": 3.942307692307692e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4379
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.75,
      "epoch": 0.21163509856977195,
      "grad_norm": 0.0,
      "learning_rate": 3.9420660997294166e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 4380
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.1875,
      "epoch": 0.21168341708542712,
      "grad_norm": 0.0,
      "learning_rate": 3.94182450715114e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4381
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.25,
      "epoch": 0.21173173560108233,
      "grad_norm": 0.0,
      "learning_rate": 3.9415829145728643e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4382
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.75,
      "epoch": 0.21178005411673753,
      "grad_norm": 0.0,
      "learning_rate": 3.941341321994588e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4383
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.75,
      "epoch": 0.21182837263239274,
      "grad_norm": 0.0,
      "learning_rate": 3.941099729416312e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4384
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.34375,
      "epoch": 0.21187669114804794,
      "grad_norm": 0.0,
      "learning_rate": 3.940858136838036e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4385
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.84375,
      "epoch": 0.21192500966370312,
      "grad_norm": 0.0,
      "learning_rate": 3.94061654425976e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4386
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.53125,
      "epoch": 0.21197332817935832,
      "grad_norm": 0.0,
      "learning_rate": 3.9403749516814845e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4387
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.78125,
      "epoch": 0.21202164669501353,
      "grad_norm": 0.0,
      "learning_rate": 3.940133359103208e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4388
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.15625,
      "epoch": 0.21206996521066873,
      "grad_norm": 0.0,
      "learning_rate": 3.939891766524932e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4389
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.53125,
      "epoch": 0.21211828372632394,
      "grad_norm": 0.0,
      "learning_rate": 3.9396501739466563e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4390
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.1875,
      "epoch": 0.2121666022419791,
      "grad_norm": 0.0,
      "learning_rate": 3.9394085813683805e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4391
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.59375,
      "epoch": 0.21221492075763432,
      "grad_norm": 0.0,
      "learning_rate": 3.939166988790104e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4392
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.9375,
      "epoch": 0.21226323927328952,
      "grad_norm": 0.0,
      "learning_rate": 3.9389253962118277e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4393
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.8125,
      "epoch": 0.21231155778894473,
      "grad_norm": 0.0,
      "learning_rate": 3.9386838036335524e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4394
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.375,
      "epoch": 0.21235987630459993,
      "grad_norm": 0.0,
      "learning_rate": 3.938442211055276e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4395
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.28125,
      "epoch": 0.2124081948202551,
      "grad_norm": 0.0,
      "learning_rate": 3.938200618477e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4396
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.75,
      "epoch": 0.2124565133359103,
      "grad_norm": 0.0,
      "learning_rate": 3.937959025898724e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4397
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.09375,
      "epoch": 0.21250483185156552,
      "grad_norm": 0.0,
      "learning_rate": 3.9377174333204484e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4398
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.84375,
      "epoch": 0.21255315036722072,
      "grad_norm": 0.0,
      "learning_rate": 3.937475840742172e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4399
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.4375,
      "epoch": 0.21260146888287593,
      "grad_norm": 0.0,
      "learning_rate": 3.9372342481638966e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4400
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.1875,
      "epoch": 0.2126497873985311,
      "grad_norm": 0.0,
      "learning_rate": 3.93699265558562e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4401
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.125,
      "epoch": 0.2126981059141863,
      "grad_norm": 0.0,
      "learning_rate": 3.9367510630073444e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4402
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.21875,
      "epoch": 0.2127464244298415,
      "grad_norm": 0.0,
      "learning_rate": 3.936509470429068e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4403
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 113.59375,
      "epoch": 0.21279474294549672,
      "grad_norm": 15.113315206559259,
      "learning_rate": 3.936267877850792e-07,
      "loss": 0.0,
      "reward": 1.125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.09375,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 4404
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.15625,
      "epoch": 0.21284306146115192,
      "grad_norm": 0.0,
      "learning_rate": 3.9360262852725163e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4405
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.4375,
      "epoch": 0.2128913799768071,
      "grad_norm": 0.0,
      "learning_rate": 3.93578469269424e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4406
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.0625,
      "epoch": 0.2129396984924623,
      "grad_norm": 0.0,
      "learning_rate": 3.9355431001159645e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4407
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 107.375,
      "epoch": 0.2129880170081175,
      "grad_norm": 0.0,
      "learning_rate": 3.935301507537688e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4408
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 100.8125,
      "epoch": 0.2130363355237727,
      "grad_norm": 0.0,
      "learning_rate": 3.9350599149594123e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4409
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.0625,
      "epoch": 0.21308465403942792,
      "grad_norm": 0.0,
      "learning_rate": 3.9348183223811364e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4410
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.28125,
      "epoch": 0.2131329725550831,
      "grad_norm": 0.0,
      "learning_rate": 3.9345767298028606e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4411
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 107.40625,
      "epoch": 0.2131812910707383,
      "grad_norm": 0.0,
      "learning_rate": 3.934335137224584e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4412
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.625,
      "epoch": 0.2132296095863935,
      "grad_norm": 0.0,
      "learning_rate": 3.9340935446463083e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4413
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 105.6875,
      "epoch": 0.2132779281020487,
      "grad_norm": 0.0,
      "learning_rate": 3.9338519520680324e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 4414
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 110.84375,
      "epoch": 0.2133262466177039,
      "grad_norm": 0.0,
      "learning_rate": 3.933610359489756e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4415
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.3125,
      "epoch": 0.2133745651333591,
      "grad_norm": 0.0,
      "learning_rate": 3.93336876691148e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4416
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.59375,
      "epoch": 0.2134228836490143,
      "grad_norm": 0.0,
      "learning_rate": 3.9331271743332043e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4417
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.8125,
      "epoch": 0.2134712021646695,
      "grad_norm": 0.0,
      "learning_rate": 3.9328855817549285e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4418
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.625,
      "epoch": 0.2135195206803247,
      "grad_norm": 0.0,
      "learning_rate": 3.932643989176652e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4419
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.71875,
      "epoch": 0.2135678391959799,
      "grad_norm": 0.0,
      "learning_rate": 3.9324023965983767e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4420
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.84375,
      "epoch": 0.21361615771163509,
      "grad_norm": 0.0,
      "learning_rate": 3.9321608040201003e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4421
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.375,
      "epoch": 0.2136644762272903,
      "grad_norm": 0.0,
      "learning_rate": 3.9319192114418245e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4422
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.8125,
      "epoch": 0.2137127947429455,
      "grad_norm": 0.0,
      "learning_rate": 3.931677618863548e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4423
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.71875,
      "epoch": 0.2137611132586007,
      "grad_norm": 0.0,
      "learning_rate": 3.931436026285273e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4424
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.1875,
      "epoch": 0.2138094317742559,
      "grad_norm": 0.0,
      "learning_rate": 3.9311944337069964e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4425
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.46875,
      "epoch": 0.21385775028991108,
      "grad_norm": 0.0,
      "learning_rate": 3.93095284112872e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4426
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.25,
      "epoch": 0.21390606880556629,
      "grad_norm": 0.0,
      "learning_rate": 3.9307112485504446e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4427
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.90625,
      "epoch": 0.2139543873212215,
      "grad_norm": 0.0,
      "learning_rate": 3.930469655972168e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4428
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.03125,
      "epoch": 0.2140027058368767,
      "grad_norm": 0.0,
      "learning_rate": 3.9302280633938924e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4429
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.59375,
      "epoch": 0.2140510243525319,
      "grad_norm": 0.0,
      "learning_rate": 3.9299864708156165e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4430
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.28125,
      "epoch": 0.21409934286818708,
      "grad_norm": 0.0,
      "learning_rate": 3.9297448782373406e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4431
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.9375,
      "epoch": 0.21414766138384228,
      "grad_norm": 0.0,
      "learning_rate": 3.929503285659064e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4432
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.5625,
      "epoch": 0.21419597989949749,
      "grad_norm": 0.0,
      "learning_rate": 3.9292616930807884e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880052089691,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4433
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 111.25,
      "epoch": 0.2142442984151527,
      "grad_norm": 0.0,
      "learning_rate": 3.9290201005025125e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4434
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.09375,
      "epoch": 0.2142926169308079,
      "grad_norm": 0.0,
      "learning_rate": 3.928778507924236e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4435
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.125,
      "epoch": 0.21434093544646307,
      "grad_norm": 0.0,
      "learning_rate": 3.92853691534596e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4436
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.0,
      "epoch": 0.21438925396211828,
      "grad_norm": 0.0,
      "learning_rate": 3.9282953227676844e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4437
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.34375,
      "epoch": 0.21443757247777348,
      "grad_norm": 0.0,
      "learning_rate": 3.9280537301894085e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4438
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.0625,
      "epoch": 0.21448589099342869,
      "grad_norm": 0.0,
      "learning_rate": 3.927812137611132e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4439
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.8125,
      "epoch": 0.2145342095090839,
      "grad_norm": 0.0,
      "learning_rate": 3.927570545032857e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4440
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.625,
      "epoch": 0.21458252802473907,
      "grad_norm": 0.0,
      "learning_rate": 3.9273289524545804e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4441
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.46875,
      "epoch": 0.21463084654039427,
      "grad_norm": 0.0,
      "learning_rate": 3.9270873598763046e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4442
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.03125,
      "epoch": 0.21467916505604948,
      "grad_norm": 0.0,
      "learning_rate": 3.926845767298028e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4443
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.53125,
      "epoch": 0.21472748357170468,
      "grad_norm": 0.0,
      "learning_rate": 3.926604174719753e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4444
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 106.84375,
      "epoch": 0.21477580208735988,
      "grad_norm": 15.391262063215883,
      "learning_rate": 3.9263625821414764e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.22017991542816162,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4445
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.6875,
      "epoch": 0.21482412060301506,
      "grad_norm": 0.0,
      "learning_rate": 3.9261209895632e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4446
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.1875,
      "epoch": 0.21487243911867027,
      "grad_norm": 0.0,
      "learning_rate": 3.9258793969849247e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4447
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.5625,
      "epoch": 0.21492075763432547,
      "grad_norm": 0.0,
      "learning_rate": 3.9256378044066483e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4448
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.15625,
      "epoch": 0.21496907614998068,
      "grad_norm": 0.0,
      "learning_rate": 3.9253962118283724e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 4449
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.21875,
      "epoch": 0.21501739466563588,
      "grad_norm": 0.0,
      "learning_rate": 3.9251546192500966e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4450
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 114.90625,
      "epoch": 0.21506571318129106,
      "grad_norm": 0.0,
      "learning_rate": 3.9249130266718207e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4451
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.09375,
      "epoch": 0.21511403169694626,
      "grad_norm": 0.0,
      "learning_rate": 3.9246714340935443e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4452
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.40625,
      "epoch": 0.21516235021260147,
      "grad_norm": 0.0,
      "learning_rate": 3.9244298415152685e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4453
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 106.9375,
      "epoch": 0.21521066872825667,
      "grad_norm": 10.750765172874043,
      "learning_rate": 3.9241882489369926e-07,
      "loss": -0.0,
      "reward": 2.359375,
      "reward_std": 0.2579136788845062,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4454
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.0,
      "epoch": 0.21525898724391188,
      "grad_norm": 0.0,
      "learning_rate": 3.923946656358716e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4455
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.625,
      "epoch": 0.21530730575956705,
      "grad_norm": 0.0,
      "learning_rate": 3.9237050637804403e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4456
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 112.3125,
      "epoch": 0.21535562427522226,
      "grad_norm": 0.0,
      "learning_rate": 3.9234634712021645e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4457
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.46875,
      "epoch": 0.21540394279087746,
      "grad_norm": 0.0,
      "learning_rate": 3.9232218786238886e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4458
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 116.65625,
      "epoch": 0.21545226130653267,
      "grad_norm": 0.0,
      "learning_rate": 3.922980286045612e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 4459
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.15625,
      "epoch": 0.21550057982218787,
      "grad_norm": 0.0,
      "learning_rate": 3.922738693467337e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4460
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.625,
      "epoch": 0.21554889833784305,
      "grad_norm": 0.0,
      "learning_rate": 3.9224971008890605e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4461
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.875,
      "epoch": 0.21559721685349825,
      "grad_norm": 0.0,
      "learning_rate": 3.9222555083107846e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4462
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 119.25,
      "epoch": 0.21564553536915346,
      "grad_norm": 0.0,
      "learning_rate": 3.922013915732508e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 4463
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.40625,
      "epoch": 0.21569385388480866,
      "grad_norm": 0.0,
      "learning_rate": 3.921772323154233e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4464
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 121.65625,
      "epoch": 0.21574217240046387,
      "grad_norm": 0.0,
      "learning_rate": 3.9215307305759565e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 4465
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.46875,
      "epoch": 0.21579049091611904,
      "grad_norm": 0.0,
      "learning_rate": 3.92128913799768e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4466
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.4375,
      "epoch": 0.21583880943177425,
      "grad_norm": 0.0,
      "learning_rate": 3.921047545419405e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4467
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.9375,
      "epoch": 0.21588712794742945,
      "grad_norm": 0.0,
      "learning_rate": 3.9208059528411284e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4468
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 112.625,
      "epoch": 0.21593544646308466,
      "grad_norm": 0.0,
      "learning_rate": 3.9205643602628525e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 4469
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.65625,
      "epoch": 0.21598376497873986,
      "grad_norm": 0.0,
      "learning_rate": 3.9203227676845767e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4470
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.71875,
      "epoch": 0.21603208349439504,
      "grad_norm": 0.0,
      "learning_rate": 3.920081175106301e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4471
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.3125,
      "epoch": 0.21608040201005024,
      "grad_norm": 0.0,
      "learning_rate": 3.9198395825280244e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4472
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.34375,
      "epoch": 0.21612872052570545,
      "grad_norm": 0.0,
      "learning_rate": 3.9195979899497485e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4473
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 124.65625,
      "epoch": 0.21617703904136065,
      "grad_norm": 0.0,
      "learning_rate": 3.9193563973714727e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 4474
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.125,
      "epoch": 0.21622535755701586,
      "grad_norm": 0.0,
      "learning_rate": 3.919114804793197e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4475
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.75,
      "epoch": 0.21627367607267103,
      "grad_norm": 0.0,
      "learning_rate": 3.9188732122149204e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4476
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.625,
      "epoch": 0.21632199458832624,
      "grad_norm": 0.0,
      "learning_rate": 3.9186316196366446e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4477
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 114.8125,
      "epoch": 0.21637031310398144,
      "grad_norm": 0.0,
      "learning_rate": 3.9183900270583687e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4478
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.28125,
      "epoch": 0.21641863161963665,
      "grad_norm": 0.0,
      "learning_rate": 3.9181484344800923e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4479
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 104.3125,
      "epoch": 0.21646695013529185,
      "grad_norm": 32.46726069466582,
      "learning_rate": 3.917906841901817e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4480
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.65625,
      "epoch": 0.21651526865094703,
      "grad_norm": 0.0,
      "learning_rate": 3.9176652493235406e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4481
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 117.53125,
      "epoch": 0.21656358716660223,
      "grad_norm": 7.118795729578217,
      "learning_rate": 3.9174236567452647e-07,
      "loss": -0.0,
      "reward": 1.578125,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4482
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.875,
      "epoch": 0.21661190568225744,
      "grad_norm": 0.0,
      "learning_rate": 3.9171820641669883e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4483
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 115.71875,
      "epoch": 0.21666022419791264,
      "grad_norm": 10.610036325153363,
      "learning_rate": 3.916940471588713e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4484
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.9375,
      "epoch": 0.21670854271356785,
      "grad_norm": 0.0,
      "learning_rate": 3.9166988790104366e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4485
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.0625,
      "epoch": 0.21675686122922302,
      "grad_norm": 0.0,
      "learning_rate": 3.91645728643216e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4486
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.8125,
      "epoch": 0.21680517974487823,
      "grad_norm": 0.0,
      "learning_rate": 3.916215693853885e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4487
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.4375,
      "epoch": 0.21685349826053343,
      "grad_norm": 6.120216331409301,
      "learning_rate": 3.9159741012756085e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4488
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.0625,
      "epoch": 0.21690181677618864,
      "grad_norm": 0.0,
      "learning_rate": 3.9157325086973326e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4489
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.59375,
      "epoch": 0.21695013529184384,
      "grad_norm": 0.0,
      "learning_rate": 3.915490916119057e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4490
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.53125,
      "epoch": 0.21699845380749902,
      "grad_norm": 0.0,
      "learning_rate": 3.915249323540781e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4491
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.9375,
      "epoch": 0.21704677232315422,
      "grad_norm": 0.0,
      "learning_rate": 3.9150077309625045e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4492
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.34375,
      "epoch": 0.21709509083880943,
      "grad_norm": 0.0,
      "learning_rate": 3.9147661383842286e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4493
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.75,
      "epoch": 0.21714340935446463,
      "grad_norm": 0.0,
      "learning_rate": 3.914524545805953e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4494
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.625,
      "epoch": 0.21719172787011984,
      "grad_norm": 0.0,
      "learning_rate": 3.914282953227677e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4495
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 102.125,
      "epoch": 0.21724004638577502,
      "grad_norm": 5.894236470937479,
      "learning_rate": 3.9140413606494005e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4496
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.84375,
      "epoch": 0.21728836490143022,
      "grad_norm": 0.0,
      "learning_rate": 3.9137997680711246e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4497
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.25,
      "epoch": 0.21733668341708542,
      "grad_norm": 0.0,
      "learning_rate": 3.913558175492849e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4498
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.59375,
      "epoch": 0.21738500193274063,
      "grad_norm": 0.0,
      "learning_rate": 3.9133165829145724e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4499
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 97.5625,
      "epoch": 0.21743332044839583,
      "grad_norm": 11.348234949851314,
      "learning_rate": 3.913074990336297e-07,
      "loss": -0.0,
      "reward": 2.375,
      "reward_std": 0.24359199404716492,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4500
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 101.0,
      "epoch": 0.217481638964051,
      "grad_norm": 11.632712215174779,
      "learning_rate": 3.9128333977580206e-07,
      "loss": -0.0,
      "reward": 2.140625,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4501
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.875,
      "epoch": 0.21752995747970622,
      "grad_norm": 0.0,
      "learning_rate": 3.912591805179745e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4502
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.0,
      "epoch": 0.21757827599536142,
      "grad_norm": 0.0,
      "learning_rate": 3.9123502126014684e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4503
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 100.375,
      "epoch": 0.21762659451101662,
      "grad_norm": 11.311571599795318,
      "learning_rate": 3.912108620023193e-07,
      "loss": -0.0,
      "reward": 1.078125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.0625,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 4504
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.59375,
      "epoch": 0.21767491302667183,
      "grad_norm": 0.0,
      "learning_rate": 3.9118670274449167e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4505
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.09375,
      "epoch": 0.217723231542327,
      "grad_norm": 0.0,
      "learning_rate": 3.911625434866641e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 4506
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.875,
      "epoch": 0.2177715500579822,
      "grad_norm": 0.0,
      "learning_rate": 3.911383842288365e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4507
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.0625,
      "epoch": 0.21781986857363742,
      "grad_norm": 0.0,
      "learning_rate": 3.9111422497100885e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4508
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.96875,
      "epoch": 0.21786818708929262,
      "grad_norm": 0.0,
      "learning_rate": 3.9109006571318127e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4509
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.28125,
      "epoch": 0.21791650560494782,
      "grad_norm": 0.0,
      "learning_rate": 3.910659064553537e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4510
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.75,
      "epoch": 0.217964824120603,
      "grad_norm": 0.0,
      "learning_rate": 3.910417471975261e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4511
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.03125,
      "epoch": 0.2180131426362582,
      "grad_norm": 0.0,
      "learning_rate": 3.9101758793969846e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4512
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.375,
      "epoch": 0.2180614611519134,
      "grad_norm": 0.0,
      "learning_rate": 3.9099342868187087e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4513
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 102.15625,
      "epoch": 0.21810977966756862,
      "grad_norm": 0.0,
      "learning_rate": 3.909692694240433e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 4514
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.96875,
      "epoch": 0.21815809818322382,
      "grad_norm": 0.0,
      "learning_rate": 3.909451101662157e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4515
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.3125,
      "epoch": 0.218206416698879,
      "grad_norm": 0.0,
      "learning_rate": 3.9092095090838806e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4516
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.21875,
      "epoch": 0.2182547352145342,
      "grad_norm": 0.0,
      "learning_rate": 3.908967916505605e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4517
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.71875,
      "epoch": 0.2183030537301894,
      "grad_norm": 0.0,
      "learning_rate": 3.908726323927329e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4518
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.0625,
      "epoch": 0.2183513722458446,
      "grad_norm": 0.0,
      "learning_rate": 3.9084847313490525e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4519
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 96.4375,
      "epoch": 0.21839969076149982,
      "grad_norm": 9.437936557411636,
      "learning_rate": 3.908243138770777e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.2608962655067444,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4520
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.90625,
      "epoch": 0.218448009277155,
      "grad_norm": 0.0,
      "learning_rate": 3.9080015461925007e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4521
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 101.8125,
      "epoch": 0.2184963277928102,
      "grad_norm": 0.0,
      "learning_rate": 3.907759953614225e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4522
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.4375,
      "epoch": 0.2185446463084654,
      "grad_norm": 0.0,
      "learning_rate": 3.9075183610359485e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4523
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.0,
      "epoch": 0.2185929648241206,
      "grad_norm": 0.0,
      "learning_rate": 3.907276768457673e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4524
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 96.8125,
      "epoch": 0.2186412833397758,
      "grad_norm": 10.543394570604423,
      "learning_rate": 3.907035175879397e-07,
      "loss": 0.0,
      "reward": 1.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.03125,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 4525
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.65625,
      "epoch": 0.218689601855431,
      "grad_norm": 0.0,
      "learning_rate": 3.906793583301121e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4526
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.375,
      "epoch": 0.2187379203710862,
      "grad_norm": 0.0,
      "learning_rate": 3.906551990722845e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4527
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.75,
      "epoch": 0.2187862388867414,
      "grad_norm": 58.621508403882984,
      "learning_rate": 3.9063103981445686e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.3074946999549866,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4528
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 100.90625,
      "epoch": 0.2188345574023966,
      "grad_norm": 0.0,
      "learning_rate": 3.906068805566293e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4529
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.3125,
      "epoch": 0.2188828759180518,
      "grad_norm": 0.0,
      "learning_rate": 3.905827212988017e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4530
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.5625,
      "epoch": 0.218931194433707,
      "grad_norm": 0.0,
      "learning_rate": 3.905585620409741e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4531
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.625,
      "epoch": 0.2189795129493622,
      "grad_norm": 0.0,
      "learning_rate": 3.9053440278314646e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4532
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.0625,
      "epoch": 0.2190278314650174,
      "grad_norm": 0.0,
      "learning_rate": 3.905102435253189e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4533
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.09375,
      "epoch": 0.2190761499806726,
      "grad_norm": 0.0,
      "learning_rate": 3.904860842674913e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4534
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.6875,
      "epoch": 0.2191244684963278,
      "grad_norm": 0.0,
      "learning_rate": 3.904619250096637e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4535
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.09375,
      "epoch": 0.219172787011983,
      "grad_norm": 0.0,
      "learning_rate": 3.9043776575183607e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4536
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 89.4375,
      "epoch": 0.21922110552763818,
      "grad_norm": 6.115455830758629,
      "learning_rate": 3.9041360649400853e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4537
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.90625,
      "epoch": 0.2192694240432934,
      "grad_norm": 0.0,
      "learning_rate": 3.903894472361809e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4538
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.21875,
      "epoch": 0.2193177425589486,
      "grad_norm": 0.0,
      "learning_rate": 3.9036528797835325e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4539
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.03125,
      "epoch": 0.2193660610746038,
      "grad_norm": 0.0,
      "learning_rate": 3.903411287205257e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4540
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.375,
      "epoch": 0.219414379590259,
      "grad_norm": 0.0,
      "learning_rate": 3.903169694626981e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4541
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.9375,
      "epoch": 0.21946269810591418,
      "grad_norm": 0.0,
      "learning_rate": 3.902928102048705e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4542
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.90625,
      "epoch": 0.21951101662156938,
      "grad_norm": 0.0,
      "learning_rate": 3.9026865094704286e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4543
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.34375,
      "epoch": 0.2195593351372246,
      "grad_norm": 0.0,
      "learning_rate": 3.902444916892153e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4544
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.9375,
      "epoch": 0.2196076536528798,
      "grad_norm": 0.0,
      "learning_rate": 3.902203324313877e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4545
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.875,
      "epoch": 0.219655972168535,
      "grad_norm": 0.0,
      "learning_rate": 3.901961731735601e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4546
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.96875,
      "epoch": 0.21970429068419017,
      "grad_norm": 0.0,
      "learning_rate": 3.901720139157325e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4547
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.28125,
      "epoch": 0.21975260919984538,
      "grad_norm": 0.0,
      "learning_rate": 3.9014785465790487e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4548
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.5,
      "epoch": 0.21980092771550058,
      "grad_norm": 0.0,
      "learning_rate": 3.901236954000773e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4549
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.4375,
      "epoch": 0.2198492462311558,
      "grad_norm": 0.0,
      "learning_rate": 3.900995361422497e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4550
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.59375,
      "epoch": 0.219897564746811,
      "grad_norm": 0.0,
      "learning_rate": 3.900753768844221e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4551
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.96875,
      "epoch": 0.21994588326246617,
      "grad_norm": 0.0,
      "learning_rate": 3.9005121762659447e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4552
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.34375,
      "epoch": 0.21999420177812137,
      "grad_norm": 0.0,
      "learning_rate": 3.900270583687669e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4553
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.15625,
      "epoch": 0.22004252029377658,
      "grad_norm": 0.0,
      "learning_rate": 3.900028991109393e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4554
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.21875,
      "epoch": 0.22009083880943178,
      "grad_norm": 0.0,
      "learning_rate": 3.899787398531117e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4555
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.71875,
      "epoch": 0.220139157325087,
      "grad_norm": 0.0,
      "learning_rate": 3.8995458059528407e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4556
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.4375,
      "epoch": 0.22018747584074216,
      "grad_norm": 0.0,
      "learning_rate": 3.8993042133745654e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4557
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.625,
      "epoch": 0.22023579435639737,
      "grad_norm": 0.0,
      "learning_rate": 3.899062620796289e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4558
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.6875,
      "epoch": 0.22028411287205257,
      "grad_norm": 0.0,
      "learning_rate": 3.8988210282180126e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4559
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.90625,
      "epoch": 0.22033243138770778,
      "grad_norm": 0.0,
      "learning_rate": 3.8985794356397373e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4560
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.34375,
      "epoch": 0.22038074990336298,
      "grad_norm": 0.0,
      "learning_rate": 3.898337843061461e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4561
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.3125,
      "epoch": 0.22042906841901816,
      "grad_norm": 14.068018998140081,
      "learning_rate": 3.898096250483185e-07,
      "loss": -0.0,
      "reward": 2.28125,
      "reward_std": 0.3188670873641968,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4562
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 85.1875,
      "epoch": 0.22047738693467336,
      "grad_norm": 30.37797425811645,
      "learning_rate": 3.8978546579049086e-07,
      "loss": -0.0,
      "reward": 1.734375,
      "reward_std": 0.2731459438800812,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4563
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.1875,
      "epoch": 0.22052570545032857,
      "grad_norm": 0.0,
      "learning_rate": 3.8976130653266333e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4564
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.15625,
      "epoch": 0.22057402396598377,
      "grad_norm": 0.0,
      "learning_rate": 3.897371472748357e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4565
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 98.9375,
      "epoch": 0.22062234248163898,
      "grad_norm": 0.0,
      "learning_rate": 3.897129880170081e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 4566
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.1875,
      "epoch": 0.22067066099729415,
      "grad_norm": 0.0,
      "learning_rate": 3.896888287591805e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4567
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.96875,
      "epoch": 0.22071897951294936,
      "grad_norm": 0.0,
      "learning_rate": 3.8966466950135293e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4568
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.34375,
      "epoch": 0.22076729802860456,
      "grad_norm": 0.0,
      "learning_rate": 3.896405102435253e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4569
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.5,
      "epoch": 0.22081561654425977,
      "grad_norm": 0.0,
      "learning_rate": 3.896163509856977e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4570
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.1875,
      "epoch": 0.22086393505991497,
      "grad_norm": 0.0,
      "learning_rate": 3.895921917278701e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4571
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.0625,
      "epoch": 0.22091225357557015,
      "grad_norm": 0.0,
      "learning_rate": 3.895680324700425e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4572
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 106.1875,
      "epoch": 0.22096057209122535,
      "grad_norm": 0.0,
      "learning_rate": 3.895438732122149e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4573
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.625,
      "epoch": 0.22100889060688056,
      "grad_norm": 0.0,
      "learning_rate": 3.895197139543873e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4574
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.5625,
      "epoch": 0.22105720912253576,
      "grad_norm": 0.0,
      "learning_rate": 3.894955546965597e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4575
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.9375,
      "epoch": 0.22110552763819097,
      "grad_norm": 0.0,
      "learning_rate": 3.894713954387321e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4576
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.46875,
      "epoch": 0.22115384615384615,
      "grad_norm": 0.0,
      "learning_rate": 3.8944723618090455e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4577
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.65625,
      "epoch": 0.22120216466950135,
      "grad_norm": 0.0,
      "learning_rate": 3.894230769230769e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4578
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.75,
      "epoch": 0.22125048318515655,
      "grad_norm": 0.0,
      "learning_rate": 3.8939891766524927e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4579
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.15625,
      "epoch": 0.22129880170081176,
      "grad_norm": 0.0,
      "learning_rate": 3.8937475840742174e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4580
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.96875,
      "epoch": 0.22134712021646696,
      "grad_norm": 0.0,
      "learning_rate": 3.893505991495941e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4581
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.15625,
      "epoch": 0.22139543873212214,
      "grad_norm": 18.041093073959622,
      "learning_rate": 3.893264398917665e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.32878512144088745,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4582
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.59375,
      "epoch": 0.22144375724777735,
      "grad_norm": 0.0,
      "learning_rate": 3.8930228063393887e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4583
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.40625,
      "epoch": 0.22149207576343255,
      "grad_norm": 0.0,
      "learning_rate": 3.8927812137611134e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4584
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 101.75,
      "epoch": 0.22154039427908775,
      "grad_norm": 33.05390016545061,
      "learning_rate": 3.892539621182837e-07,
      "loss": -0.0,
      "reward": 1.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.03125,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 4585
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.84375,
      "epoch": 0.22158871279474296,
      "grad_norm": 0.0,
      "learning_rate": 3.892298028604561e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4586
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.28125,
      "epoch": 0.22163703131039814,
      "grad_norm": 0.0,
      "learning_rate": 3.892056436026285e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4587
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.09375,
      "epoch": 0.22168534982605334,
      "grad_norm": 0.0,
      "learning_rate": 3.8918148434480094e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4588
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.3125,
      "epoch": 0.22173366834170855,
      "grad_norm": 0.0,
      "learning_rate": 3.891573250869733e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4589
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.84375,
      "epoch": 0.22178198685736375,
      "grad_norm": 0.0,
      "learning_rate": 3.891331658291457e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4590
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.75,
      "epoch": 0.22183030537301895,
      "grad_norm": 0.0,
      "learning_rate": 3.8910900657131813e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4591
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.3125,
      "epoch": 0.22187862388867413,
      "grad_norm": 9.605268590109826,
      "learning_rate": 3.890848473134905e-07,
      "loss": -0.0,
      "reward": 2.328125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4592
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.03125,
      "epoch": 0.22192694240432934,
      "grad_norm": 0.0,
      "learning_rate": 3.890606880556629e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4593
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.65625,
      "epoch": 0.22197526091998454,
      "grad_norm": 0.0,
      "learning_rate": 3.890365287978353e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4594
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.15625,
      "epoch": 0.22202357943563975,
      "grad_norm": 0.0,
      "learning_rate": 3.8901236954000773e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4595
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.71875,
      "epoch": 0.22207189795129495,
      "grad_norm": 0.0,
      "learning_rate": 3.889882102821801e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 4596
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.84375,
      "epoch": 0.22212021646695013,
      "grad_norm": 0.0,
      "learning_rate": 3.8896405102435256e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4597
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.84375,
      "epoch": 0.22216853498260533,
      "grad_norm": 0.0,
      "learning_rate": 3.889398917665249e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4598
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 98.1875,
      "epoch": 0.22221685349826054,
      "grad_norm": 0.0,
      "learning_rate": 3.889157325086973e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 4599
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.9375,
      "epoch": 0.22226517201391574,
      "grad_norm": 0.0,
      "learning_rate": 3.8889157325086974e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4600
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.625,
      "epoch": 0.22231349052957095,
      "grad_norm": 0.0,
      "learning_rate": 3.888674139930421e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4601
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.21875,
      "epoch": 0.22236180904522612,
      "grad_norm": 0.0,
      "learning_rate": 3.888432547352145e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4602
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.65625,
      "epoch": 0.22241012756088133,
      "grad_norm": 0.0,
      "learning_rate": 3.888190954773869e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4603
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.1875,
      "epoch": 0.22245844607653653,
      "grad_norm": 0.0,
      "learning_rate": 3.8879493621955935e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4604
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.34375,
      "epoch": 0.22250676459219174,
      "grad_norm": 0.0,
      "learning_rate": 3.887707769617317e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4605
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.6875,
      "epoch": 0.22255508310784694,
      "grad_norm": 0.0,
      "learning_rate": 3.887466177039041e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4606
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.28125,
      "epoch": 0.22260340162350212,
      "grad_norm": 0.0,
      "learning_rate": 3.8872245844607653e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4607
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.3125,
      "epoch": 0.22265172013915732,
      "grad_norm": 0.0,
      "learning_rate": 3.8869829918824895e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4608
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.15625,
      "epoch": 0.22270003865481253,
      "grad_norm": 0.0,
      "learning_rate": 3.886741399304213e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4609
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.78125,
      "epoch": 0.22274835717046773,
      "grad_norm": 0.0,
      "learning_rate": 3.886499806725937e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4610
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.5,
      "epoch": 0.22279667568612294,
      "grad_norm": 0.0,
      "learning_rate": 3.8862582141476613e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4611
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.625,
      "epoch": 0.2228449942017781,
      "grad_norm": 0.0,
      "learning_rate": 3.886016621569385e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4612
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.5625,
      "epoch": 0.22289331271743332,
      "grad_norm": 0.0,
      "learning_rate": 3.885775028991109e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4613
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.5625,
      "epoch": 0.22294163123308852,
      "grad_norm": 0.0,
      "learning_rate": 3.885533436412833e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4614
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.90625,
      "epoch": 0.22298994974874373,
      "grad_norm": 0.0,
      "learning_rate": 3.8852918438345574e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4615
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.9375,
      "epoch": 0.22303826826439893,
      "grad_norm": 0.0,
      "learning_rate": 3.885050251256281e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4616
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.625,
      "epoch": 0.2230865867800541,
      "grad_norm": 0.0,
      "learning_rate": 3.8848086586780056e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4617
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.4375,
      "epoch": 0.2231349052957093,
      "grad_norm": 0.0,
      "learning_rate": 3.884567066099729e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4618
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.3125,
      "epoch": 0.22318322381136452,
      "grad_norm": 0.0,
      "learning_rate": 3.8843254735214534e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4619
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.90625,
      "epoch": 0.22323154232701972,
      "grad_norm": 0.0,
      "learning_rate": 3.8840838809431775e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4620
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.21875,
      "epoch": 0.22327986084267493,
      "grad_norm": 0.0,
      "learning_rate": 3.883842288364901e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4621
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 98.3125,
      "epoch": 0.2233281793583301,
      "grad_norm": 9.801805249355985,
      "learning_rate": 3.883600695786625e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4622
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.9375,
      "epoch": 0.2233764978739853,
      "grad_norm": 0.0,
      "learning_rate": 3.883359103208349e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4623
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.5625,
      "epoch": 0.2234248163896405,
      "grad_norm": 0.0,
      "learning_rate": 3.8831175106300735e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4624
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.625,
      "epoch": 0.22347313490529572,
      "grad_norm": 0.0,
      "learning_rate": 3.882875918051797e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4625
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.75,
      "epoch": 0.22352145342095092,
      "grad_norm": 0.0,
      "learning_rate": 3.8826343254735213e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4626
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.625,
      "epoch": 0.2235697719366061,
      "grad_norm": 0.0,
      "learning_rate": 3.8823927328952454e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4627
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 100.5625,
      "epoch": 0.2236180904522613,
      "grad_norm": 17.842245010697845,
      "learning_rate": 3.8821511403169695e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4628
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 93.40625,
      "epoch": 0.2236664089679165,
      "grad_norm": 5.190605863494789,
      "learning_rate": 3.881909547738693e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.32195520401000977,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4629
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.90625,
      "epoch": 0.2237147274835717,
      "grad_norm": 0.0,
      "learning_rate": 3.881667955160418e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4630
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.8125,
      "epoch": 0.22376304599922692,
      "grad_norm": 0.0,
      "learning_rate": 3.8814263625821414e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4631
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.6875,
      "epoch": 0.2238113645148821,
      "grad_norm": 0.0,
      "learning_rate": 3.881184770003865e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 4632
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.09375,
      "epoch": 0.2238596830305373,
      "grad_norm": 0.0,
      "learning_rate": 3.880943177425589e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4633
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.875,
      "epoch": 0.2239080015461925,
      "grad_norm": 0.0,
      "learning_rate": 3.8807015848473133e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 4634
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.3125,
      "epoch": 0.2239563200618477,
      "grad_norm": 0.0,
      "learning_rate": 3.8804599922690374e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4635
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.4375,
      "epoch": 0.2240046385775029,
      "grad_norm": 0.0,
      "learning_rate": 3.880218399690761e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4636
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.125,
      "epoch": 0.2240529570931581,
      "grad_norm": 0.0,
      "learning_rate": 3.8799768071124857e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4637
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.6875,
      "epoch": 0.2241012756088133,
      "grad_norm": 0.0,
      "learning_rate": 3.8797352145342093e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4638
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.375,
      "epoch": 0.2241495941244685,
      "grad_norm": 0.0,
      "learning_rate": 3.8794936219559335e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4639
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 99.78125,
      "epoch": 0.2241979126401237,
      "grad_norm": 31.550054988724067,
      "learning_rate": 3.8792520293776576e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.21528217196464539,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4640
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.78125,
      "epoch": 0.2242462311557789,
      "grad_norm": 0.0,
      "learning_rate": 3.879010436799381e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4641
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.625,
      "epoch": 0.22429454967143408,
      "grad_norm": 0.0,
      "learning_rate": 3.8787688442211053e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4642
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 102.03125,
      "epoch": 0.2243428681870893,
      "grad_norm": 0.0,
      "learning_rate": 3.878527251642829e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4643
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 107.5625,
      "epoch": 0.2243911867027445,
      "grad_norm": 9.75076374307553,
      "learning_rate": 3.8782856590645536e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4644
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.9375,
      "epoch": 0.2244395052183997,
      "grad_norm": 0.0,
      "learning_rate": 3.878044066486277e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4645
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.65625,
      "epoch": 0.2244878237340549,
      "grad_norm": 0.0,
      "learning_rate": 3.8778024739080014e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4646
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.0625,
      "epoch": 0.22453614224971008,
      "grad_norm": 0.0,
      "learning_rate": 3.8775608813297255e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4647
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.59375,
      "epoch": 0.22458446076536528,
      "grad_norm": 0.0,
      "learning_rate": 3.8773192887514496e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4648
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 98.96875,
      "epoch": 0.2246327792810205,
      "grad_norm": 20.2280822479928,
      "learning_rate": 3.877077696173173e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.3093337416648865,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4649
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.6875,
      "epoch": 0.2246810977966757,
      "grad_norm": 0.0,
      "learning_rate": 3.876836103594898e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4650
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.6875,
      "epoch": 0.2247294163123309,
      "grad_norm": 0.0,
      "learning_rate": 3.8765945110166215e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4651
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.5625,
      "epoch": 0.22477773482798608,
      "grad_norm": 0.0,
      "learning_rate": 3.876352918438345e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4652
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.1875,
      "epoch": 0.22482605334364128,
      "grad_norm": 0.0,
      "learning_rate": 3.876111325860069e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4653
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.90625,
      "epoch": 0.22487437185929648,
      "grad_norm": 0.0,
      "learning_rate": 3.8758697332817934e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4654
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.125,
      "epoch": 0.2249226903749517,
      "grad_norm": 0.0,
      "learning_rate": 3.8756281407035175e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4655
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 98.1875,
      "epoch": 0.2249710088906069,
      "grad_norm": 0.0,
      "learning_rate": 3.875386548125241e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 4656
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.90625,
      "epoch": 0.22501932740626207,
      "grad_norm": 0.0,
      "learning_rate": 3.875144955546966e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4657
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.0625,
      "epoch": 0.22506764592191728,
      "grad_norm": 0.0,
      "learning_rate": 3.8749033629686894e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4658
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.1875,
      "epoch": 0.22511596443757248,
      "grad_norm": 0.0,
      "learning_rate": 3.8746617703904135e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4659
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.34375,
      "epoch": 0.22516428295322768,
      "grad_norm": 0.0,
      "learning_rate": 3.8744201778121377e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4660
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.15625,
      "epoch": 0.2252126014688829,
      "grad_norm": 0.0,
      "learning_rate": 3.8741785852338613e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4661
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 98.46875,
      "epoch": 0.22526091998453807,
      "grad_norm": 0.0,
      "learning_rate": 3.8739369926555854e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 4662
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.0,
      "epoch": 0.22530923850019327,
      "grad_norm": 0.0,
      "learning_rate": 3.873695400077309e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4663
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.40625,
      "epoch": 0.22535755701584848,
      "grad_norm": 0.0,
      "learning_rate": 3.8734538074990337e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4664
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 100.0,
      "epoch": 0.22540587553150368,
      "grad_norm": 9.423722428083833,
      "learning_rate": 3.8732122149207573e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4665
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.03125,
      "epoch": 0.22545419404715888,
      "grad_norm": 0.0,
      "learning_rate": 3.8729706223424814e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4666
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.71875,
      "epoch": 0.22550251256281406,
      "grad_norm": 0.0,
      "learning_rate": 3.8727290297642056e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4667
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.9375,
      "epoch": 0.22555083107846927,
      "grad_norm": 0.0,
      "learning_rate": 3.8724874371859297e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4668
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.75,
      "epoch": 0.22559914959412447,
      "grad_norm": 0.0,
      "learning_rate": 3.8722458446076533e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4669
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.125,
      "epoch": 0.22564746810977968,
      "grad_norm": 0.0,
      "learning_rate": 3.872004252029378e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4670
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.6875,
      "epoch": 0.22569578662543488,
      "grad_norm": 0.0,
      "learning_rate": 3.8717626594511016e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4671
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.21875,
      "epoch": 0.22574410514109006,
      "grad_norm": 0.0,
      "learning_rate": 3.871521066872825e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4672
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.40625,
      "epoch": 0.22579242365674526,
      "grad_norm": 0.0,
      "learning_rate": 3.8712794742945493e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4673
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.25,
      "epoch": 0.22584074217240047,
      "grad_norm": 0.0,
      "learning_rate": 3.8710378817162735e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4674
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.1875,
      "epoch": 0.22588906068805567,
      "grad_norm": 0.0,
      "learning_rate": 3.8707962891379976e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4675
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.6875,
      "epoch": 0.22593737920371088,
      "grad_norm": 9.996695639954002,
      "learning_rate": 3.870554696559721e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4676
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.53125,
      "epoch": 0.22598569771936605,
      "grad_norm": 0.0,
      "learning_rate": 3.870313103981446e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4677
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.40625,
      "epoch": 0.22603401623502126,
      "grad_norm": 0.0,
      "learning_rate": 3.8700715114031695e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4678
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.21875,
      "epoch": 0.22608233475067646,
      "grad_norm": 0.0,
      "learning_rate": 3.8698299188248936e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4679
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.875,
      "epoch": 0.22613065326633167,
      "grad_norm": 0.0,
      "learning_rate": 3.869588326246618e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4680
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.40625,
      "epoch": 0.22617897178198687,
      "grad_norm": 0.0,
      "learning_rate": 3.869346733668342e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4681
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.09375,
      "epoch": 0.22622729029764205,
      "grad_norm": 0.0,
      "learning_rate": 3.8691051410900655e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4682
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.34375,
      "epoch": 0.22627560881329725,
      "grad_norm": 0.0,
      "learning_rate": 3.868863548511789e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4683
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 112.9375,
      "epoch": 0.22632392732895246,
      "grad_norm": 0.0,
      "learning_rate": 3.868621955933514e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 4684
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.0625,
      "epoch": 0.22637224584460766,
      "grad_norm": 0.0,
      "learning_rate": 3.8683803633552374e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4685
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.40625,
      "epoch": 0.22642056436026287,
      "grad_norm": 0.0,
      "learning_rate": 3.8681387707769615e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4686
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.03125,
      "epoch": 0.22646888287591804,
      "grad_norm": 0.0,
      "learning_rate": 3.8678971781986856e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4687
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.125,
      "epoch": 0.22651720139157325,
      "grad_norm": 0.0,
      "learning_rate": 3.86765558562041e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4688
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 118.96875,
      "epoch": 0.22656551990722845,
      "grad_norm": 8.838951940628537,
      "learning_rate": 3.8674139930421334e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4689
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.90625,
      "epoch": 0.22661383842288366,
      "grad_norm": 0.0,
      "learning_rate": 3.867172400463858e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4690
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.875,
      "epoch": 0.22666215693853886,
      "grad_norm": 0.0,
      "learning_rate": 3.8669308078855817e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4691
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.875,
      "epoch": 0.22671047545419404,
      "grad_norm": 0.0,
      "learning_rate": 3.8666892153073053e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4692
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.125,
      "epoch": 0.22675879396984924,
      "grad_norm": 0.0,
      "learning_rate": 3.8664476227290294e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4693
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.21875,
      "epoch": 0.22680711248550445,
      "grad_norm": 0.0,
      "learning_rate": 3.8662060301507535e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4694
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.53125,
      "epoch": 0.22685543100115965,
      "grad_norm": 0.0,
      "learning_rate": 3.8659644375724777e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4695
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.53125,
      "epoch": 0.22690374951681486,
      "grad_norm": 0.0,
      "learning_rate": 3.8657228449942013e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4696
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.375,
      "epoch": 0.22695206803247003,
      "grad_norm": 0.0,
      "learning_rate": 3.865481252415926e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4697
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.84375,
      "epoch": 0.22700038654812524,
      "grad_norm": 0.0,
      "learning_rate": 3.8652396598376496e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4698
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 114.59375,
      "epoch": 0.22704870506378044,
      "grad_norm": 0.0,
      "learning_rate": 3.8649980672593737e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4699
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.78125,
      "epoch": 0.22709702357943565,
      "grad_norm": 0.0,
      "learning_rate": 3.864756474681098e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4700
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.28125,
      "epoch": 0.22714534209509085,
      "grad_norm": 0.0,
      "learning_rate": 3.864514882102822e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4701
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 111.59375,
      "epoch": 0.22719366061074603,
      "grad_norm": 9.30042499053579,
      "learning_rate": 3.8642732895245456e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4702
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.9375,
      "epoch": 0.22724197912640123,
      "grad_norm": 0.0,
      "learning_rate": 3.864031696946269e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4703
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 121.59375,
      "epoch": 0.22729029764205644,
      "grad_norm": 0.0,
      "learning_rate": 3.863790104367994e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 4704
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 111.0625,
      "epoch": 0.22733861615771164,
      "grad_norm": 7.768093112175119,
      "learning_rate": 3.8635485117897175e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.20785759389400482,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4705
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.65625,
      "epoch": 0.22738693467336685,
      "grad_norm": 0.0,
      "learning_rate": 3.8633069192114416e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4706
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.78125,
      "epoch": 0.22743525318902202,
      "grad_norm": 0.0,
      "learning_rate": 3.8630653266331657e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4707
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.03125,
      "epoch": 0.22748357170467723,
      "grad_norm": 0.0,
      "learning_rate": 3.86282373405489e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4708
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.21875,
      "epoch": 0.22753189022033243,
      "grad_norm": 0.0,
      "learning_rate": 3.8625821414766135e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4709
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.96875,
      "epoch": 0.22758020873598764,
      "grad_norm": 0.0,
      "learning_rate": 3.862340548898338e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4710
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.6875,
      "epoch": 0.22762852725164284,
      "grad_norm": 0.0,
      "learning_rate": 3.862098956320062e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4711
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 109.25,
      "epoch": 0.22767684576729802,
      "grad_norm": 13.761906927228672,
      "learning_rate": 3.8618573637417853e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.26726123690605164,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4712
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.90625,
      "epoch": 0.22772516428295322,
      "grad_norm": 0.0,
      "learning_rate": 3.8616157711635095e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4713
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.8125,
      "epoch": 0.22777348279860843,
      "grad_norm": 0.0,
      "learning_rate": 3.8613741785852336e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4714
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.03125,
      "epoch": 0.22782180131426363,
      "grad_norm": 0.0,
      "learning_rate": 3.861132586006958e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4715
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.4375,
      "epoch": 0.22787011982991884,
      "grad_norm": 0.0,
      "learning_rate": 3.8608909934286814e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4716
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 117.28125,
      "epoch": 0.22791843834557401,
      "grad_norm": 0.0,
      "learning_rate": 3.860649400850406e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4717
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 104.125,
      "epoch": 0.22796675686122922,
      "grad_norm": 8.922190160328523,
      "learning_rate": 3.8604078082721296e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4718
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.9375,
      "epoch": 0.22801507537688442,
      "grad_norm": 0.0,
      "learning_rate": 3.860166215693854e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4719
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.21875,
      "epoch": 0.22806339389253963,
      "grad_norm": 0.0,
      "learning_rate": 3.859924623115578e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4720
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.8125,
      "epoch": 0.22811171240819483,
      "grad_norm": 0.0,
      "learning_rate": 3.859683030537302e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4721
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.84375,
      "epoch": 0.22816003092385,
      "grad_norm": 0.0,
      "learning_rate": 3.8594414379590257e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4722
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.6875,
      "epoch": 0.22820834943950521,
      "grad_norm": 0.0,
      "learning_rate": 3.859199845380749e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4723
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.21875,
      "epoch": 0.22825666795516042,
      "grad_norm": 0.0,
      "learning_rate": 3.858958252802474e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4724
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.875,
      "epoch": 0.22830498647081562,
      "grad_norm": 0.0,
      "learning_rate": 3.8587166602241975e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4725
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.40625,
      "epoch": 0.22835330498647083,
      "grad_norm": 0.0,
      "learning_rate": 3.8584750676459217e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4726
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.1875,
      "epoch": 0.228401623502126,
      "grad_norm": 0.0,
      "learning_rate": 3.858233475067646e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4727
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.34375,
      "epoch": 0.2284499420177812,
      "grad_norm": 0.0,
      "learning_rate": 3.85799188248937e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4728
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 116.15625,
      "epoch": 0.22849826053343641,
      "grad_norm": 0.0,
      "learning_rate": 3.8577502899110935e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4729
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.84375,
      "epoch": 0.22854657904909162,
      "grad_norm": 0.0,
      "learning_rate": 3.857508697332818e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4730
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.15625,
      "epoch": 0.22859489756474682,
      "grad_norm": 0.0,
      "learning_rate": 3.857267104754542e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4731
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.9375,
      "epoch": 0.228643216080402,
      "grad_norm": 0.0,
      "learning_rate": 3.857025512176266e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4732
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.78125,
      "epoch": 0.2286915345960572,
      "grad_norm": 0.0,
      "learning_rate": 3.8567839195979896e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4733
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.09375,
      "epoch": 0.2287398531117124,
      "grad_norm": 0.0,
      "learning_rate": 3.8565423270197137e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4734
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 108.3125,
      "epoch": 0.22878817162736761,
      "grad_norm": 10.087883818610702,
      "learning_rate": 3.856300734441438e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4735
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 109.84375,
      "epoch": 0.22883649014302282,
      "grad_norm": 5.587341486382611,
      "learning_rate": 3.8560591418631614e-07,
      "loss": -0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 4736
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.3125,
      "epoch": 0.228884808658678,
      "grad_norm": 0.0,
      "learning_rate": 3.855817549284886e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4737
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 111.03125,
      "epoch": 0.2289331271743332,
      "grad_norm": 0.0,
      "learning_rate": 3.8555759567066097e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 4738
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.875,
      "epoch": 0.2289814456899884,
      "grad_norm": 0.0,
      "learning_rate": 3.855334364128334e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4739
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.59375,
      "epoch": 0.2290297642056436,
      "grad_norm": 0.0,
      "learning_rate": 3.855092771550058e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4740
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 116.46875,
      "epoch": 0.22907808272129881,
      "grad_norm": 14.701111992089876,
      "learning_rate": 3.854851178971782e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.3636062741279602,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4741
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 107.59375,
      "epoch": 0.229126401236954,
      "grad_norm": 6.096097450570277,
      "learning_rate": 3.8546095863935057e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4742
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.0625,
      "epoch": 0.2291747197526092,
      "grad_norm": 0.0,
      "learning_rate": 3.8543679938152293e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4743
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.34375,
      "epoch": 0.2292230382682644,
      "grad_norm": 0.0,
      "learning_rate": 3.854126401236954e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4744
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.34375,
      "epoch": 0.2292713567839196,
      "grad_norm": 0.0,
      "learning_rate": 3.8538848086586776e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4745
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.625,
      "epoch": 0.2293196752995748,
      "grad_norm": 0.0,
      "learning_rate": 3.853643216080402e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4746
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.5,
      "epoch": 0.22936799381523,
      "grad_norm": 0.0,
      "learning_rate": 3.853401623502126e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4747
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.9375,
      "epoch": 0.2294163123308852,
      "grad_norm": 0.0,
      "learning_rate": 3.85316003092385e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4748
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.875,
      "epoch": 0.2294646308465404,
      "grad_norm": 0.0,
      "learning_rate": 3.8529184383455736e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4749
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.1875,
      "epoch": 0.2295129493621956,
      "grad_norm": 0.0,
      "learning_rate": 3.8526768457672983e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4750
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.375,
      "epoch": 0.2295612678778508,
      "grad_norm": 0.0,
      "learning_rate": 3.852435253189022e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4751
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 110.5625,
      "epoch": 0.22960958639350598,
      "grad_norm": 0.0,
      "learning_rate": 3.852193660610746e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 4752
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.34375,
      "epoch": 0.2296579049091612,
      "grad_norm": 0.0,
      "learning_rate": 3.8519520680324696e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4753
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 107.15625,
      "epoch": 0.2297062234248164,
      "grad_norm": 5.717516061097599,
      "learning_rate": 3.851710475454194e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4754
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.34375,
      "epoch": 0.2297545419404716,
      "grad_norm": 0.0,
      "learning_rate": 3.851468882875918e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4755
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.90625,
      "epoch": 0.2298028604561268,
      "grad_norm": 0.0,
      "learning_rate": 3.8512272902976415e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4756
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.78125,
      "epoch": 0.22985117897178198,
      "grad_norm": 0.0,
      "learning_rate": 3.850985697719366e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4757
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.40625,
      "epoch": 0.22989949748743718,
      "grad_norm": 0.0,
      "learning_rate": 3.85074410514109e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4758
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.375,
      "epoch": 0.2299478160030924,
      "grad_norm": 0.0,
      "learning_rate": 3.850502512562814e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4759
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 108.71875,
      "epoch": 0.2299961345187476,
      "grad_norm": 18.748339997609644,
      "learning_rate": 3.850260919984538e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4760
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.59375,
      "epoch": 0.2300444530344028,
      "grad_norm": 0.0,
      "learning_rate": 3.850019327406262e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4761
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.21875,
      "epoch": 0.23009277155005797,
      "grad_norm": 0.0,
      "learning_rate": 3.849777734827986e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4762
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.90625,
      "epoch": 0.23014109006571318,
      "grad_norm": 0.0,
      "learning_rate": 3.84953614224971e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4763
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.5,
      "epoch": 0.23018940858136838,
      "grad_norm": 0.0,
      "learning_rate": 3.849294549671434e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 4764
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 124.5625,
      "epoch": 0.2302377270970236,
      "grad_norm": 0.0,
      "learning_rate": 3.8490529570931577e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 4765
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 115.875,
      "epoch": 0.2302860456126788,
      "grad_norm": 9.892569543144578,
      "learning_rate": 3.848811364514882e-07,
      "loss": -0.0,
      "reward": 1.453125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 4766
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.84375,
      "epoch": 0.23033436412833397,
      "grad_norm": 0.0,
      "learning_rate": 3.848569771936606e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4767
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.25,
      "epoch": 0.23038268264398917,
      "grad_norm": 0.0,
      "learning_rate": 3.84832817935833e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4768
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.625,
      "epoch": 0.23043100115964438,
      "grad_norm": 0.0,
      "learning_rate": 3.8480865867800537e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4769
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 106.90625,
      "epoch": 0.23047931967529958,
      "grad_norm": 10.645060444438121,
      "learning_rate": 3.8478449942017784e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.2651395797729492,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4770
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.3125,
      "epoch": 0.2305276381909548,
      "grad_norm": 0.0,
      "learning_rate": 3.847603401623502e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4771
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.53125,
      "epoch": 0.23057595670660996,
      "grad_norm": 0.0,
      "learning_rate": 3.847361809045226e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4772
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.5,
      "epoch": 0.23062427522226517,
      "grad_norm": 0.0,
      "learning_rate": 3.8471202164669497e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4773
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 115.71875,
      "epoch": 0.23067259373792037,
      "grad_norm": 7.022194841848557,
      "learning_rate": 3.8468786238886744e-07,
      "loss": -0.0,
      "reward": 2.375,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4774
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.96875,
      "epoch": 0.23072091225357558,
      "grad_norm": 0.0,
      "learning_rate": 3.846637031310398e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4775
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 111.875,
      "epoch": 0.23076923076923078,
      "grad_norm": 0.0,
      "learning_rate": 3.8463954387321216e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4776
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.8125,
      "epoch": 0.23081754928488596,
      "grad_norm": 0.0,
      "learning_rate": 3.8461538461538463e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4777
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.25,
      "epoch": 0.23086586780054116,
      "grad_norm": 0.0,
      "learning_rate": 3.84591225357557e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4778
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.09375,
      "epoch": 0.23091418631619637,
      "grad_norm": 0.0,
      "learning_rate": 3.845670660997294e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4779
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 106.90625,
      "epoch": 0.23096250483185157,
      "grad_norm": 7.908102388781982,
      "learning_rate": 3.845429068419018e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.2920827269554138,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4780
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 109.96875,
      "epoch": 0.23101082334750678,
      "grad_norm": 5.374095641077153,
      "learning_rate": 3.8451874758407423e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4781
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.90625,
      "epoch": 0.23105914186316195,
      "grad_norm": 0.0,
      "learning_rate": 3.844945883262466e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4782
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.21875,
      "epoch": 0.23110746037881716,
      "grad_norm": 0.0,
      "learning_rate": 3.84470429068419e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4783
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 211.34375,
      "epoch": 0.23115577889447236,
      "grad_norm": 6.107268635702705,
      "learning_rate": 3.844462698105914e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.36722296476364136,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 0.9375,
      "step": 4784
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 118.5,
      "epoch": 0.23120409741012757,
      "grad_norm": 0.0,
      "learning_rate": 3.844221105527638e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4785
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.9375,
      "epoch": 0.23125241592578277,
      "grad_norm": 0.0,
      "learning_rate": 3.843979512949362e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4786
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.625,
      "epoch": 0.23130073444143795,
      "grad_norm": 0.0,
      "learning_rate": 3.843737920371086e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4787
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 112.9375,
      "epoch": 0.23134905295709315,
      "grad_norm": 7.7033937929324106,
      "learning_rate": 3.84349632779281e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19999711215496063,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4788
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.90625,
      "epoch": 0.23139737147274836,
      "grad_norm": 0.0,
      "learning_rate": 3.843254735214534e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4789
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.25,
      "epoch": 0.23144568998840356,
      "grad_norm": 0.0,
      "learning_rate": 3.8430131426362584e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4790
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 106.40625,
      "epoch": 0.23149400850405877,
      "grad_norm": 44.739185133923826,
      "learning_rate": 3.842771550057982e-07,
      "loss": -0.0,
      "reward": 2.328125,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4791
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.34375,
      "epoch": 0.23154232701971394,
      "grad_norm": 0.0,
      "learning_rate": 3.842529957479706e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4792
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.3125,
      "epoch": 0.23159064553536915,
      "grad_norm": 0.0,
      "learning_rate": 3.84228836490143e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4793
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.0625,
      "epoch": 0.23163896405102435,
      "grad_norm": 0.0,
      "learning_rate": 3.8420467723231545e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4794
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.625,
      "epoch": 0.23168728256667956,
      "grad_norm": 0.0,
      "learning_rate": 3.841805179744878e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4795
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.15625,
      "epoch": 0.23173560108233476,
      "grad_norm": 0.0,
      "learning_rate": 3.8415635871666017e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4796
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.71875,
      "epoch": 0.23178391959798994,
      "grad_norm": 0.0,
      "learning_rate": 3.8413219945883263e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4797
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 125.1875,
      "epoch": 0.23183223811364514,
      "grad_norm": 0.0,
      "learning_rate": 3.84108040201005e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4798
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.625,
      "epoch": 0.23188055662930035,
      "grad_norm": 0.0,
      "learning_rate": 3.840838809431774e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4799
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.46875,
      "epoch": 0.23192887514495555,
      "grad_norm": 0.0,
      "learning_rate": 3.840597216853498e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4800
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 104.15625,
      "epoch": 0.23197719366061076,
      "grad_norm": 10.165846770216513,
      "learning_rate": 3.8403556242752224e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4801
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 163.96875,
      "epoch": 0.23202551217626594,
      "grad_norm": 1.7009763043025006,
      "learning_rate": 3.840114031696946e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.28778618574142456,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 0.96875,
      "step": 4802
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 112.6875,
      "epoch": 0.23207383069192114,
      "grad_norm": 0.0,
      "learning_rate": 3.83987243911867e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4803
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.4375,
      "epoch": 0.23212214920757634,
      "grad_norm": 0.0,
      "learning_rate": 3.839630846540394e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4804
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.5,
      "epoch": 0.23217046772323155,
      "grad_norm": 0.0,
      "learning_rate": 3.839389253962118e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4805
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 157.96875,
      "epoch": 0.23221878623888675,
      "grad_norm": 1.6079076521407947,
      "learning_rate": 3.839147661383842e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 0.96875,
      "step": 4806
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.71875,
      "epoch": 0.23226710475454193,
      "grad_norm": 0.0,
      "learning_rate": 3.838906068805566e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4807
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 128.59375,
      "epoch": 0.23231542327019714,
      "grad_norm": 0.0,
      "learning_rate": 3.83866447622729e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4808
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.71875,
      "epoch": 0.23236374178585234,
      "grad_norm": 0.0,
      "learning_rate": 3.838422883649014e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4809
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.75,
      "epoch": 0.23241206030150754,
      "grad_norm": 0.0,
      "learning_rate": 3.8381812910707385e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4810
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 196.71875,
      "epoch": 0.23246037881716275,
      "grad_norm": 1.5412727743942811,
      "learning_rate": 3.837939698492462e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 0.96875,
      "step": 4811
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 128.0625,
      "epoch": 0.23250869733281793,
      "grad_norm": 0.0,
      "learning_rate": 3.8376981059141863e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4812
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.875,
      "epoch": 0.23255701584847313,
      "grad_norm": 0.0,
      "learning_rate": 3.83745651333591e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4813
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 178.0,
      "epoch": 0.23260533436412834,
      "grad_norm": 9.20884367182524,
      "learning_rate": 3.8372149207576345e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.45561128854751587,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 0.96875,
      "step": 4814
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 109.53125,
      "epoch": 0.23265365287978354,
      "grad_norm": 0.0,
      "learning_rate": 3.836973328179358e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4815
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 172.96875,
      "epoch": 0.23270197139543874,
      "grad_norm": 0.0,
      "learning_rate": 3.836731735601082e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 0.96875,
      "step": 4816
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.15625,
      "epoch": 0.23275028991109392,
      "grad_norm": 0.0,
      "learning_rate": 3.8364901430228064e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4817
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 232.375,
      "epoch": 0.23279860842674913,
      "grad_norm": 6.365262908739057,
      "learning_rate": 3.83624855044453e-07,
      "loss": -0.0,
      "reward": 2.21875,
      "reward_std": 0.4187060594558716,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 0.9375,
      "step": 4818
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.5625,
      "epoch": 0.23284692694240433,
      "grad_norm": 0.0,
      "learning_rate": 3.836006957866254e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4819
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 120.8125,
      "epoch": 0.23289524545805954,
      "grad_norm": 0.0,
      "learning_rate": 3.8357653652879783e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 4820
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.03125,
      "epoch": 0.23294356397371474,
      "grad_norm": 0.0,
      "learning_rate": 3.8355237727097024e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4821
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.78125,
      "epoch": 0.23299188248936992,
      "grad_norm": 0.0,
      "learning_rate": 3.835282180131426e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4822
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.03125,
      "epoch": 0.23304020100502512,
      "grad_norm": 0.0,
      "learning_rate": 3.83504058755315e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4823
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 110.8125,
      "epoch": 0.23308851952068033,
      "grad_norm": 7.413637543727143,
      "learning_rate": 3.8347989949748743e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.2226182520389557,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4824
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 231.34375,
      "epoch": 0.23313683803633553,
      "grad_norm": 3.5032038177633402,
      "learning_rate": 3.8345574023965985e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.36722299456596375,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 0.9375,
      "step": 4825
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 167.46875,
      "epoch": 0.23318515655199074,
      "grad_norm": 0.0,
      "learning_rate": 3.834315809818322e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 0.96875,
      "step": 4826
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 166.9375,
      "epoch": 0.2332334750676459,
      "grad_norm": 2.5382582289963285,
      "learning_rate": 3.834074217240046e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.35247981548309326,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 0.96875,
      "step": 4827
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.8125,
      "epoch": 0.23328179358330112,
      "grad_norm": 0.0,
      "learning_rate": 3.8338326246617703e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4828
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.375,
      "epoch": 0.23333011209895632,
      "grad_norm": 0.0,
      "learning_rate": 3.833591032083494e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4829
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.65625,
      "epoch": 0.23337843061461153,
      "grad_norm": 0.0,
      "learning_rate": 3.8333494395052186e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4830
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 147.21875,
      "epoch": 0.23342674913026673,
      "grad_norm": 0.0,
      "learning_rate": 3.833107846926942e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 0.96875,
      "step": 4831
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.90625,
      "epoch": 0.2334750676459219,
      "grad_norm": 0.0,
      "learning_rate": 3.8328662543486663e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4832
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.5625,
      "epoch": 0.2335233861615771,
      "grad_norm": 0.0,
      "learning_rate": 3.83262466177039e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4833
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.90625,
      "epoch": 0.23357170467723232,
      "grad_norm": 0.0,
      "learning_rate": 3.8323830691921146e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4834
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.8125,
      "epoch": 0.23362002319288752,
      "grad_norm": 0.0,
      "learning_rate": 3.832141476613838e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4835
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.9375,
      "epoch": 0.23366834170854273,
      "grad_norm": 0.0,
      "learning_rate": 3.831899884035562e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4836
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 161.65625,
      "epoch": 0.2337166602241979,
      "grad_norm": 0.0,
      "learning_rate": 3.8316582914572865e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 0.96875,
      "step": 4837
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.0,
      "epoch": 0.2337649787398531,
      "grad_norm": 0.0,
      "learning_rate": 3.83141669887901e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4838
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.28125,
      "epoch": 0.2338132972555083,
      "grad_norm": 0.0,
      "learning_rate": 3.831175106300734e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4839
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 117.8125,
      "epoch": 0.23386161577116352,
      "grad_norm": 119.6415519953957,
      "learning_rate": 3.8309335137224584e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4840
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.625,
      "epoch": 0.23390993428681872,
      "grad_norm": 0.0,
      "learning_rate": 3.8306919211441825e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4841
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.125,
      "epoch": 0.2339582528024739,
      "grad_norm": 0.0,
      "learning_rate": 3.830450328565906e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4842
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 195.5,
      "epoch": 0.2340065713181291,
      "grad_norm": 5.07885146177238,
      "learning_rate": 3.83020873598763e-07,
      "loss": -0.0,
      "reward": 2.28125,
      "reward_std": 0.405046284198761,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 0.9375,
      "step": 4843
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.15625,
      "epoch": 0.2340548898337843,
      "grad_norm": 0.0,
      "learning_rate": 3.8299671434093544e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4844
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.90625,
      "epoch": 0.2341032083494395,
      "grad_norm": 0.0,
      "learning_rate": 3.8297255508310785e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4845
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 93.46875,
      "epoch": 0.23415152686509472,
      "grad_norm": 11.119394436418519,
      "learning_rate": 3.829483958252802e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.2805348336696625,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4846
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.1875,
      "epoch": 0.2341998453807499,
      "grad_norm": 0.0,
      "learning_rate": 3.8292423656745263e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4847
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.75,
      "epoch": 0.2342481638964051,
      "grad_norm": 0.0,
      "learning_rate": 3.8290007730962504e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 4848
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.875,
      "epoch": 0.2342964824120603,
      "grad_norm": 0.0,
      "learning_rate": 3.828759180517974e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4849
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 135.28125,
      "epoch": 0.2343448009277155,
      "grad_norm": 0.0,
      "learning_rate": 3.8285175879396987e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 0.96875,
      "step": 4850
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.875,
      "epoch": 0.2343931194433707,
      "grad_norm": 0.0,
      "learning_rate": 3.8282759953614223e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4851
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.78125,
      "epoch": 0.2344414379590259,
      "grad_norm": 0.0,
      "learning_rate": 3.8280344027831464e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4852
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 91.625,
      "epoch": 0.2344897564746811,
      "grad_norm": 20.44379140028743,
      "learning_rate": 3.82779281020487e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.28778618574142456,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4853
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.25,
      "epoch": 0.2345380749903363,
      "grad_norm": 0.0,
      "learning_rate": 3.8275512176265947e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4854
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.21875,
      "epoch": 0.2345863935059915,
      "grad_norm": 0.0,
      "learning_rate": 3.8273096250483183e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4855
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.34375,
      "epoch": 0.2346347120216467,
      "grad_norm": 0.0,
      "learning_rate": 3.827068032470042e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4856
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.8125,
      "epoch": 0.23468303053730188,
      "grad_norm": 0.0,
      "learning_rate": 3.8268264398917666e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4857
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.84375,
      "epoch": 0.2347313490529571,
      "grad_norm": 0.0,
      "learning_rate": 3.82658484731349e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4858
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.9375,
      "epoch": 0.2347796675686123,
      "grad_norm": 0.0,
      "learning_rate": 3.8263432547352143e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4859
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.9375,
      "epoch": 0.2348279860842675,
      "grad_norm": 0.0,
      "learning_rate": 3.8261016621569385e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4860
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.46875,
      "epoch": 0.2348763045999227,
      "grad_norm": 0.0,
      "learning_rate": 3.8258600695786626e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 4861
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.1875,
      "epoch": 0.23492462311557788,
      "grad_norm": 0.0,
      "learning_rate": 3.825618477000386e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4862
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.6875,
      "epoch": 0.23497294163123308,
      "grad_norm": 0.0,
      "learning_rate": 3.8253768844221103e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4863
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 127.71875,
      "epoch": 0.2350212601468883,
      "grad_norm": 0.0,
      "learning_rate": 3.8251352918438345e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 0.96875,
      "step": 4864
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.40625,
      "epoch": 0.2350695786625435,
      "grad_norm": 0.0,
      "learning_rate": 3.8248936992655586e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4865
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.0625,
      "epoch": 0.2351178971781987,
      "grad_norm": 0.0,
      "learning_rate": 3.824652106687282e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4866
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.8125,
      "epoch": 0.23516621569385388,
      "grad_norm": 0.0,
      "learning_rate": 3.8244105141090064e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 4867
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.75,
      "epoch": 0.23521453420950908,
      "grad_norm": 0.0,
      "learning_rate": 3.8241689215307305e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4868
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.78125,
      "epoch": 0.23526285272516428,
      "grad_norm": 0.0,
      "learning_rate": 3.823927328952454e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 4869
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.375,
      "epoch": 0.2353111712408195,
      "grad_norm": 0.0,
      "learning_rate": 3.823685736374179e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4870
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.09375,
      "epoch": 0.2353594897564747,
      "grad_norm": 0.0,
      "learning_rate": 3.8234441437959024e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4871
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.5,
      "epoch": 0.23540780827212987,
      "grad_norm": 0.0,
      "learning_rate": 3.8232025512176265e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4872
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.53125,
      "epoch": 0.23545612678778507,
      "grad_norm": 0.0,
      "learning_rate": 3.82296095863935e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4873
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.78125,
      "epoch": 0.23550444530344028,
      "grad_norm": 0.0,
      "learning_rate": 3.822719366061075e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4874
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.84375,
      "epoch": 0.23555276381909548,
      "grad_norm": 0.0,
      "learning_rate": 3.8224777734827984e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4875
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.0,
      "epoch": 0.2356010823347507,
      "grad_norm": 0.0,
      "learning_rate": 3.8222361809045225e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4876
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.59375,
      "epoch": 0.23564940085040587,
      "grad_norm": 0.0,
      "learning_rate": 3.8219945883262467e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 4877
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.84375,
      "epoch": 0.23569771936606107,
      "grad_norm": 0.0,
      "learning_rate": 3.8217529957479703e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4878
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.375,
      "epoch": 0.23574603788171627,
      "grad_norm": 0.0,
      "learning_rate": 3.8215114031696944e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4879
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.25,
      "epoch": 0.23579435639737148,
      "grad_norm": 0.0,
      "learning_rate": 3.8212698105914185e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4880
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.71875,
      "epoch": 0.23584267491302668,
      "grad_norm": 0.0,
      "learning_rate": 3.8210282180131427e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4881
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.875,
      "epoch": 0.23589099342868186,
      "grad_norm": 0.0,
      "learning_rate": 3.8207866254348663e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4882
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.09375,
      "epoch": 0.23593931194433707,
      "grad_norm": 0.0,
      "learning_rate": 3.8205450328565904e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 4883
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.40625,
      "epoch": 0.23598763045999227,
      "grad_norm": 0.0,
      "learning_rate": 3.8203034402783146e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4884
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 99.75,
      "epoch": 0.23603594897564747,
      "grad_norm": 24.43439299581009,
      "learning_rate": 3.8200618477000387e-07,
      "loss": -0.0,
      "reward": 2.40625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4885
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.53125,
      "epoch": 0.23608426749130268,
      "grad_norm": 0.0,
      "learning_rate": 3.8198202551217623e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4886
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 105.875,
      "epoch": 0.23613258600695786,
      "grad_norm": 26.620510406547908,
      "learning_rate": 3.819578662543487e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 4887
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 131.6875,
      "epoch": 0.23618090452261306,
      "grad_norm": 0.0,
      "learning_rate": 3.8193370699652106e-07,
      "loss": 0.0,
      "reward": 1.28125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 0.96875,
      "step": 4888
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.1875,
      "epoch": 0.23622922303826827,
      "grad_norm": 0.0,
      "learning_rate": 3.819095477386934e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4889
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.125,
      "epoch": 0.23627754155392347,
      "grad_norm": 0.0,
      "learning_rate": 3.818853884808659e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4890
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.5,
      "epoch": 0.23632586006957867,
      "grad_norm": 0.0,
      "learning_rate": 3.8186122922303824e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4891
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.5,
      "epoch": 0.23637417858523385,
      "grad_norm": 0.0,
      "learning_rate": 3.8183706996521066e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4892
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.1875,
      "epoch": 0.23642249710088906,
      "grad_norm": 0.0,
      "learning_rate": 3.81812910707383e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4893
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.71875,
      "epoch": 0.23647081561654426,
      "grad_norm": 0.0,
      "learning_rate": 3.817887514495555e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4894
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.5,
      "epoch": 0.23651913413219947,
      "grad_norm": 0.0,
      "learning_rate": 3.8176459219172785e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4895
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.78125,
      "epoch": 0.23656745264785467,
      "grad_norm": 0.0,
      "learning_rate": 3.8174043293390026e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4896
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.28125,
      "epoch": 0.23661577116350985,
      "grad_norm": 0.0,
      "learning_rate": 3.817162736760727e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4897
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.625,
      "epoch": 0.23666408967916505,
      "grad_norm": 0.0,
      "learning_rate": 3.8169211441824503e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4898
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.71875,
      "epoch": 0.23671240819482026,
      "grad_norm": 0.0,
      "learning_rate": 3.8166795516041745e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4899
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.09375,
      "epoch": 0.23676072671047546,
      "grad_norm": 0.0,
      "learning_rate": 3.8164379590258986e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4900
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.375,
      "epoch": 0.23680904522613067,
      "grad_norm": 0.0,
      "learning_rate": 3.816196366447623e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4901
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 126.375,
      "epoch": 0.23685736374178584,
      "grad_norm": 0.0,
      "learning_rate": 3.8159547738693464e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 0.96875,
      "step": 4902
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.5625,
      "epoch": 0.23690568225744105,
      "grad_norm": 0.0,
      "learning_rate": 3.8157131812910705e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4903
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.34375,
      "epoch": 0.23695400077309625,
      "grad_norm": 0.0,
      "learning_rate": 3.8154715887127946e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4904
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.15625,
      "epoch": 0.23700231928875146,
      "grad_norm": 0.0,
      "learning_rate": 3.815229996134519e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4905
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.53125,
      "epoch": 0.23705063780440666,
      "grad_norm": 0.0,
      "learning_rate": 3.8149884035562424e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4906
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.75,
      "epoch": 0.23709895632006184,
      "grad_norm": 0.0,
      "learning_rate": 3.814746810977967e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4907
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.34375,
      "epoch": 0.23714727483571704,
      "grad_norm": 0.0,
      "learning_rate": 3.8145052183996906e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.19044628739356995,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4908
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.71875,
      "epoch": 0.23719559335137225,
      "grad_norm": 0.0,
      "learning_rate": 3.814263625821414e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4909
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.5,
      "epoch": 0.23724391186702745,
      "grad_norm": 0.0,
      "learning_rate": 3.814022033243139e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4910
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.03125,
      "epoch": 0.23729223038268266,
      "grad_norm": 0.0,
      "learning_rate": 3.8137804406648625e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4911
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.28125,
      "epoch": 0.23734054889833783,
      "grad_norm": 0.0,
      "learning_rate": 3.8135388480865867e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4912
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 101.28125,
      "epoch": 0.23738886741399304,
      "grad_norm": 95.41568900123983,
      "learning_rate": 3.8132972555083103e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4913
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.21875,
      "epoch": 0.23743718592964824,
      "grad_norm": 0.0,
      "learning_rate": 3.813055662930035e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4914
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.3125,
      "epoch": 0.23748550444530345,
      "grad_norm": 0.0,
      "learning_rate": 3.8128140703517585e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4915
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.5625,
      "epoch": 0.23753382296095865,
      "grad_norm": 0.0,
      "learning_rate": 3.8125724777734827e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 4916
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.5625,
      "epoch": 0.23758214147661383,
      "grad_norm": 0.0,
      "learning_rate": 3.812330885195207e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4917
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.21875,
      "epoch": 0.23763045999226903,
      "grad_norm": 0.0,
      "learning_rate": 3.8120892926169304e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 4918
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.15625,
      "epoch": 0.23767877850792424,
      "grad_norm": 0.0,
      "learning_rate": 3.8118477000386546e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4919
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.28125,
      "epoch": 0.23772709702357944,
      "grad_norm": 0.0,
      "learning_rate": 3.8116061074603787e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4920
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.53125,
      "epoch": 0.23777541553923465,
      "grad_norm": 0.0,
      "learning_rate": 3.811364514882103e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4921
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.3125,
      "epoch": 0.23782373405488982,
      "grad_norm": 0.0,
      "learning_rate": 3.8111229223038264e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4922
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.78125,
      "epoch": 0.23787205257054503,
      "grad_norm": 0.0,
      "learning_rate": 3.8108813297255506e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4923
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.21875,
      "epoch": 0.23792037108620023,
      "grad_norm": 0.0,
      "learning_rate": 3.8106397371472747e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4924
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.09375,
      "epoch": 0.23796868960185544,
      "grad_norm": 0.0,
      "learning_rate": 3.810398144568999e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4925
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.28125,
      "epoch": 0.23801700811751064,
      "grad_norm": 0.0,
      "learning_rate": 3.8101565519907225e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4926
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.625,
      "epoch": 0.23806532663316582,
      "grad_norm": 0.0,
      "learning_rate": 3.809914959412447e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4927
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 99.28125,
      "epoch": 0.23811364514882102,
      "grad_norm": 36.33446843164056,
      "learning_rate": 3.8096733668341707e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.3051002323627472,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4928
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.09375,
      "epoch": 0.23816196366447623,
      "grad_norm": 0.0,
      "learning_rate": 3.8094317742558943e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4929
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.3125,
      "epoch": 0.23821028218013143,
      "grad_norm": 0.0,
      "learning_rate": 3.809190181677619e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4930
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 101.875,
      "epoch": 0.23825860069578664,
      "grad_norm": 121.59015460251467,
      "learning_rate": 3.8089485890993426e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4931
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.03125,
      "epoch": 0.23830691921144181,
      "grad_norm": 0.0,
      "learning_rate": 3.808706996521067e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4932
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.5,
      "epoch": 0.23835523772709702,
      "grad_norm": 0.0,
      "learning_rate": 3.8084654039427903e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4933
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.625,
      "epoch": 0.23840355624275222,
      "grad_norm": 0.0,
      "learning_rate": 3.808223811364515e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4934
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.09375,
      "epoch": 0.23845187475840743,
      "grad_norm": 0.0,
      "learning_rate": 3.8079822187862386e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4935
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.21875,
      "epoch": 0.23850019327406263,
      "grad_norm": 0.0,
      "learning_rate": 3.807740626207963e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4936
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.375,
      "epoch": 0.2385485117897178,
      "grad_norm": 0.0,
      "learning_rate": 3.807499033629687e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4937
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.4375,
      "epoch": 0.23859683030537301,
      "grad_norm": 0.0,
      "learning_rate": 3.807257441051411e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4938
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.90625,
      "epoch": 0.23864514882102822,
      "grad_norm": 0.0,
      "learning_rate": 3.8070158484731346e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4939
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.8125,
      "epoch": 0.23869346733668342,
      "grad_norm": 0.0,
      "learning_rate": 3.806774255894859e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4940
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.375,
      "epoch": 0.23874178585233863,
      "grad_norm": 0.0,
      "learning_rate": 3.806532663316583e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4941
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.15625,
      "epoch": 0.2387901043679938,
      "grad_norm": 0.0,
      "learning_rate": 3.8062910707383065e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4942
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.21875,
      "epoch": 0.238838422883649,
      "grad_norm": 0.0,
      "learning_rate": 3.8060494781600307e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4943
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.125,
      "epoch": 0.23888674139930421,
      "grad_norm": 0.0,
      "learning_rate": 3.805807885581755e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4944
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.8125,
      "epoch": 0.23893505991495942,
      "grad_norm": 0.0,
      "learning_rate": 3.805566293003479e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 4945
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.4375,
      "epoch": 0.23898337843061462,
      "grad_norm": 9.781988833042197,
      "learning_rate": 3.8053247004252025e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4946
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.21875,
      "epoch": 0.2390316969462698,
      "grad_norm": 0.0,
      "learning_rate": 3.805083107846927e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 4947
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.53125,
      "epoch": 0.239080015461925,
      "grad_norm": 0.0,
      "learning_rate": 3.804841515268651e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4948
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 96.75,
      "epoch": 0.2391283339775802,
      "grad_norm": 0.0,
      "learning_rate": 3.8045999226903744e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 4949
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.53125,
      "epoch": 0.23917665249323541,
      "grad_norm": 0.0,
      "learning_rate": 3.804358330112099e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4950
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.59375,
      "epoch": 0.23922497100889062,
      "grad_norm": 0.0,
      "learning_rate": 3.8041167375338227e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4951
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.0625,
      "epoch": 0.2392732895245458,
      "grad_norm": 0.0,
      "learning_rate": 3.803875144955547e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4952
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 133.53125,
      "epoch": 0.239321608040201,
      "grad_norm": 0.0,
      "learning_rate": 3.8036335523772704e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 0.96875,
      "step": 4953
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.15625,
      "epoch": 0.2393699265558562,
      "grad_norm": 0.0,
      "learning_rate": 3.803391959798995e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4954
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.1875,
      "epoch": 0.2394182450715114,
      "grad_norm": 0.0,
      "learning_rate": 3.8031503672207187e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4955
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.9375,
      "epoch": 0.23946656358716661,
      "grad_norm": 0.0,
      "learning_rate": 3.802908774642443e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4956
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.59375,
      "epoch": 0.2395148821028218,
      "grad_norm": 0.0,
      "learning_rate": 3.802667182064167e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4957
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.75,
      "epoch": 0.239563200618477,
      "grad_norm": 0.0,
      "learning_rate": 3.802425589485891e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4958
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.84375,
      "epoch": 0.2396115191341322,
      "grad_norm": 21.567135461476024,
      "learning_rate": 3.8021839969076147e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4959
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 196.15625,
      "epoch": 0.2396598376497874,
      "grad_norm": 3.89694086057857,
      "learning_rate": 3.801942404329339e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.36532074213027954,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 0.9375,
      "step": 4960
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.0625,
      "epoch": 0.2397081561654426,
      "grad_norm": 0.0,
      "learning_rate": 3.801700811751063e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4961
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.375,
      "epoch": 0.2397564746810978,
      "grad_norm": 0.0,
      "learning_rate": 3.8014592191727866e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 4962
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.15625,
      "epoch": 0.239804793196753,
      "grad_norm": 0.0,
      "learning_rate": 3.8012176265945107e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4963
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.90625,
      "epoch": 0.2398531117124082,
      "grad_norm": 0.0,
      "learning_rate": 3.800976034016235e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 4964
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 95.875,
      "epoch": 0.2399014302280634,
      "grad_norm": 8.361384983664509,
      "learning_rate": 3.800734441437959e-07,
      "loss": -0.0,
      "reward": 2.375,
      "reward_std": 0.2925041913986206,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4965
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.5,
      "epoch": 0.2399497487437186,
      "grad_norm": 0.0,
      "learning_rate": 3.8004928488596826e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 4966
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.125,
      "epoch": 0.23999806725937378,
      "grad_norm": 0.0,
      "learning_rate": 3.8002512562814073e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4967
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.625,
      "epoch": 0.240046385775029,
      "grad_norm": 0.0,
      "learning_rate": 3.800009663703131e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4968
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.5,
      "epoch": 0.2400947042906842,
      "grad_norm": 0.0,
      "learning_rate": 3.799768071124855e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4969
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.75,
      "epoch": 0.2401430228063394,
      "grad_norm": 0.0,
      "learning_rate": 3.799526478546579e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4970
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 118.375,
      "epoch": 0.2401913413219946,
      "grad_norm": 10.121083972059695,
      "learning_rate": 3.799284885968303e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.37677380442619324,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 0.96875,
      "step": 4971
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 97.1875,
      "epoch": 0.24023965983764978,
      "grad_norm": 0.0,
      "learning_rate": 3.799043293390027e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 4972
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.5625,
      "epoch": 0.24028797835330498,
      "grad_norm": 0.0,
      "learning_rate": 3.7988017008117505e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 4973
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.65625,
      "epoch": 0.2403362968689602,
      "grad_norm": 0.0,
      "learning_rate": 3.798560108233475e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4974
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.59375,
      "epoch": 0.2403846153846154,
      "grad_norm": 0.0,
      "learning_rate": 3.798318515655199e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4975
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.6875,
      "epoch": 0.2404329339002706,
      "grad_norm": 0.0,
      "learning_rate": 3.798076923076923e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4976
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.96875,
      "epoch": 0.24048125241592577,
      "grad_norm": 0.0,
      "learning_rate": 3.797835330498647e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4977
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.15625,
      "epoch": 0.24052957093158098,
      "grad_norm": 0.0,
      "learning_rate": 3.797593737920371e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4978
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.5,
      "epoch": 0.24057788944723618,
      "grad_norm": 0.0,
      "learning_rate": 3.797352145342095e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4979
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 95.53125,
      "epoch": 0.2406262079628914,
      "grad_norm": 8.826775972875524,
      "learning_rate": 3.7971105527638195e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4980
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.21875,
      "epoch": 0.2406745264785466,
      "grad_norm": 0.0,
      "learning_rate": 3.796868960185543e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 4981
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.875,
      "epoch": 0.24072284499420177,
      "grad_norm": 0.0,
      "learning_rate": 3.7966273676072667e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4982
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.15625,
      "epoch": 0.24077116350985697,
      "grad_norm": 0.0,
      "learning_rate": 3.796385775028991e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 4983
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.0,
      "epoch": 0.24081948202551218,
      "grad_norm": 0.0,
      "learning_rate": 3.796144182450715e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4984
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.21875,
      "epoch": 0.24086780054116738,
      "grad_norm": 0.0,
      "learning_rate": 3.795902589872439e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 4985
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 84.03125,
      "epoch": 0.2409161190568226,
      "grad_norm": 20.863054780876567,
      "learning_rate": 3.7956609972941627e-07,
      "loss": 0.0,
      "reward": 1.5625,
      "reward_std": 0.18857255578041077,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 4986
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.59375,
      "epoch": 0.24096443757247776,
      "grad_norm": 0.0,
      "learning_rate": 3.7954194047158874e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 4987
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.28125,
      "epoch": 0.24101275608813297,
      "grad_norm": 0.0,
      "learning_rate": 3.795177812137611e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 4988
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.875,
      "epoch": 0.24106107460378817,
      "grad_norm": 0.0,
      "learning_rate": 3.794936219559335e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 4989
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.3125,
      "epoch": 0.24110939311944338,
      "grad_norm": 0.0,
      "learning_rate": 3.794694626981059e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 4990
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 99.03125,
      "epoch": 0.24115771163509858,
      "grad_norm": 17.238032552515932,
      "learning_rate": 3.794453034402783e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4991
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.34375,
      "epoch": 0.24120603015075376,
      "grad_norm": 0.0,
      "learning_rate": 3.794211441824507e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4992
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.59375,
      "epoch": 0.24125434866640896,
      "grad_norm": 0.0,
      "learning_rate": 3.7939698492462306e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 4993
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.5,
      "epoch": 0.24130266718206417,
      "grad_norm": 0.0,
      "learning_rate": 3.793728256667955e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 4994
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.625,
      "epoch": 0.24135098569771937,
      "grad_norm": 0.0,
      "learning_rate": 3.793486664089679e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 4995
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.78125,
      "epoch": 0.24139930421337458,
      "grad_norm": 0.0,
      "learning_rate": 3.793245071511403e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 4996
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.3125,
      "epoch": 0.24144762272902975,
      "grad_norm": 0.0,
      "learning_rate": 3.793003478933127e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 4997
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.34375,
      "epoch": 0.24149594124468496,
      "grad_norm": 0.0,
      "learning_rate": 3.7927618863548513e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 4998
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.0,
      "epoch": 0.24154425976034016,
      "grad_norm": 0.0,
      "learning_rate": 3.792520293776575e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 4999
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.3125,
      "epoch": 0.24159257827599537,
      "grad_norm": 0.0,
      "learning_rate": 3.7922787011982995e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5000
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.84375,
      "epoch": 0.24164089679165057,
      "grad_norm": 0.0,
      "learning_rate": 3.792037108620023e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5001
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.125,
      "epoch": 0.24168921530730575,
      "grad_norm": 0.0,
      "learning_rate": 3.791795516041747e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5002
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.5625,
      "epoch": 0.24173753382296095,
      "grad_norm": 0.0,
      "learning_rate": 3.791553923463471e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5003
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.84375,
      "epoch": 0.24178585233861616,
      "grad_norm": 0.0,
      "learning_rate": 3.791312330885195e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5004
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.25,
      "epoch": 0.24183417085427136,
      "grad_norm": 0.0,
      "learning_rate": 3.791070738306919e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5005
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.75,
      "epoch": 0.24188248936992657,
      "grad_norm": 0.0,
      "learning_rate": 3.790829145728643e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5006
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 84.78125,
      "epoch": 0.24193080788558174,
      "grad_norm": 15.739811798234422,
      "learning_rate": 3.7905875531503674e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5007
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.46875,
      "epoch": 0.24197912640123695,
      "grad_norm": 0.0,
      "learning_rate": 3.790345960572091e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 5008
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.5625,
      "epoch": 0.24202744491689215,
      "grad_norm": 0.0,
      "learning_rate": 3.790104367993815e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5009
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.3125,
      "epoch": 0.24207576343254736,
      "grad_norm": 0.0,
      "learning_rate": 3.7898627754155393e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5010
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.5,
      "epoch": 0.24212408194820256,
      "grad_norm": 0.0,
      "learning_rate": 3.789621182837263e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5011
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.25,
      "epoch": 0.24217240046385774,
      "grad_norm": 0.0,
      "learning_rate": 3.789379590258987e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5012
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.15625,
      "epoch": 0.24222071897951294,
      "grad_norm": 0.0,
      "learning_rate": 3.7891379976807107e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5013
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.90625,
      "epoch": 0.24226903749516815,
      "grad_norm": 0.0,
      "learning_rate": 3.7888964051024353e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5014
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.0625,
      "epoch": 0.24231735601082335,
      "grad_norm": 0.0,
      "learning_rate": 3.788654812524159e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5015
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.375,
      "epoch": 0.24236567452647856,
      "grad_norm": 0.0,
      "learning_rate": 3.788413219945883e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5016
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.78125,
      "epoch": 0.24241399304213374,
      "grad_norm": 0.0,
      "learning_rate": 3.788171627367607e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5017
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.59375,
      "epoch": 0.24246231155778894,
      "grad_norm": 0.0,
      "learning_rate": 3.7879300347893313e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5018
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.90625,
      "epoch": 0.24251063007344414,
      "grad_norm": 0.0,
      "learning_rate": 3.787688442211055e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 5019
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 88.34375,
      "epoch": 0.24255894858909935,
      "grad_norm": 15.401580423092959,
      "learning_rate": 3.7874468496327796e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5020
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.875,
      "epoch": 0.24260726710475455,
      "grad_norm": 0.0,
      "learning_rate": 3.787205257054503e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5021
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.09375,
      "epoch": 0.24265558562040973,
      "grad_norm": 0.0,
      "learning_rate": 3.786963664476227e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5022
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.59375,
      "epoch": 0.24270390413606494,
      "grad_norm": 0.0,
      "learning_rate": 3.786722071897951e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5023
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.53125,
      "epoch": 0.24275222265172014,
      "grad_norm": 0.0,
      "learning_rate": 3.786480479319675e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.19727617502212524,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5024
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.46875,
      "epoch": 0.24280054116737534,
      "grad_norm": 0.0,
      "learning_rate": 3.786238886741399e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5025
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.03125,
      "epoch": 0.24284885968303055,
      "grad_norm": 0.0,
      "learning_rate": 3.785997294163123e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5026
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.78125,
      "epoch": 0.24289717819868573,
      "grad_norm": 0.0,
      "learning_rate": 3.7857557015848475e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5027
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.59375,
      "epoch": 0.24294549671434093,
      "grad_norm": 13.939977269013957,
      "learning_rate": 3.785514109006571e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5028
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.40625,
      "epoch": 0.24299381522999614,
      "grad_norm": 0.0,
      "learning_rate": 3.785272516428295e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 5029
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.0625,
      "epoch": 0.24304213374565134,
      "grad_norm": 0.0,
      "learning_rate": 3.7850309238500194e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5030
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.4375,
      "epoch": 0.24309045226130654,
      "grad_norm": 0.0,
      "learning_rate": 3.7847893312717435e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5031
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.03125,
      "epoch": 0.24313877077696172,
      "grad_norm": 0.0,
      "learning_rate": 3.784547738693467e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5032
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.59375,
      "epoch": 0.24318708929261693,
      "grad_norm": 0.0,
      "learning_rate": 3.784306146115191e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5033
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.3125,
      "epoch": 0.24323540780827213,
      "grad_norm": 0.0,
      "learning_rate": 3.7840645535369154e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5034
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.3125,
      "epoch": 0.24328372632392734,
      "grad_norm": 0.0,
      "learning_rate": 3.783822960958639e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5035
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.90625,
      "epoch": 0.24333204483958254,
      "grad_norm": 0.0,
      "learning_rate": 3.783581368380363e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5036
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.84375,
      "epoch": 0.24338036335523772,
      "grad_norm": 0.0,
      "learning_rate": 3.7833397758020873e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5037
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.84375,
      "epoch": 0.24342868187089292,
      "grad_norm": 0.0,
      "learning_rate": 3.7830981832238114e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5038
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.59375,
      "epoch": 0.24347700038654813,
      "grad_norm": 0.0,
      "learning_rate": 3.782856590645535e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5039
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.5,
      "epoch": 0.24352531890220333,
      "grad_norm": 0.0,
      "learning_rate": 3.7826149980672597e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5040
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 105.5625,
      "epoch": 0.24357363741785854,
      "grad_norm": 7.781372650006614,
      "learning_rate": 3.7823734054889833e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 0.96875,
      "step": 5041
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.5,
      "epoch": 0.2436219559335137,
      "grad_norm": 0.0,
      "learning_rate": 3.782131812910707e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5042
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.15625,
      "epoch": 0.24367027444916892,
      "grad_norm": 0.0,
      "learning_rate": 3.781890220332431e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5043
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.65625,
      "epoch": 0.24371859296482412,
      "grad_norm": 0.0,
      "learning_rate": 3.781648627754155e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5044
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.28125,
      "epoch": 0.24376691148047933,
      "grad_norm": 0.0,
      "learning_rate": 3.7814070351758793e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5045
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.0,
      "epoch": 0.24381522999613453,
      "grad_norm": 0.0,
      "learning_rate": 3.781165442597603e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5046
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.21875,
      "epoch": 0.2438635485117897,
      "grad_norm": 0.0,
      "learning_rate": 3.7809238500193276e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5047
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.75,
      "epoch": 0.2439118670274449,
      "grad_norm": 0.0,
      "learning_rate": 3.780682257441051e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5048
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.78125,
      "epoch": 0.24396018554310012,
      "grad_norm": 0.0,
      "learning_rate": 3.7804406648627753e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5049
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.5625,
      "epoch": 0.24400850405875532,
      "grad_norm": 0.0,
      "learning_rate": 3.7801990722844995e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5050
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.21875,
      "epoch": 0.24405682257441053,
      "grad_norm": 0.0,
      "learning_rate": 3.7799574797062236e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5051
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.59375,
      "epoch": 0.2441051410900657,
      "grad_norm": 0.0,
      "learning_rate": 3.779715887127947e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5052
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.625,
      "epoch": 0.2441534596057209,
      "grad_norm": 9.733046675794297,
      "learning_rate": 3.779474294549671e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5053
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.625,
      "epoch": 0.2442017781213761,
      "grad_norm": 12.566311856907799,
      "learning_rate": 3.7792327019713955e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5054
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.03125,
      "epoch": 0.24425009663703132,
      "grad_norm": 0.0,
      "learning_rate": 3.778991109393119e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5055
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.34375,
      "epoch": 0.24429841515268652,
      "grad_norm": 0.0,
      "learning_rate": 3.778749516814843e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5056
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.84375,
      "epoch": 0.2443467336683417,
      "grad_norm": 0.0,
      "learning_rate": 3.7785079242365674e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5057
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.875,
      "epoch": 0.2443950521839969,
      "grad_norm": 0.0,
      "learning_rate": 3.7782663316582915e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5058
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.40625,
      "epoch": 0.2444433706996521,
      "grad_norm": 0.0,
      "learning_rate": 3.778024739080015e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5059
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 85.09375,
      "epoch": 0.2444916892153073,
      "grad_norm": 14.377686370275466,
      "learning_rate": 3.77778314650174e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5060
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 88.96875,
      "epoch": 0.24454000773096252,
      "grad_norm": 9.098429907291301,
      "learning_rate": 3.7775415539234634e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5061
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.03125,
      "epoch": 0.2445883262466177,
      "grad_norm": 0.0,
      "learning_rate": 3.777299961345187e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5062
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.59375,
      "epoch": 0.2446366447622729,
      "grad_norm": 0.0,
      "learning_rate": 3.777058368766911e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5063
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.65625,
      "epoch": 0.2446849632779281,
      "grad_norm": 0.0,
      "learning_rate": 3.776816776188635e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5064
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.125,
      "epoch": 0.2447332817935833,
      "grad_norm": 0.0,
      "learning_rate": 3.7765751836103594e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5065
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.78125,
      "epoch": 0.2447816003092385,
      "grad_norm": 0.0,
      "learning_rate": 3.776333591032083e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5066
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.59375,
      "epoch": 0.2448299188248937,
      "grad_norm": 0.0,
      "learning_rate": 3.7760919984538077e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5067
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.8125,
      "epoch": 0.2448782373405489,
      "grad_norm": 0.0,
      "learning_rate": 3.7758504058755313e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5068
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.5625,
      "epoch": 0.2449265558562041,
      "grad_norm": 0.0,
      "learning_rate": 3.7756088132972554e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5069
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.25,
      "epoch": 0.2449748743718593,
      "grad_norm": 0.0,
      "learning_rate": 3.7753672207189795e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5070
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.0625,
      "epoch": 0.2450231928875145,
      "grad_norm": 0.0,
      "learning_rate": 3.7751256281407037e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5071
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.46875,
      "epoch": 0.24507151140316968,
      "grad_norm": 0.0,
      "learning_rate": 3.7748840355624273e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5072
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.5,
      "epoch": 0.2451198299188249,
      "grad_norm": 0.0,
      "learning_rate": 3.774642442984151e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 5073
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.4375,
      "epoch": 0.2451681484344801,
      "grad_norm": 0.0,
      "learning_rate": 3.7744008504058756e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5074
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.28125,
      "epoch": 0.2452164669501353,
      "grad_norm": 0.0,
      "learning_rate": 3.774159257827599e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5075
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.375,
      "epoch": 0.2452647854657905,
      "grad_norm": 0.0,
      "learning_rate": 3.7739176652493233e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5076
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.78125,
      "epoch": 0.24531310398144568,
      "grad_norm": 697.3382584293217,
      "learning_rate": 3.7736760726710474e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5077
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.09375,
      "epoch": 0.24536142249710088,
      "grad_norm": 0.0,
      "learning_rate": 3.7734344800927716e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5078
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.5,
      "epoch": 0.2454097410127561,
      "grad_norm": 0.0,
      "learning_rate": 3.773192887514495e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5079
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.09375,
      "epoch": 0.2454580595284113,
      "grad_norm": 0.0,
      "learning_rate": 3.77295129493622e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5080
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.375,
      "epoch": 0.2455063780440665,
      "grad_norm": 0.0,
      "learning_rate": 3.7727097023579435e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5081
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.875,
      "epoch": 0.24555469655972167,
      "grad_norm": 0.0,
      "learning_rate": 3.7724681097796676e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5082
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.3125,
      "epoch": 0.24560301507537688,
      "grad_norm": 0.0,
      "learning_rate": 3.772226517201391e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5083
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.53125,
      "epoch": 0.24565133359103208,
      "grad_norm": 0.0,
      "learning_rate": 3.7719849246231153e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5084
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.8125,
      "epoch": 0.2456996521066873,
      "grad_norm": 0.0,
      "learning_rate": 3.7717433320448395e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5085
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.1875,
      "epoch": 0.2457479706223425,
      "grad_norm": 0.0,
      "learning_rate": 3.771501739466563e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 5086
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.4375,
      "epoch": 0.24579628913799767,
      "grad_norm": 0.0,
      "learning_rate": 3.771260146888288e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5087
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.34375,
      "epoch": 0.24584460765365287,
      "grad_norm": 0.0,
      "learning_rate": 3.7710185543100114e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.24251842498779297,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5088
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.46875,
      "epoch": 0.24589292616930808,
      "grad_norm": 0.0,
      "learning_rate": 3.7707769617317355e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5089
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.90625,
      "epoch": 0.24594124468496328,
      "grad_norm": 0.0,
      "learning_rate": 3.7705353691534596e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5090
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.0,
      "epoch": 0.2459895632006185,
      "grad_norm": 0.0,
      "learning_rate": 3.770293776575184e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5091
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.78125,
      "epoch": 0.24603788171627367,
      "grad_norm": 0.0,
      "learning_rate": 3.7700521839969074e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5092
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.84375,
      "epoch": 0.24608620023192887,
      "grad_norm": 0.0,
      "learning_rate": 3.769810591418631e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5093
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.53125,
      "epoch": 0.24613451874758407,
      "grad_norm": 0.0,
      "learning_rate": 3.7695689988403556e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5094
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.15625,
      "epoch": 0.24618283726323928,
      "grad_norm": 0.0,
      "learning_rate": 3.769327406262079e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5095
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.5,
      "epoch": 0.24623115577889448,
      "grad_norm": 0.0,
      "learning_rate": 3.7690858136838034e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5096
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.9375,
      "epoch": 0.24627947429454966,
      "grad_norm": 0.0,
      "learning_rate": 3.7688442211055275e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5097
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.34375,
      "epoch": 0.24632779281020487,
      "grad_norm": 0.0,
      "learning_rate": 3.7686026285272517e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5098
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.0,
      "epoch": 0.24637611132586007,
      "grad_norm": 0.0,
      "learning_rate": 3.7683610359489753e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5099
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.8125,
      "epoch": 0.24642442984151527,
      "grad_norm": 0.0,
      "learning_rate": 3.7681194433707e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5100
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.28125,
      "epoch": 0.24647274835717048,
      "grad_norm": 25.229426585405715,
      "learning_rate": 3.7678778507924235e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5101
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.53125,
      "epoch": 0.24652106687282566,
      "grad_norm": 0.0,
      "learning_rate": 3.7676362582141477e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5102
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.15625,
      "epoch": 0.24656938538848086,
      "grad_norm": 0.0,
      "learning_rate": 3.7673946656358713e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5103
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.03125,
      "epoch": 0.24661770390413607,
      "grad_norm": 0.0,
      "learning_rate": 3.7671530730575954e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5104
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.5625,
      "epoch": 0.24666602241979127,
      "grad_norm": 0.0,
      "learning_rate": 3.7669114804793196e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5105
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.15625,
      "epoch": 0.24671434093544647,
      "grad_norm": 0.0,
      "learning_rate": 3.766669887901043e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5106
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 85.3125,
      "epoch": 0.24676265945110165,
      "grad_norm": 29.79699640366703,
      "learning_rate": 3.766428295322768e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5107
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.6875,
      "epoch": 0.24681097796675686,
      "grad_norm": 0.0,
      "learning_rate": 3.7661867027444914e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5108
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.875,
      "epoch": 0.24685929648241206,
      "grad_norm": 0.0,
      "learning_rate": 3.7659451101662156e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5109
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.03125,
      "epoch": 0.24690761499806727,
      "grad_norm": 0.0,
      "learning_rate": 3.7657035175879397e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5110
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.21875,
      "epoch": 0.24695593351372247,
      "grad_norm": 0.0,
      "learning_rate": 3.765461925009664e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5111
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.46875,
      "epoch": 0.24700425202937765,
      "grad_norm": 0.0,
      "learning_rate": 3.7652203324313875e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5112
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 95.65625,
      "epoch": 0.24705257054503285,
      "grad_norm": 0.0,
      "learning_rate": 3.764978739853111e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 5113
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.0625,
      "epoch": 0.24710088906068806,
      "grad_norm": 0.0,
      "learning_rate": 3.7647371472748357e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5114
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.0,
      "epoch": 0.24714920757634326,
      "grad_norm": 0.0,
      "learning_rate": 3.7644955546965593e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5115
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.875,
      "epoch": 0.24719752609199847,
      "grad_norm": 0.0,
      "learning_rate": 3.7642539621182835e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5116
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.65625,
      "epoch": 0.24724584460765364,
      "grad_norm": 0.0,
      "learning_rate": 3.7640123695400076e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5117
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.0625,
      "epoch": 0.24729416312330885,
      "grad_norm": 0.0,
      "learning_rate": 3.763770776961732e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5118
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.375,
      "epoch": 0.24734248163896405,
      "grad_norm": 0.0,
      "learning_rate": 3.7635291843834553e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5119
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.75,
      "epoch": 0.24739080015461926,
      "grad_norm": 0.0,
      "learning_rate": 3.76328759180518e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5120
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.875,
      "epoch": 0.24743911867027446,
      "grad_norm": 0.0,
      "learning_rate": 3.7630459992269036e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5121
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.53125,
      "epoch": 0.24748743718592964,
      "grad_norm": 0.0,
      "learning_rate": 3.762804406648628e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5122
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.5,
      "epoch": 0.24753575570158484,
      "grad_norm": 0.0,
      "learning_rate": 3.7625628140703514e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 5123
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.28125,
      "epoch": 0.24758407421724005,
      "grad_norm": 0.0,
      "learning_rate": 3.7623212214920755e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5124
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.25,
      "epoch": 0.24763239273289525,
      "grad_norm": 0.0,
      "learning_rate": 3.7620796289137996e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5125
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 86.21875,
      "epoch": 0.24768071124855046,
      "grad_norm": 5.590573268008451,
      "learning_rate": 3.761838036335523e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5126
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.90625,
      "epoch": 0.24772902976420563,
      "grad_norm": 0.0,
      "learning_rate": 3.761596443757248e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5127
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.0625,
      "epoch": 0.24777734827986084,
      "grad_norm": 0.0,
      "learning_rate": 3.7613548511789715e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5128
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 94.78125,
      "epoch": 0.24782566679551604,
      "grad_norm": 14.226280487196815,
      "learning_rate": 3.7611132586006956e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.20633062720298767,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 5129
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.46875,
      "epoch": 0.24787398531117125,
      "grad_norm": 0.0,
      "learning_rate": 3.76087166602242e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5130
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.03125,
      "epoch": 0.24792230382682645,
      "grad_norm": 0.0,
      "learning_rate": 3.760630073444144e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5131
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.90625,
      "epoch": 0.24797062234248163,
      "grad_norm": 0.0,
      "learning_rate": 3.7603884808658675e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5132
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.40625,
      "epoch": 0.24801894085813683,
      "grad_norm": 0.0,
      "learning_rate": 3.7601468882875917e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5133
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.5,
      "epoch": 0.24806725937379204,
      "grad_norm": 0.0,
      "learning_rate": 3.759905295709316e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 5134
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.59375,
      "epoch": 0.24811557788944724,
      "grad_norm": 0.0,
      "learning_rate": 3.7596637031310394e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5135
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.125,
      "epoch": 0.24816389640510245,
      "grad_norm": 0.0,
      "learning_rate": 3.7594221105527635e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5136
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.90625,
      "epoch": 0.24821221492075762,
      "grad_norm": 0.0,
      "learning_rate": 3.7591805179744877e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5137
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.15625,
      "epoch": 0.24826053343641283,
      "grad_norm": 0.0,
      "learning_rate": 3.758938925396212e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5138
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.90625,
      "epoch": 0.24830885195206803,
      "grad_norm": 0.0,
      "learning_rate": 3.7586973328179354e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 5139
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.65625,
      "epoch": 0.24835717046772324,
      "grad_norm": 0.0,
      "learning_rate": 3.75845574023966e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5140
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.375,
      "epoch": 0.24840548898337844,
      "grad_norm": 0.0,
      "learning_rate": 3.7582141476613837e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5141
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.875,
      "epoch": 0.24845380749903362,
      "grad_norm": 57.73204703101295,
      "learning_rate": 3.757972555083108e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5142
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.0,
      "epoch": 0.24850212601468882,
      "grad_norm": 0.0,
      "learning_rate": 3.7577309625048314e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5143
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.4375,
      "epoch": 0.24855044453034403,
      "grad_norm": 0.0,
      "learning_rate": 3.757489369926556e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5144
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.21875,
      "epoch": 0.24859876304599923,
      "grad_norm": 0.0,
      "learning_rate": 3.7572477773482797e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5145
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.375,
      "epoch": 0.24864708156165444,
      "grad_norm": 0.0,
      "learning_rate": 3.7570061847700033e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5146
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.375,
      "epoch": 0.24869540007730961,
      "grad_norm": 0.0,
      "learning_rate": 3.756764592191728e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5147
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.28125,
      "epoch": 0.24874371859296482,
      "grad_norm": 0.0,
      "learning_rate": 3.7565229996134516e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5148
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.25,
      "epoch": 0.24879203710862002,
      "grad_norm": 0.0,
      "learning_rate": 3.7562814070351757e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5149
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.1875,
      "epoch": 0.24884035562427523,
      "grad_norm": 0.0,
      "learning_rate": 3.7560398144569e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5150
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.6875,
      "epoch": 0.24888867413993043,
      "grad_norm": 0.0,
      "learning_rate": 3.755798221878624e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5151
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.65625,
      "epoch": 0.2489369926555856,
      "grad_norm": 0.0,
      "learning_rate": 3.7555566293003476e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5152
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.1875,
      "epoch": 0.24898531117124081,
      "grad_norm": 0.0,
      "learning_rate": 3.755315036722072e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5153
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.9375,
      "epoch": 0.24903362968689602,
      "grad_norm": 0.0,
      "learning_rate": 3.755073444143796e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5154
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.5,
      "epoch": 0.24908194820255122,
      "grad_norm": 0.0,
      "learning_rate": 3.7548318515655195e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5155
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.6875,
      "epoch": 0.24913026671820643,
      "grad_norm": 0.0,
      "learning_rate": 3.7545902589872436e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5156
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.53125,
      "epoch": 0.2491785852338616,
      "grad_norm": 0.0,
      "learning_rate": 3.754348666408968e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5157
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.1875,
      "epoch": 0.2492269037495168,
      "grad_norm": 0.0,
      "learning_rate": 3.754107073830692e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5158
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.21875,
      "epoch": 0.24927522226517201,
      "grad_norm": 0.0,
      "learning_rate": 3.7538654812524155e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5159
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.59375,
      "epoch": 0.24932354078082722,
      "grad_norm": 0.0,
      "learning_rate": 3.75362388867414e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5160
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.09375,
      "epoch": 0.24937185929648242,
      "grad_norm": 0.0,
      "learning_rate": 3.753382296095864e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 5161
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.5625,
      "epoch": 0.2494201778121376,
      "grad_norm": 0.0,
      "learning_rate": 3.753140703517588e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5162
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.3125,
      "epoch": 0.2494684963277928,
      "grad_norm": 0.0,
      "learning_rate": 3.7528991109393115e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5163
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 99.34375,
      "epoch": 0.249516814843448,
      "grad_norm": 15.702519898302532,
      "learning_rate": 3.752657518361036e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 5164
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.84375,
      "epoch": 0.24956513335910321,
      "grad_norm": 0.0,
      "learning_rate": 3.75241592578276e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5165
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.59375,
      "epoch": 0.24961345187475842,
      "grad_norm": 0.0,
      "learning_rate": 3.7521743332044834e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5166
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.71875,
      "epoch": 0.2496617703904136,
      "grad_norm": 0.0,
      "learning_rate": 3.751932740626208e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5167
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.4375,
      "epoch": 0.2497100889060688,
      "grad_norm": 0.0,
      "learning_rate": 3.7516911480479317e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5168
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.09375,
      "epoch": 0.249758407421724,
      "grad_norm": 0.0,
      "learning_rate": 3.751449555469656e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5169
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.90625,
      "epoch": 0.2498067259373792,
      "grad_norm": 0.0,
      "learning_rate": 3.75120796289138e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5170
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.90625,
      "epoch": 0.24985504445303441,
      "grad_norm": 0.0,
      "learning_rate": 3.750966370313104e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5171
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.59375,
      "epoch": 0.2499033629686896,
      "grad_norm": 0.0,
      "learning_rate": 3.7507247777348277e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5172
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.25,
      "epoch": 0.2499516814843448,
      "grad_norm": 0.0,
      "learning_rate": 3.750483185156552e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5173
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.59375,
      "epoch": 0.25,
      "grad_norm": 0.0,
      "learning_rate": 3.750241592578276e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5174
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.0,
      "epoch": 0.2500483185156552,
      "grad_norm": 0.0,
      "learning_rate": 3.75e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5175
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.78125,
      "epoch": 0.2500966370313104,
      "grad_norm": 0.0,
      "learning_rate": 3.7497584074217237e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5176
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.65625,
      "epoch": 0.2501449555469656,
      "grad_norm": 0.0,
      "learning_rate": 3.749516814843448e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5177
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.90625,
      "epoch": 0.2501932740626208,
      "grad_norm": 0.0,
      "learning_rate": 3.749275222265172e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5178
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 101.40625,
      "epoch": 0.250241592578276,
      "grad_norm": 0.0,
      "learning_rate": 3.7490336296868956e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5179
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.375,
      "epoch": 0.25028991109393117,
      "grad_norm": 0.0,
      "learning_rate": 3.7487920371086197e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5180
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.3125,
      "epoch": 0.2503382296095864,
      "grad_norm": 0.0,
      "learning_rate": 3.748550444530344e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5181
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.21875,
      "epoch": 0.2503865481252416,
      "grad_norm": 0.0,
      "learning_rate": 3.748308851952068e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5182
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.9375,
      "epoch": 0.2504348666408968,
      "grad_norm": 0.0,
      "learning_rate": 3.7480672593737916e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 5183
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.25,
      "epoch": 0.250483185156552,
      "grad_norm": 0.0,
      "learning_rate": 3.747825666795516e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5184
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.96875,
      "epoch": 0.25053150367220717,
      "grad_norm": 0.0,
      "learning_rate": 3.74758407421724e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5185
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.46875,
      "epoch": 0.2505798221878624,
      "grad_norm": 0.0,
      "learning_rate": 3.7473424816389635e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5186
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.4375,
      "epoch": 0.2506281407035176,
      "grad_norm": 0.0,
      "learning_rate": 3.747100889060688e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5187
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.71875,
      "epoch": 0.2506764592191728,
      "grad_norm": 0.0,
      "learning_rate": 3.746859296482412e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5188
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.25,
      "epoch": 0.250724777734828,
      "grad_norm": 0.0,
      "learning_rate": 3.746617703904136e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2130674123764038,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5189
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.25,
      "epoch": 0.25077309625048316,
      "grad_norm": 0.0,
      "learning_rate": 3.7463761113258595e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5190
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.40625,
      "epoch": 0.2508214147661384,
      "grad_norm": 0.0,
      "learning_rate": 3.746134518747584e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5191
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.0,
      "epoch": 0.25086973328179357,
      "grad_norm": 0.0,
      "learning_rate": 3.745892926169308e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5192
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.6875,
      "epoch": 0.2509180517974488,
      "grad_norm": 0.0,
      "learning_rate": 3.745651333591032e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5193
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.75,
      "epoch": 0.250966370313104,
      "grad_norm": 0.0,
      "learning_rate": 3.745409741012756e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5194
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.4375,
      "epoch": 0.25101468882875916,
      "grad_norm": 0.0,
      "learning_rate": 3.74516814843448e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5195
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 93.0,
      "epoch": 0.2510630073444144,
      "grad_norm": 10.973720509179202,
      "learning_rate": 3.744926555856204e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5196
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.28125,
      "epoch": 0.25111132586006957,
      "grad_norm": 0.0,
      "learning_rate": 3.744684963277928e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5197
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.40625,
      "epoch": 0.2511596443757248,
      "grad_norm": 0.0,
      "learning_rate": 3.744443370699652e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5198
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.65625,
      "epoch": 0.25120796289138,
      "grad_norm": 0.0,
      "learning_rate": 3.7442017781213757e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5199
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.5625,
      "epoch": 0.25125628140703515,
      "grad_norm": 0.0,
      "learning_rate": 3.7439601855431e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5200
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.0625,
      "epoch": 0.2513045999226904,
      "grad_norm": 0.0,
      "learning_rate": 3.743718592964824e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5201
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.34375,
      "epoch": 0.25135291843834556,
      "grad_norm": 0.0,
      "learning_rate": 3.743477000386548e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5202
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.46875,
      "epoch": 0.2514012369540008,
      "grad_norm": 0.0,
      "learning_rate": 3.7432354078082717e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5203
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.6875,
      "epoch": 0.25144955546965597,
      "grad_norm": 0.0,
      "learning_rate": 3.7429938152299963e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5204
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.8125,
      "epoch": 0.25149787398531115,
      "grad_norm": 0.0,
      "learning_rate": 3.74275222265172e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5205
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.0,
      "epoch": 0.2515461925009664,
      "grad_norm": 0.0,
      "learning_rate": 3.7425106300734436e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5206
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.90625,
      "epoch": 0.25159451101662156,
      "grad_norm": 0.0,
      "learning_rate": 3.742269037495168e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5207
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.25,
      "epoch": 0.2516428295322768,
      "grad_norm": 0.0,
      "learning_rate": 3.742027444916892e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5208
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.3125,
      "epoch": 0.25169114804793197,
      "grad_norm": 0.0,
      "learning_rate": 3.741785852338616e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5209
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.75,
      "epoch": 0.25173946656358714,
      "grad_norm": 0.0,
      "learning_rate": 3.7415442597603396e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5210
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 95.625,
      "epoch": 0.2517877850792424,
      "grad_norm": 9.336257117630488,
      "learning_rate": 3.741302667182064e-07,
      "loss": 0.0,
      "reward": 1.40625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5211
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.875,
      "epoch": 0.25183610359489755,
      "grad_norm": 0.0,
      "learning_rate": 3.741061074603788e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5212
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.25,
      "epoch": 0.2518844221105528,
      "grad_norm": 0.0,
      "learning_rate": 3.740819482025512e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5213
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.375,
      "epoch": 0.25193274062620796,
      "grad_norm": 0.0,
      "learning_rate": 3.740577889447236e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 5214
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.84375,
      "epoch": 0.25198105914186314,
      "grad_norm": 0.0,
      "learning_rate": 3.74033629686896e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5215
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.25,
      "epoch": 0.25202937765751837,
      "grad_norm": 0.0,
      "learning_rate": 3.740094704290684e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5216
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.59375,
      "epoch": 0.25207769617317355,
      "grad_norm": 0.0,
      "learning_rate": 3.739853111712408e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5217
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.375,
      "epoch": 0.2521260146888288,
      "grad_norm": 0.0,
      "learning_rate": 3.739611519134132e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5218
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.0625,
      "epoch": 0.25217433320448396,
      "grad_norm": 0.0,
      "learning_rate": 3.739369926555856e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5219
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.40625,
      "epoch": 0.25222265172013913,
      "grad_norm": 0.0,
      "learning_rate": 3.73912833397758e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5220
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.375,
      "epoch": 0.25227097023579437,
      "grad_norm": 0.0,
      "learning_rate": 3.738886741399304e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5221
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.53125,
      "epoch": 0.25231928875144954,
      "grad_norm": 0.0,
      "learning_rate": 3.738645148821028e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5222
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.71875,
      "epoch": 0.2523676072671048,
      "grad_norm": 0.0,
      "learning_rate": 3.738403556242752e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5223
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.0,
      "epoch": 0.25241592578275995,
      "grad_norm": 0.0,
      "learning_rate": 3.7381619636644764e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5224
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.71875,
      "epoch": 0.25246424429841513,
      "grad_norm": 0.0,
      "learning_rate": 3.7379203710862e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5225
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.46875,
      "epoch": 0.25251256281407036,
      "grad_norm": 0.0,
      "learning_rate": 3.737678778507924e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5226
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.96875,
      "epoch": 0.25256088132972554,
      "grad_norm": 0.0,
      "learning_rate": 3.7374371859296483e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5227
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.6875,
      "epoch": 0.25260919984538077,
      "grad_norm": 0.0,
      "learning_rate": 3.737195593351372e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5228
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.15625,
      "epoch": 0.25265751836103595,
      "grad_norm": 0.0,
      "learning_rate": 3.736954000773096e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5229
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.21875,
      "epoch": 0.2527058368766911,
      "grad_norm": 0.0,
      "learning_rate": 3.7367124081948196e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5230
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.0,
      "epoch": 0.25275415539234636,
      "grad_norm": 0.0,
      "learning_rate": 3.7364708156165443e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5231
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.84375,
      "epoch": 0.25280247390800153,
      "grad_norm": 0.0,
      "learning_rate": 3.736229223038268e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5232
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.875,
      "epoch": 0.25285079242365677,
      "grad_norm": 0.0,
      "learning_rate": 3.735987630459992e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5233
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.6875,
      "epoch": 0.25289911093931194,
      "grad_norm": 0.0,
      "learning_rate": 3.735746037881716e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5234
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.65625,
      "epoch": 0.2529474294549671,
      "grad_norm": 14.886223378816883,
      "learning_rate": 3.7355044453034403e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5235
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.46875,
      "epoch": 0.25299574797062235,
      "grad_norm": 0.0,
      "learning_rate": 3.735262852725164e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5236
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.1875,
      "epoch": 0.25304406648627753,
      "grad_norm": 0.0,
      "learning_rate": 3.7350212601468886e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5237
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.65625,
      "epoch": 0.25309238500193276,
      "grad_norm": 0.0,
      "learning_rate": 3.734779667568612e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5238
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 79.1875,
      "epoch": 0.25314070351758794,
      "grad_norm": 9.446438169572543,
      "learning_rate": 3.734538074990336e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5239
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.5625,
      "epoch": 0.2531890220332431,
      "grad_norm": 0.0,
      "learning_rate": 3.73429648241206e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5240
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.25,
      "epoch": 0.25323734054889835,
      "grad_norm": 0.0,
      "learning_rate": 3.734054889833784e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5241
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.3125,
      "epoch": 0.2532856590645535,
      "grad_norm": 5.487824978689983,
      "learning_rate": 3.733813297255508e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5242
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.9375,
      "epoch": 0.25333397758020876,
      "grad_norm": 0.0,
      "learning_rate": 3.733571704677232e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 5243
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.65625,
      "epoch": 0.25338229609586393,
      "grad_norm": 0.0,
      "learning_rate": 3.7333301120989565e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5244
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.75,
      "epoch": 0.2534306146115191,
      "grad_norm": 0.0,
      "learning_rate": 3.73308851952068e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 5245
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.96875,
      "epoch": 0.25347893312717434,
      "grad_norm": 0.0,
      "learning_rate": 3.732846926942404e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 5246
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.40625,
      "epoch": 0.2535272516428295,
      "grad_norm": 0.0,
      "learning_rate": 3.7326053343641284e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5247
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.96875,
      "epoch": 0.25357557015848475,
      "grad_norm": 0.0,
      "learning_rate": 3.732363741785852e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5248
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.90625,
      "epoch": 0.25362388867413993,
      "grad_norm": 0.0,
      "learning_rate": 3.732122149207576e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5249
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.8125,
      "epoch": 0.2536722071897951,
      "grad_norm": 0.0,
      "learning_rate": 3.7318805566292997e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 5250
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.1875,
      "epoch": 0.25372052570545034,
      "grad_norm": 0.0,
      "learning_rate": 3.7316389640510244e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5251
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.5625,
      "epoch": 0.2537688442211055,
      "grad_norm": 0.0,
      "learning_rate": 3.731397371472748e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5252
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.875,
      "epoch": 0.25381716273676075,
      "grad_norm": 0.0,
      "learning_rate": 3.731155778894472e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5253
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.3125,
      "epoch": 0.2538654812524159,
      "grad_norm": 0.0,
      "learning_rate": 3.7309141863161963e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5254
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.75,
      "epoch": 0.2539137997680711,
      "grad_norm": 0.0,
      "learning_rate": 3.7306725937379204e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5255
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.8125,
      "epoch": 0.25396211828372633,
      "grad_norm": 0.0,
      "learning_rate": 3.730431001159644e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5256
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.71875,
      "epoch": 0.2540104367993815,
      "grad_norm": 0.0,
      "learning_rate": 3.7301894085813687e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5257
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.03125,
      "epoch": 0.25405875531503674,
      "grad_norm": 0.0,
      "learning_rate": 3.7299478160030923e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5258
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.3125,
      "epoch": 0.2541070738306919,
      "grad_norm": 0.0,
      "learning_rate": 3.729706223424816e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5259
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.25,
      "epoch": 0.2541553923463471,
      "grad_norm": 0.0,
      "learning_rate": 3.72946463084654e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5260
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.0,
      "epoch": 0.25420371086200233,
      "grad_norm": 9.24098357618512,
      "learning_rate": 3.729223038268264e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5261
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.46875,
      "epoch": 0.2542520293776575,
      "grad_norm": 0.0,
      "learning_rate": 3.7289814456899883e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5262
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.1875,
      "epoch": 0.25430034789331274,
      "grad_norm": 0.0,
      "learning_rate": 3.728739853111712e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5263
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 85.21875,
      "epoch": 0.2543486664089679,
      "grad_norm": 21.88918502061233,
      "learning_rate": 3.7284982605334366e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5264
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.40625,
      "epoch": 0.2543969849246231,
      "grad_norm": 0.0,
      "learning_rate": 3.72825666795516e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5265
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.15625,
      "epoch": 0.2544453034402783,
      "grad_norm": 0.0,
      "learning_rate": 3.7280150753768843e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5266
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.375,
      "epoch": 0.2544936219559335,
      "grad_norm": 0.0,
      "learning_rate": 3.7277734827986085e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5267
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.6875,
      "epoch": 0.25454194047158873,
      "grad_norm": 0.0,
      "learning_rate": 3.727531890220332e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5268
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.71875,
      "epoch": 0.2545902589872439,
      "grad_norm": 0.0,
      "learning_rate": 3.727290297642056e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.19727617502212524,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5269
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.28125,
      "epoch": 0.2546385775028991,
      "grad_norm": 0.0,
      "learning_rate": 3.72704870506378e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5270
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 93.6875,
      "epoch": 0.2546868960185543,
      "grad_norm": 10.268068937407095,
      "learning_rate": 3.7268071124855045e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5271
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.5625,
      "epoch": 0.2547352145342095,
      "grad_norm": 0.0,
      "learning_rate": 3.726565519907228e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5272
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.34375,
      "epoch": 0.25478353304986473,
      "grad_norm": 0.0,
      "learning_rate": 3.726323927328952e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5273
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.0,
      "epoch": 0.2548318515655199,
      "grad_norm": 0.0,
      "learning_rate": 3.7260823347506764e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5274
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.4375,
      "epoch": 0.2548801700811751,
      "grad_norm": 0.0,
      "learning_rate": 3.7258407421724005e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5275
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.21875,
      "epoch": 0.2549284885968303,
      "grad_norm": 0.0,
      "learning_rate": 3.725599149594124e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5276
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.71875,
      "epoch": 0.2549768071124855,
      "grad_norm": 0.0,
      "learning_rate": 3.725357557015849e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5277
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.8125,
      "epoch": 0.2550251256281407,
      "grad_norm": 0.0,
      "learning_rate": 3.7251159644375724e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 5278
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 104.5,
      "epoch": 0.2550734441437959,
      "grad_norm": 0.0,
      "learning_rate": 3.724874371859296e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 5279
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.5625,
      "epoch": 0.2551217626594511,
      "grad_norm": 0.0,
      "learning_rate": 3.72463277928102e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 5280
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 106.75,
      "epoch": 0.2551700811751063,
      "grad_norm": 0.0,
      "learning_rate": 3.724391186702744e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 5281
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.75,
      "epoch": 0.2552183996907615,
      "grad_norm": 0.0,
      "learning_rate": 3.7241495941244684e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5282
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.8125,
      "epoch": 0.2552667182064167,
      "grad_norm": 0.0,
      "learning_rate": 3.723908001546192e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5283
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.21875,
      "epoch": 0.2553150367220719,
      "grad_norm": 0.0,
      "learning_rate": 3.7236664089679167e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5284
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.90625,
      "epoch": 0.2553633552377271,
      "grad_norm": 0.0,
      "learning_rate": 3.72342481638964e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5285
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 120.21875,
      "epoch": 0.2554116737533823,
      "grad_norm": 0.0,
      "learning_rate": 3.7231832238113644e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5286
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.40625,
      "epoch": 0.2554599922690375,
      "grad_norm": 0.0,
      "learning_rate": 3.7229416312330885e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5287
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 113.1875,
      "epoch": 0.2555083107846927,
      "grad_norm": 51.90108726158192,
      "learning_rate": 3.7227000386548127e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.31751489639282227,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5288
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.8125,
      "epoch": 0.2555566293003479,
      "grad_norm": 0.0,
      "learning_rate": 3.7224584460765363e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5289
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.5,
      "epoch": 0.25560494781600307,
      "grad_norm": 0.0,
      "learning_rate": 3.72221685349826e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5290
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.1875,
      "epoch": 0.2556532663316583,
      "grad_norm": 0.0,
      "learning_rate": 3.7219752609199846e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5291
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.59375,
      "epoch": 0.2557015848473135,
      "grad_norm": 0.0,
      "learning_rate": 3.721733668341708e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5292
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.15625,
      "epoch": 0.2557499033629687,
      "grad_norm": 0.0,
      "learning_rate": 3.7214920757634323e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5293
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.59375,
      "epoch": 0.2557982218786239,
      "grad_norm": 0.0,
      "learning_rate": 3.7212504831851564e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5294
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 109.65625,
      "epoch": 0.25584654039427907,
      "grad_norm": 6.94235034068538,
      "learning_rate": 3.7210088906068806e-07,
      "loss": -0.0,
      "reward": 2.28125,
      "reward_std": 0.32195520401000977,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5295
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.21875,
      "epoch": 0.2558948589099343,
      "grad_norm": 0.0,
      "learning_rate": 3.720767298028604e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5296
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.3125,
      "epoch": 0.2559431774255895,
      "grad_norm": 0.0,
      "learning_rate": 3.720525705450329e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5297
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.28125,
      "epoch": 0.2559914959412447,
      "grad_norm": 0.0,
      "learning_rate": 3.7202841128720524e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5298
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.125,
      "epoch": 0.2560398144568999,
      "grad_norm": 0.0,
      "learning_rate": 3.720042520293776e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5299
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.34375,
      "epoch": 0.25608813297255506,
      "grad_norm": 0.0,
      "learning_rate": 3.7198009277155e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5300
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.125,
      "epoch": 0.2561364514882103,
      "grad_norm": 0.0,
      "learning_rate": 3.7195593351372243e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5301
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.4375,
      "epoch": 0.25618477000386547,
      "grad_norm": 0.0,
      "learning_rate": 3.7193177425589485e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5302
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.5,
      "epoch": 0.2562330885195207,
      "grad_norm": 0.0,
      "learning_rate": 3.719076149980672e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5303
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.5625,
      "epoch": 0.2562814070351759,
      "grad_norm": 0.0,
      "learning_rate": 3.7188345574023967e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5304
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.125,
      "epoch": 0.25632972555083106,
      "grad_norm": 0.0,
      "learning_rate": 3.7185929648241203e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5305
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.875,
      "epoch": 0.2563780440664863,
      "grad_norm": 0.0,
      "learning_rate": 3.7183513722458445e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5306
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.0,
      "epoch": 0.25642636258214146,
      "grad_norm": 0.0,
      "learning_rate": 3.7181097796675686e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5307
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.34375,
      "epoch": 0.2564746810977967,
      "grad_norm": 0.0,
      "learning_rate": 3.717868187089293e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5308
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.96875,
      "epoch": 0.2565229996134519,
      "grad_norm": 0.0,
      "learning_rate": 3.7176265945110164e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5309
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.03125,
      "epoch": 0.25657131812910705,
      "grad_norm": 0.0,
      "learning_rate": 3.71738500193274e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5310
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.78125,
      "epoch": 0.2566196366447623,
      "grad_norm": 0.0,
      "learning_rate": 3.7171434093544646e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5311
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.15625,
      "epoch": 0.25666795516041746,
      "grad_norm": 0.0,
      "learning_rate": 3.716901816776188e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5312
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 114.875,
      "epoch": 0.2567162736760727,
      "grad_norm": 0.0,
      "learning_rate": 3.7166602241979124e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 5313
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.46875,
      "epoch": 0.25676459219172787,
      "grad_norm": 0.0,
      "learning_rate": 3.7164186316196365e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 5314
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.65625,
      "epoch": 0.25681291070738305,
      "grad_norm": 0.0,
      "learning_rate": 3.7161770390413606e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5315
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.375,
      "epoch": 0.2568612292230383,
      "grad_norm": 0.0,
      "learning_rate": 3.715935446463084e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5316
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.84375,
      "epoch": 0.25690954773869346,
      "grad_norm": 0.0,
      "learning_rate": 3.715693853884809e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5317
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.6875,
      "epoch": 0.2569578662543487,
      "grad_norm": 0.0,
      "learning_rate": 3.7154522613065325e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5318
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 126.4375,
      "epoch": 0.25700618477000386,
      "grad_norm": 0.0,
      "learning_rate": 3.715210668728256e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5319
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.375,
      "epoch": 0.25705450328565904,
      "grad_norm": 0.0,
      "learning_rate": 3.7149690761499803e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5320
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.5625,
      "epoch": 0.2571028218013143,
      "grad_norm": 0.0,
      "learning_rate": 3.7147274835717044e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5321
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 125.15625,
      "epoch": 0.25715114031696945,
      "grad_norm": 24.174781042288036,
      "learning_rate": 3.7144858909934285e-07,
      "loss": -0.0,
      "reward": 1.578125,
      "reward_std": 0.37992191314697266,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 5322
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.09375,
      "epoch": 0.2571994588326247,
      "grad_norm": 0.0,
      "learning_rate": 3.714244298415152e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5323
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.3125,
      "epoch": 0.25724777734827986,
      "grad_norm": 0.0,
      "learning_rate": 3.714002705836877e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5324
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.34375,
      "epoch": 0.25729609586393504,
      "grad_norm": 0.0,
      "learning_rate": 3.7137611132586004e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5325
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.3125,
      "epoch": 0.25734441437959027,
      "grad_norm": 0.0,
      "learning_rate": 3.7135195206803246e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5326
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.5,
      "completion_length": 118.4375,
      "epoch": 0.25739273289524545,
      "grad_norm": 39.62030980536969,
      "learning_rate": 3.7132779281020487e-07,
      "loss": -0.0,
      "reward": 1.203125,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.15625,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 5327
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.5,
      "epoch": 0.2574410514109007,
      "grad_norm": 0.0,
      "learning_rate": 3.713036335523773e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5328
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 103.6875,
      "epoch": 0.25748936992655586,
      "grad_norm": 0.0,
      "learning_rate": 3.7127947429454964e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5329
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.75,
      "epoch": 0.25753768844221103,
      "grad_norm": 0.0,
      "learning_rate": 3.71255315036722e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 5330
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 87.96875,
      "epoch": 0.25758600695786626,
      "grad_norm": 9.303278016525356,
      "learning_rate": 3.7123115577889447e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.2993341088294983,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5331
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.3125,
      "epoch": 0.25763432547352144,
      "grad_norm": 0.0,
      "learning_rate": 3.7120699652106683e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5332
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.375,
      "epoch": 0.2576826439891767,
      "grad_norm": 0.0,
      "learning_rate": 3.7118283726323925e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5333
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.40625,
      "epoch": 0.25773096250483185,
      "grad_norm": 0.0,
      "learning_rate": 3.7115867800541166e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5334
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.9375,
      "epoch": 0.257779281020487,
      "grad_norm": 0.0,
      "learning_rate": 3.7113451874758407e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5335
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.3125,
      "epoch": 0.25782759953614226,
      "grad_norm": 0.0,
      "learning_rate": 3.7111035948975643e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5336
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 98.96875,
      "epoch": 0.25787591805179744,
      "grad_norm": 0.0,
      "learning_rate": 3.710862002319289e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 5337
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.96875,
      "epoch": 0.25792423656745267,
      "grad_norm": 0.0,
      "learning_rate": 3.7106204097410126e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5338
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.6875,
      "epoch": 0.25797255508310785,
      "grad_norm": 0.0,
      "learning_rate": 3.710378817162737e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5339
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 105.90625,
      "epoch": 0.258020873598763,
      "grad_norm": 32.634868531564145,
      "learning_rate": 3.7101372245844603e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5340
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.875,
      "epoch": 0.25806919211441826,
      "grad_norm": 0.0,
      "learning_rate": 3.7098956320061845e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5341
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.4375,
      "epoch": 0.25811751063007343,
      "grad_norm": 0.0,
      "learning_rate": 3.7096540394279086e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5342
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.40625,
      "epoch": 0.25816582914572866,
      "grad_norm": 0.0,
      "learning_rate": 3.709412446849632e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5343
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.1875,
      "epoch": 0.25821414766138384,
      "grad_norm": 0.0,
      "learning_rate": 3.709170854271357e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5344
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 102.34375,
      "epoch": 0.258262466177039,
      "grad_norm": 0.0,
      "learning_rate": 3.7089292616930805e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 5345
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 101.375,
      "epoch": 0.25831078469269425,
      "grad_norm": 0.0,
      "learning_rate": 3.7086876691148046e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5346
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.875,
      "epoch": 0.2583591032083494,
      "grad_norm": 0.0,
      "learning_rate": 3.708446076536529e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5347
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.71875,
      "epoch": 0.25840742172400466,
      "grad_norm": 0.0,
      "learning_rate": 3.708204483958253e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5348
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.40625,
      "epoch": 0.25845574023965984,
      "grad_norm": 0.0,
      "learning_rate": 3.7079628913799765e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5349
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.625,
      "epoch": 0.258504058755315,
      "grad_norm": 0.0,
      "learning_rate": 3.7077212988017e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5350
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.9375,
      "epoch": 0.25855237727097025,
      "grad_norm": 0.0,
      "learning_rate": 3.707479706223425e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5351
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.3125,
      "epoch": 0.2586006957866254,
      "grad_norm": 0.0,
      "learning_rate": 3.7072381136451484e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5352
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 99.34375,
      "epoch": 0.25864901430228066,
      "grad_norm": 9.48019742715268,
      "learning_rate": 3.7069965210668725e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5353
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.1875,
      "epoch": 0.25869733281793583,
      "grad_norm": 0.0,
      "learning_rate": 3.7067549284885967e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5354
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.6875,
      "epoch": 0.258745651333591,
      "grad_norm": 0.0,
      "learning_rate": 3.706513335910321e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5355
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.15625,
      "epoch": 0.25879396984924624,
      "grad_norm": 0.0,
      "learning_rate": 3.7062717433320444e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5356
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.8125,
      "epoch": 0.2588422883649014,
      "grad_norm": 0.0,
      "learning_rate": 3.706030150753769e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5357
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 97.21875,
      "epoch": 0.25889060688055665,
      "grad_norm": 0.0,
      "learning_rate": 3.7057885581754927e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5358
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.46875,
      "epoch": 0.2589389253962118,
      "grad_norm": 0.0,
      "learning_rate": 3.705546965597217e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5359
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 87.65625,
      "epoch": 0.258987243911867,
      "grad_norm": 32.27519024792103,
      "learning_rate": 3.7053053730189404e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.3693191707134247,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5360
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.5,
      "epoch": 0.25903556242752224,
      "grad_norm": 0.0,
      "learning_rate": 3.7050637804406646e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5361
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.8125,
      "epoch": 0.2590838809431774,
      "grad_norm": 0.0,
      "learning_rate": 3.7048221878623887e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5362
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.9375,
      "epoch": 0.25913219945883265,
      "grad_norm": 0.0,
      "learning_rate": 3.7045805952841123e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5363
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.03125,
      "epoch": 0.2591805179744878,
      "grad_norm": 0.0,
      "learning_rate": 3.704339002705837e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5364
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.5,
      "epoch": 0.259228836490143,
      "grad_norm": 0.0,
      "learning_rate": 3.7040974101275606e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 5365
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.0625,
      "epoch": 0.25927715500579823,
      "grad_norm": 0.0,
      "learning_rate": 3.7038558175492847e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5366
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.21875,
      "epoch": 0.2593254735214534,
      "grad_norm": 0.0,
      "learning_rate": 3.703614224971009e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5367
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 98.28125,
      "epoch": 0.25937379203710864,
      "grad_norm": 37.252999592970546,
      "learning_rate": 3.703372632392733e-07,
      "loss": -0.0,
      "reward": 1.4375,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5368
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.125,
      "epoch": 0.2594221105527638,
      "grad_norm": 0.0,
      "learning_rate": 3.7031310398144566e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5369
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.8125,
      "epoch": 0.259470429068419,
      "grad_norm": 0.0,
      "learning_rate": 3.70288944723618e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5370
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.21875,
      "epoch": 0.2595187475840742,
      "grad_norm": 0.0,
      "learning_rate": 3.702647854657905e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5371
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.78125,
      "epoch": 0.2595670660997294,
      "grad_norm": 0.0,
      "learning_rate": 3.7024062620796285e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5372
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.09375,
      "epoch": 0.25961538461538464,
      "grad_norm": 0.0,
      "learning_rate": 3.7021646695013526e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5373
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 86.4375,
      "epoch": 0.2596637031310398,
      "grad_norm": 13.110690956842664,
      "learning_rate": 3.701923076923077e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.37402740120887756,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5374
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.125,
      "epoch": 0.259712021646695,
      "grad_norm": 0.0,
      "learning_rate": 3.701681484344801e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5375
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.3125,
      "epoch": 0.2597603401623502,
      "grad_norm": 0.0,
      "learning_rate": 3.7014398917665245e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5376
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.34375,
      "epoch": 0.2598086586780054,
      "grad_norm": 0.0,
      "learning_rate": 3.701198299188249e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5377
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.1875,
      "epoch": 0.25985697719366063,
      "grad_norm": 0.0,
      "learning_rate": 3.700956706609973e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5378
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.03125,
      "epoch": 0.2599052957093158,
      "grad_norm": 0.0,
      "learning_rate": 3.700715114031697e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5379
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.84375,
      "epoch": 0.259953614224971,
      "grad_norm": 0.0,
      "learning_rate": 3.7004735214534205e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5380
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.9375,
      "epoch": 0.2600019327406262,
      "grad_norm": 0.0,
      "learning_rate": 3.7002319288751446e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5381
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 86.96875,
      "epoch": 0.2600502512562814,
      "grad_norm": 19.06762742883854,
      "learning_rate": 3.699990336296869e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5382
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.3125,
      "epoch": 0.2600985697719366,
      "grad_norm": 0.0,
      "learning_rate": 3.6997487437185924e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5383
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.90625,
      "epoch": 0.2601468882875918,
      "grad_norm": 0.0,
      "learning_rate": 3.699507151140317e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5384
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.3125,
      "epoch": 0.260195206803247,
      "grad_norm": 0.0,
      "learning_rate": 3.6992655585620407e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5385
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.40625,
      "epoch": 0.2602435253189022,
      "grad_norm": 0.0,
      "learning_rate": 3.699023965983765e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5386
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.65625,
      "epoch": 0.2602918438345574,
      "grad_norm": 0.0,
      "learning_rate": 3.698782373405489e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5387
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.4375,
      "epoch": 0.2603401623502126,
      "grad_norm": 0.0,
      "learning_rate": 3.698540780827213e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5388
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.4375,
      "epoch": 0.2603884808658678,
      "grad_norm": 0.0,
      "learning_rate": 3.6982991882489367e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5389
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.9375,
      "epoch": 0.260436799381523,
      "grad_norm": 0.0,
      "learning_rate": 3.698057595670661e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 5390
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.1875,
      "epoch": 0.2604851178971782,
      "grad_norm": 0.0,
      "learning_rate": 3.697816003092385e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5391
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.8125,
      "epoch": 0.2605334364128334,
      "grad_norm": 0.0,
      "learning_rate": 3.6975744105141086e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5392
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.34375,
      "epoch": 0.2605817549284886,
      "grad_norm": 0.0,
      "learning_rate": 3.6973328179358327e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5393
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.84375,
      "epoch": 0.2606300734441438,
      "grad_norm": 0.0,
      "learning_rate": 3.697091225357557e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5394
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.34375,
      "epoch": 0.26067839195979897,
      "grad_norm": 0.0,
      "learning_rate": 3.696849632779281e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5395
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.1875,
      "epoch": 0.2607267104754542,
      "grad_norm": 0.0,
      "learning_rate": 3.6966080402010046e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5396
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.5,
      "epoch": 0.2607750289911094,
      "grad_norm": 0.0,
      "learning_rate": 3.696366447622729e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5397
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.40625,
      "epoch": 0.2608233475067646,
      "grad_norm": 0.0,
      "learning_rate": 3.696124855044453e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5398
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.84375,
      "epoch": 0.2608716660224198,
      "grad_norm": 0.0,
      "learning_rate": 3.695883262466177e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5399
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.5,
      "epoch": 0.26091998453807497,
      "grad_norm": 0.0,
      "learning_rate": 3.6956416698879006e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5400
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.1875,
      "epoch": 0.2609683030537302,
      "grad_norm": 0.0,
      "learning_rate": 3.695400077309625e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5401
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.9375,
      "epoch": 0.2610166215693854,
      "grad_norm": 0.0,
      "learning_rate": 3.695158484731349e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5402
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.40625,
      "epoch": 0.2610649400850406,
      "grad_norm": 20.51828897944532,
      "learning_rate": 3.6949168921530725e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5403
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.28125,
      "epoch": 0.2611132586006958,
      "grad_norm": 0.0,
      "learning_rate": 3.694675299574797e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5404
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.84375,
      "epoch": 0.26116157711635096,
      "grad_norm": 0.0,
      "learning_rate": 3.6944337069965207e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5405
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.15625,
      "epoch": 0.2612098956320062,
      "grad_norm": 0.0,
      "learning_rate": 3.694192114418245e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5406
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.90625,
      "epoch": 0.26125821414766137,
      "grad_norm": 11.09876011478955,
      "learning_rate": 3.693950521839969e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.31983357667922974,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5407
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.0625,
      "epoch": 0.2613065326633166,
      "grad_norm": 0.0,
      "learning_rate": 3.693708929261693e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5408
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.15625,
      "epoch": 0.2613548511789718,
      "grad_norm": 0.0,
      "learning_rate": 3.693467336683417e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5409
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.125,
      "epoch": 0.26140316969462696,
      "grad_norm": 0.0,
      "learning_rate": 3.693225744105141e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5410
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.75,
      "epoch": 0.2614514882102822,
      "grad_norm": 0.0,
      "learning_rate": 3.692984151526865e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5411
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.34375,
      "epoch": 0.26149980672593737,
      "grad_norm": 0.0,
      "learning_rate": 3.6927425589485886e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5412
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.59375,
      "epoch": 0.2615481252415926,
      "grad_norm": 0.0,
      "learning_rate": 3.692500966370313e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5413
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.09375,
      "epoch": 0.2615964437572478,
      "grad_norm": 0.0,
      "learning_rate": 3.692259373792037e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5414
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.46875,
      "epoch": 0.26164476227290295,
      "grad_norm": 0.0,
      "learning_rate": 3.692017781213761e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5415
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.3125,
      "epoch": 0.2616930807885582,
      "grad_norm": 0.0,
      "learning_rate": 3.6917761886354846e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5416
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.28125,
      "epoch": 0.26174139930421336,
      "grad_norm": 0.0,
      "learning_rate": 3.6915345960572093e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5417
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.90625,
      "epoch": 0.2617897178198686,
      "grad_norm": 0.0,
      "learning_rate": 3.691293003478933e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5418
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 92.125,
      "epoch": 0.26183803633552377,
      "grad_norm": 38.8679950854955,
      "learning_rate": 3.691051410900657e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.2505248188972473,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5419
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.6875,
      "epoch": 0.26188635485117895,
      "grad_norm": 0.0,
      "learning_rate": 3.6908098183223807e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5420
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 93.1875,
      "epoch": 0.2619346733668342,
      "grad_norm": 0.0,
      "learning_rate": 3.6905682257441053e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 5421
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.625,
      "epoch": 0.26198299188248936,
      "grad_norm": 0.0,
      "learning_rate": 3.690326633165829e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 5422
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.59375,
      "epoch": 0.2620313103981446,
      "grad_norm": 0.0,
      "learning_rate": 3.6900850405875525e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 5423
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.09375,
      "epoch": 0.26207962891379977,
      "grad_norm": 0.0,
      "learning_rate": 3.689843448009277e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5424
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.4375,
      "epoch": 0.26212794742945494,
      "grad_norm": 0.0,
      "learning_rate": 3.689601855431001e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5425
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 94.28125,
      "epoch": 0.2621762659451102,
      "grad_norm": 46.837045603776154,
      "learning_rate": 3.689360262852725e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5426
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.09375,
      "epoch": 0.26222458446076535,
      "grad_norm": 0.0,
      "learning_rate": 3.689118670274449e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5427
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.0,
      "epoch": 0.2622729029764206,
      "grad_norm": 0.0,
      "learning_rate": 3.688877077696173e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5428
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.34375,
      "epoch": 0.26232122149207576,
      "grad_norm": 0.0,
      "learning_rate": 3.688635485117897e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5429
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.15625,
      "epoch": 0.26236954000773094,
      "grad_norm": 0.0,
      "learning_rate": 3.688393892539621e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5430
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.65625,
      "epoch": 0.26241785852338617,
      "grad_norm": 0.0,
      "learning_rate": 3.688152299961345e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5431
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.75,
      "epoch": 0.26246617703904135,
      "grad_norm": 0.0,
      "learning_rate": 3.687910707383069e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5432
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.65625,
      "epoch": 0.2625144955546966,
      "grad_norm": 0.0,
      "learning_rate": 3.687669114804793e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5433
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.0,
      "epoch": 0.26256281407035176,
      "grad_norm": 0.0,
      "learning_rate": 3.687427522226517e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5434
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.4375,
      "epoch": 0.26261113258600693,
      "grad_norm": 0.0,
      "learning_rate": 3.687185929648241e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5435
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.0625,
      "epoch": 0.26265945110166217,
      "grad_norm": 0.0,
      "learning_rate": 3.6869443370699647e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5436
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.25,
      "epoch": 0.26270776961731734,
      "grad_norm": 0.0,
      "learning_rate": 3.6867027444916894e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5437
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.28125,
      "epoch": 0.2627560881329726,
      "grad_norm": 0.0,
      "learning_rate": 3.686461151913413e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5438
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.84375,
      "epoch": 0.26280440664862775,
      "grad_norm": 0.0,
      "learning_rate": 3.686219559335137e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5439
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.6875,
      "epoch": 0.26285272516428293,
      "grad_norm": 0.0,
      "learning_rate": 3.685977966756861e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5440
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.4375,
      "epoch": 0.26290104367993816,
      "grad_norm": 0.0,
      "learning_rate": 3.6857363741785854e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5441
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.4375,
      "epoch": 0.26294936219559334,
      "grad_norm": 0.0,
      "learning_rate": 3.685494781600309e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5442
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.40625,
      "epoch": 0.26299768071124857,
      "grad_norm": 0.0,
      "learning_rate": 3.6852531890220326e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5443
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.53125,
      "epoch": 0.26304599922690375,
      "grad_norm": 0.0,
      "learning_rate": 3.6850115964437573e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5444
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.8125,
      "epoch": 0.2630943177425589,
      "grad_norm": 0.0,
      "learning_rate": 3.684770003865481e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5445
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.125,
      "epoch": 0.26314263625821416,
      "grad_norm": 0.0,
      "learning_rate": 3.684528411287205e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5446
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.34375,
      "epoch": 0.26319095477386933,
      "grad_norm": 0.0,
      "learning_rate": 3.684286818708929e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5447
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.875,
      "epoch": 0.26323927328952457,
      "grad_norm": 0.0,
      "learning_rate": 3.6840452261306533e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5448
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.6875,
      "epoch": 0.26328759180517974,
      "grad_norm": 0.0,
      "learning_rate": 3.683803633552377e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5449
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 93.90625,
      "epoch": 0.2633359103208349,
      "grad_norm": 46.84080055713453,
      "learning_rate": 3.683562040974101e-07,
      "loss": 0.0,
      "reward": 1.5625,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5450
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.96875,
      "epoch": 0.26338422883649015,
      "grad_norm": 0.0,
      "learning_rate": 3.683320448395825e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5451
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.8125,
      "epoch": 0.26343254735214533,
      "grad_norm": 0.0,
      "learning_rate": 3.6830788558175493e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5452
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.90625,
      "epoch": 0.26348086586780056,
      "grad_norm": 0.0,
      "learning_rate": 3.682837263239273e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 5453
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.6875,
      "epoch": 0.26352918438345574,
      "grad_norm": 0.0,
      "learning_rate": 3.682595670660997e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5454
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.21875,
      "epoch": 0.2635775028991109,
      "grad_norm": 0.0,
      "learning_rate": 3.682354078082721e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5455
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.8125,
      "epoch": 0.26362582141476615,
      "grad_norm": 0.0,
      "learning_rate": 3.682112485504445e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5456
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.90625,
      "epoch": 0.2636741399304213,
      "grad_norm": 0.0,
      "learning_rate": 3.6818708929261695e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5457
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.375,
      "epoch": 0.26372245844607656,
      "grad_norm": 0.0,
      "learning_rate": 3.681629300347893e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5458
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.53125,
      "epoch": 0.26377077696173173,
      "grad_norm": 0.0,
      "learning_rate": 3.681387707769617e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5459
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.375,
      "epoch": 0.2638190954773869,
      "grad_norm": 0.0,
      "learning_rate": 3.681146115191341e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5460
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.65625,
      "epoch": 0.26386741399304214,
      "grad_norm": 0.0,
      "learning_rate": 3.6809045226130655e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5461
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.75,
      "epoch": 0.2639157325086973,
      "grad_norm": 0.0,
      "learning_rate": 3.680662930034789e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5462
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.9375,
      "epoch": 0.26396405102435255,
      "grad_norm": 0.0,
      "learning_rate": 3.6804213374565127e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5463
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.09375,
      "epoch": 0.26401236954000773,
      "grad_norm": 0.0,
      "learning_rate": 3.6801797448782374e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5464
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.625,
      "epoch": 0.2640606880556629,
      "grad_norm": 0.0,
      "learning_rate": 3.679938152299961e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5465
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.875,
      "epoch": 0.26410900657131814,
      "grad_norm": 0.0,
      "learning_rate": 3.679696559721685e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5466
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.0625,
      "epoch": 0.2641573250869733,
      "grad_norm": 0.0,
      "learning_rate": 3.679454967143409e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5467
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.03125,
      "epoch": 0.26420564360262855,
      "grad_norm": 0.0,
      "learning_rate": 3.6792133745651334e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5468
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.1875,
      "epoch": 0.2642539621182837,
      "grad_norm": 0.0,
      "learning_rate": 3.678971781986857e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5469
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.5,
      "epoch": 0.2643022806339389,
      "grad_norm": 0.0,
      "learning_rate": 3.678730189408581e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5470
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.125,
      "epoch": 0.26435059914959413,
      "grad_norm": 0.0,
      "learning_rate": 3.678488596830305e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5471
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.03125,
      "epoch": 0.2643989176652493,
      "grad_norm": 0.0,
      "learning_rate": 3.6782470042520294e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5472
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.59375,
      "epoch": 0.26444723618090454,
      "grad_norm": 0.0,
      "learning_rate": 3.678005411673753e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5473
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.09375,
      "epoch": 0.2644955546965597,
      "grad_norm": 10.469679923726972,
      "learning_rate": 3.677763819095477e-07,
      "loss": -0.0,
      "reward": 2.296875,
      "reward_std": 0.3098084628582001,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5474
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.34375,
      "epoch": 0.2645438732122149,
      "grad_norm": 0.0,
      "learning_rate": 3.6775222265172013e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5475
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.0,
      "epoch": 0.26459219172787013,
      "grad_norm": 0.0,
      "learning_rate": 3.677280633938925e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5476
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.96875,
      "epoch": 0.2646405102435253,
      "grad_norm": 0.0,
      "learning_rate": 3.6770390413606495e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5477
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.46875,
      "epoch": 0.26468882875918054,
      "grad_norm": 0.0,
      "learning_rate": 3.676797448782373e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5478
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.84375,
      "epoch": 0.2647371472748357,
      "grad_norm": 0.0,
      "learning_rate": 3.6765558562040973e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5479
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 97.75,
      "epoch": 0.2647854657904909,
      "grad_norm": 0.0,
      "learning_rate": 3.676314263625821e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5480
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.125,
      "epoch": 0.2648337843061461,
      "grad_norm": 0.0,
      "learning_rate": 3.6760726710475456e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5481
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.28125,
      "epoch": 0.2648821028218013,
      "grad_norm": 0.0,
      "learning_rate": 3.675831078469269e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 5482
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.71875,
      "epoch": 0.26493042133745653,
      "grad_norm": 0.0,
      "learning_rate": 3.6755894858909933e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5483
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.25,
      "epoch": 0.2649787398531117,
      "grad_norm": 0.0,
      "learning_rate": 3.6753478933127174e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5484
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.78125,
      "epoch": 0.2650270583687669,
      "grad_norm": 0.0,
      "learning_rate": 3.675106300734441e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5485
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.3125,
      "epoch": 0.2650753768844221,
      "grad_norm": 0.0,
      "learning_rate": 3.674864708156165e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5486
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.71875,
      "epoch": 0.2651236954000773,
      "grad_norm": 0.0,
      "learning_rate": 3.6746231155778893e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5487
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.15625,
      "epoch": 0.26517201391573253,
      "grad_norm": 0.0,
      "learning_rate": 3.6743815229996135e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5488
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.75,
      "epoch": 0.2652203324313877,
      "grad_norm": 0.0,
      "learning_rate": 3.674139930421337e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5489
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.5,
      "epoch": 0.2652686509470429,
      "grad_norm": 0.0,
      "learning_rate": 3.673898337843061e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5490
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 86.15625,
      "epoch": 0.2653169694626981,
      "grad_norm": 14.742801310369934,
      "learning_rate": 3.6736567452647853e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.2961388826370239,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5491
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 94.0,
      "epoch": 0.2653652879783533,
      "grad_norm": 0.0,
      "learning_rate": 3.6734151526865095e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 5492
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.65625,
      "epoch": 0.2654136064940085,
      "grad_norm": 0.0,
      "learning_rate": 3.673173560108233e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5493
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.6875,
      "epoch": 0.2654619250096637,
      "grad_norm": 0.0,
      "learning_rate": 3.672931967529958e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5494
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.40625,
      "epoch": 0.2655102435253189,
      "grad_norm": 0.0,
      "learning_rate": 3.6726903749516814e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5495
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.6875,
      "epoch": 0.2655585620409741,
      "grad_norm": 0.0,
      "learning_rate": 3.672448782373405e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5496
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.59375,
      "epoch": 0.2656068805566293,
      "grad_norm": 0.0,
      "learning_rate": 3.6722071897951296e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5497
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.4375,
      "epoch": 0.2656551990722845,
      "grad_norm": 0.0,
      "learning_rate": 3.671965597216853e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5498
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.1875,
      "epoch": 0.2657035175879397,
      "grad_norm": 0.0,
      "learning_rate": 3.6717240046385774e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5499
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.53125,
      "epoch": 0.2657518361035949,
      "grad_norm": 0.0,
      "learning_rate": 3.671482412060301e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5500
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.625,
      "epoch": 0.2658001546192501,
      "grad_norm": 0.0,
      "learning_rate": 3.6712408194820256e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5501
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.90625,
      "epoch": 0.2658484731349053,
      "grad_norm": 0.0,
      "learning_rate": 3.670999226903749e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5502
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.28125,
      "epoch": 0.2658967916505605,
      "grad_norm": 0.0,
      "learning_rate": 3.6707576343254734e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5503
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.25,
      "epoch": 0.2659451101662157,
      "grad_norm": 0.0,
      "learning_rate": 3.6705160417471975e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5504
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.46875,
      "epoch": 0.26599342868187087,
      "grad_norm": 0.0,
      "learning_rate": 3.670274449168921e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5505
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.9375,
      "epoch": 0.2660417471975261,
      "grad_norm": 0.0,
      "learning_rate": 3.670032856590645e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5506
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.78125,
      "epoch": 0.2660900657131813,
      "grad_norm": 0.0,
      "learning_rate": 3.6697912640123694e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5507
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.4375,
      "epoch": 0.2661383842288365,
      "grad_norm": 0.0,
      "learning_rate": 3.6695496714340935e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5508
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 90.75,
      "epoch": 0.2661867027444917,
      "grad_norm": 32.96002311718961,
      "learning_rate": 3.669308078855817e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5509
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.125,
      "epoch": 0.26623502126014686,
      "grad_norm": 0.0,
      "learning_rate": 3.6690664862775413e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5510
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.75,
      "epoch": 0.2662833397758021,
      "grad_norm": 0.0,
      "learning_rate": 3.6688248936992654e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5511
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.65625,
      "epoch": 0.2663316582914573,
      "grad_norm": 0.0,
      "learning_rate": 3.6685833011209896e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5512
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.71875,
      "epoch": 0.2663799768071125,
      "grad_norm": 0.0,
      "learning_rate": 3.668341708542713e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5513
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.125,
      "epoch": 0.2664282953227677,
      "grad_norm": 0.0,
      "learning_rate": 3.668100115964438e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5514
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.25,
      "epoch": 0.26647661383842286,
      "grad_norm": 0.0,
      "learning_rate": 3.6678585233861614e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5515
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.75,
      "epoch": 0.2665249323540781,
      "grad_norm": 0.0,
      "learning_rate": 3.667616930807885e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5516
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.5,
      "epoch": 0.26657325086973327,
      "grad_norm": 0.0,
      "learning_rate": 3.6673753382296097e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5517
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.4375,
      "epoch": 0.2666215693853885,
      "grad_norm": 0.0,
      "learning_rate": 3.6671337456513333e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5518
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.34375,
      "epoch": 0.2666698879010437,
      "grad_norm": 0.0,
      "learning_rate": 3.6668921530730574e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5519
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.25,
      "epoch": 0.26671820641669886,
      "grad_norm": 0.0,
      "learning_rate": 3.666650560494781e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5520
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.4375,
      "epoch": 0.2667665249323541,
      "grad_norm": 0.0,
      "learning_rate": 3.6664089679165057e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5521
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.21875,
      "epoch": 0.26681484344800926,
      "grad_norm": 0.0,
      "learning_rate": 3.6661673753382293e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 5522
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.90625,
      "epoch": 0.2668631619636645,
      "grad_norm": 0.0,
      "learning_rate": 3.6659257827599535e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5523
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.25,
      "epoch": 0.2669114804793197,
      "grad_norm": 0.0,
      "learning_rate": 3.6656841901816776e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5524
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.875,
      "epoch": 0.26695979899497485,
      "grad_norm": 0.0,
      "learning_rate": 3.665442597603401e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5525
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.59375,
      "epoch": 0.2670081175106301,
      "grad_norm": 0.0,
      "learning_rate": 3.6652010050251253e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5526
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 93.5625,
      "epoch": 0.26705643602628526,
      "grad_norm": 7.993043368394014,
      "learning_rate": 3.6649594124468495e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.22049658000469208,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5527
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.78125,
      "epoch": 0.2671047545419405,
      "grad_norm": 0.0,
      "learning_rate": 3.6647178198685736e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5528
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.59375,
      "epoch": 0.26715307305759567,
      "grad_norm": 0.0,
      "learning_rate": 3.664476227290297e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5529
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.5,
      "epoch": 0.26720139157325085,
      "grad_norm": 0.0,
      "learning_rate": 3.6642346347120214e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5530
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.3125,
      "epoch": 0.2672497100889061,
      "grad_norm": 0.0,
      "learning_rate": 3.6639930421337455e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5531
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.78125,
      "epoch": 0.26729802860456126,
      "grad_norm": 0.0,
      "learning_rate": 3.6637514495554696e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5532
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.25,
      "epoch": 0.2673463471202165,
      "grad_norm": 0.0,
      "learning_rate": 3.663509856977193e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5533
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.1875,
      "epoch": 0.26739466563587166,
      "grad_norm": 0.0,
      "learning_rate": 3.663268264398918e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 5534
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.4375,
      "epoch": 0.26744298415152684,
      "grad_norm": 0.0,
      "learning_rate": 3.6630266718206415e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 5535
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.875,
      "epoch": 0.2674913026671821,
      "grad_norm": 0.0,
      "learning_rate": 3.662785079242365e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5536
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.625,
      "epoch": 0.26753962118283725,
      "grad_norm": 0.0,
      "learning_rate": 3.66254348666409e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 5537
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.46875,
      "epoch": 0.2675879396984925,
      "grad_norm": 0.0,
      "learning_rate": 3.6623018940858134e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5538
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.125,
      "epoch": 0.26763625821414766,
      "grad_norm": 0.0,
      "learning_rate": 3.6620603015075375e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5539
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.875,
      "epoch": 0.26768457672980284,
      "grad_norm": 0.0,
      "learning_rate": 3.661818708929261e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5540
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.28125,
      "epoch": 0.26773289524545807,
      "grad_norm": 0.0,
      "learning_rate": 3.661577116350986e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5541
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.59375,
      "epoch": 0.26778121376111325,
      "grad_norm": 0.0,
      "learning_rate": 3.6613355237727094e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5542
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.84375,
      "epoch": 0.2678295322767685,
      "grad_norm": 0.0,
      "learning_rate": 3.6610939311944335e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5543
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.5,
      "epoch": 0.26787785079242366,
      "grad_norm": 0.0,
      "learning_rate": 3.6608523386161577e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5544
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.71875,
      "epoch": 0.26792616930807883,
      "grad_norm": 0.0,
      "learning_rate": 3.660610746037882e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5545
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.65625,
      "epoch": 0.26797448782373406,
      "grad_norm": 0.0,
      "learning_rate": 3.6603691534596054e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5546
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.15625,
      "epoch": 0.26802280633938924,
      "grad_norm": 0.0,
      "learning_rate": 3.6601275608813296e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5547
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.40625,
      "epoch": 0.2680711248550445,
      "grad_norm": 0.0,
      "learning_rate": 3.6598859683030537e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5548
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.28125,
      "epoch": 0.26811944337069965,
      "grad_norm": 0.0,
      "learning_rate": 3.6596443757247773e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5549
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.625,
      "epoch": 0.2681677618863548,
      "grad_norm": 0.0,
      "learning_rate": 3.6594027831465014e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 5550
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.9375,
      "epoch": 0.26821608040201006,
      "grad_norm": 0.0,
      "learning_rate": 3.6591611905682256e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5551
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.21875,
      "epoch": 0.26826439891766524,
      "grad_norm": 0.0,
      "learning_rate": 3.6589195979899497e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5552
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.34375,
      "epoch": 0.26831271743332047,
      "grad_norm": 0.0,
      "learning_rate": 3.6586780054116733e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5553
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.90625,
      "epoch": 0.26836103594897565,
      "grad_norm": 0.0,
      "learning_rate": 3.658436412833398e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5554
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.71875,
      "epoch": 0.2684093544646308,
      "grad_norm": 0.0,
      "learning_rate": 3.6581948202551216e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5555
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.5,
      "epoch": 0.26845767298028606,
      "grad_norm": 0.0,
      "learning_rate": 3.657953227676845e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5556
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.96875,
      "epoch": 0.26850599149594123,
      "grad_norm": 0.0,
      "learning_rate": 3.65771163509857e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5557
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.3125,
      "epoch": 0.26855431001159646,
      "grad_norm": 0.0,
      "learning_rate": 3.6574700425202935e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5558
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.1875,
      "epoch": 0.26860262852725164,
      "grad_norm": 0.0,
      "learning_rate": 3.6572284499420176e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5559
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.21875,
      "epoch": 0.2686509470429068,
      "grad_norm": 0.0,
      "learning_rate": 3.656986857363741e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5560
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.96875,
      "epoch": 0.26869926555856205,
      "grad_norm": 0.0,
      "learning_rate": 3.656745264785466e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5561
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.875,
      "epoch": 0.2687475840742172,
      "grad_norm": 0.0,
      "learning_rate": 3.6565036722071895e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5562
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.875,
      "epoch": 0.26879590258987246,
      "grad_norm": 0.0,
      "learning_rate": 3.6562620796289136e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 5563
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.96875,
      "epoch": 0.26884422110552764,
      "grad_norm": 0.0,
      "learning_rate": 3.656020487050638e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5564
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.71875,
      "epoch": 0.2688925396211828,
      "grad_norm": 0.0,
      "learning_rate": 3.655778894472362e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5565
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.96875,
      "epoch": 0.26894085813683805,
      "grad_norm": 0.0,
      "learning_rate": 3.6555373018940855e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5566
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 99.1875,
      "epoch": 0.2689891766524932,
      "grad_norm": 12.986205596573907,
      "learning_rate": 3.6552957093158096e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.30537301301956177,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5567
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.875,
      "epoch": 0.26903749516814845,
      "grad_norm": 0.0,
      "learning_rate": 3.655054116737534e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5568
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.21875,
      "epoch": 0.26908581368380363,
      "grad_norm": 0.0,
      "learning_rate": 3.6548125241592574e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5569
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.375,
      "epoch": 0.2691341321994588,
      "grad_norm": 0.0,
      "learning_rate": 3.6545709315809815e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5570
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.625,
      "epoch": 0.26918245071511404,
      "grad_norm": 0.0,
      "learning_rate": 3.6543293390027057e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5571
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.59375,
      "epoch": 0.2692307692307692,
      "grad_norm": 0.0,
      "learning_rate": 3.65408774642443e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5572
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.21875,
      "epoch": 0.26927908774642445,
      "grad_norm": 0.0,
      "learning_rate": 3.6538461538461534e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5573
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.28125,
      "epoch": 0.2693274062620796,
      "grad_norm": 0.0,
      "learning_rate": 3.653604561267878e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 5574
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.3125,
      "epoch": 0.2693757247777348,
      "grad_norm": 0.0,
      "learning_rate": 3.6533629686896017e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5575
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 97.96875,
      "epoch": 0.26942404329339004,
      "grad_norm": 0.0,
      "learning_rate": 3.6531213761113253e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5576
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.0625,
      "epoch": 0.2694723618090452,
      "grad_norm": 0.0,
      "learning_rate": 3.65287978353305e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5577
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.59375,
      "epoch": 0.26952068032470045,
      "grad_norm": 0.0,
      "learning_rate": 3.6526381909547735e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5578
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 102.03125,
      "epoch": 0.2695689988403556,
      "grad_norm": 83.05111091826188,
      "learning_rate": 3.6523965983764977e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5579
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.46875,
      "epoch": 0.2696173173560108,
      "grad_norm": 0.0,
      "learning_rate": 3.6521550057982213e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5580
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.3125,
      "epoch": 0.26966563587166603,
      "grad_norm": 0.0,
      "learning_rate": 3.651913413219946e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5581
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.28125,
      "epoch": 0.2697139543873212,
      "grad_norm": 0.0,
      "learning_rate": 3.6516718206416696e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5582
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 98.75,
      "epoch": 0.26976227290297644,
      "grad_norm": 0.0,
      "learning_rate": 3.6514302280633937e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 5583
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 92.71875,
      "epoch": 0.2698105914186316,
      "grad_norm": 13.365027220404269,
      "learning_rate": 3.651188635485118e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.23685522377490997,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5584
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.9375,
      "epoch": 0.2698589099342868,
      "grad_norm": 0.0,
      "learning_rate": 3.650947042906842e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5585
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.59375,
      "epoch": 0.269907228449942,
      "grad_norm": 0.0,
      "learning_rate": 3.6507054503285656e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5586
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 100.21875,
      "epoch": 0.2699555469655972,
      "grad_norm": 0.0,
      "learning_rate": 3.6504638577502897e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 5587
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 104.6875,
      "epoch": 0.27000386548125244,
      "grad_norm": 0.0,
      "learning_rate": 3.650222265172014e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 5588
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.34375,
      "epoch": 0.2700521839969076,
      "grad_norm": 0.0,
      "learning_rate": 3.6499806725937375e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5589
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.0625,
      "epoch": 0.2701005025125628,
      "grad_norm": 0.0,
      "learning_rate": 3.6497390800154616e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5590
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.34375,
      "epoch": 0.270148821028218,
      "grad_norm": 0.0,
      "learning_rate": 3.6494974874371857e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5591
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.90625,
      "epoch": 0.2701971395438732,
      "grad_norm": 0.0,
      "learning_rate": 3.64925589485891e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5592
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.6875,
      "epoch": 0.27024545805952843,
      "grad_norm": 0.0,
      "learning_rate": 3.6490143022806335e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5593
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.21875,
      "epoch": 0.2702937765751836,
      "grad_norm": 0.0,
      "learning_rate": 3.648772709702358e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5594
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 98.6875,
      "epoch": 0.2703420950908388,
      "grad_norm": 0.0,
      "learning_rate": 3.648531117124082e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5595
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.78125,
      "epoch": 0.270390413606494,
      "grad_norm": 0.0,
      "learning_rate": 3.648289524545806e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5596
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.0,
      "epoch": 0.2704387321221492,
      "grad_norm": 0.0,
      "learning_rate": 3.64804793196753e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5597
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.53125,
      "epoch": 0.2704870506378044,
      "grad_norm": 9.535304476044892,
      "learning_rate": 3.6478063393892536e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5598
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.59375,
      "epoch": 0.2705353691534596,
      "grad_norm": 0.0,
      "learning_rate": 3.647564746810978e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5599
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 94.25,
      "epoch": 0.2705836876691148,
      "grad_norm": 27.57958182895391,
      "learning_rate": 3.6473231542327014e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5600
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.4375,
      "epoch": 0.27063200618477,
      "grad_norm": 0.0,
      "learning_rate": 3.647081561654426e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5601
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.84375,
      "epoch": 0.2706803247004252,
      "grad_norm": 0.0,
      "learning_rate": 3.6468399690761496e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5602
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.3125,
      "epoch": 0.2707286432160804,
      "grad_norm": 0.0,
      "learning_rate": 3.646598376497874e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5603
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.40625,
      "epoch": 0.2707769617317356,
      "grad_norm": 0.0,
      "learning_rate": 3.646356783919598e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5604
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.71875,
      "epoch": 0.2708252802473908,
      "grad_norm": 0.0,
      "learning_rate": 3.646115191341322e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5605
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.375,
      "epoch": 0.270873598763046,
      "grad_norm": 0.0,
      "learning_rate": 3.6458735987630457e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5606
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.75,
      "epoch": 0.2709219172787012,
      "grad_norm": 0.0,
      "learning_rate": 3.6456320061847703e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5607
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.09375,
      "epoch": 0.2709702357943564,
      "grad_norm": 0.0,
      "learning_rate": 3.645390413606494e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5608
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 89.46875,
      "epoch": 0.2710185543100116,
      "grad_norm": 15.30962840157365,
      "learning_rate": 3.6451488210282175e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5609
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.0,
      "epoch": 0.27106687282566677,
      "grad_norm": 0.0,
      "learning_rate": 3.6449072284499417e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5610
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.46875,
      "epoch": 0.271115191341322,
      "grad_norm": 0.0,
      "learning_rate": 3.644665635871666e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5611
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.6875,
      "epoch": 0.2711635098569772,
      "grad_norm": 0.0,
      "learning_rate": 3.64442404329339e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5612
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.6875,
      "epoch": 0.2712118283726324,
      "grad_norm": 0.0,
      "learning_rate": 3.6441824507151136e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5613
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.28125,
      "epoch": 0.2712601468882876,
      "grad_norm": 0.0,
      "learning_rate": 3.643940858136838e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5614
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.71875,
      "epoch": 0.27130846540394277,
      "grad_norm": 0.0,
      "learning_rate": 3.643699265558562e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5615
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.21875,
      "epoch": 0.271356783919598,
      "grad_norm": 0.0,
      "learning_rate": 3.643457672980286e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5616
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.46875,
      "epoch": 0.2714051024352532,
      "grad_norm": 0.0,
      "learning_rate": 3.64321608040201e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5617
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.625,
      "epoch": 0.2714534209509084,
      "grad_norm": 0.0,
      "learning_rate": 3.6429744878237337e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5618
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 96.21875,
      "epoch": 0.2715017394665636,
      "grad_norm": 11.032714848983742,
      "learning_rate": 3.642732895245458e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 5619
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.875,
      "epoch": 0.27155005798221876,
      "grad_norm": 0.0,
      "learning_rate": 3.6424913026671814e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5620
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 102.59375,
      "epoch": 0.271598376497874,
      "grad_norm": 0.0,
      "learning_rate": 3.642249710088906e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 5621
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.6875,
      "epoch": 0.27164669501352917,
      "grad_norm": 0.0,
      "learning_rate": 3.6420081175106297e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5622
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.375,
      "epoch": 0.2716950135291844,
      "grad_norm": 0.0,
      "learning_rate": 3.641766524932354e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5623
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.28125,
      "epoch": 0.2717433320448396,
      "grad_norm": 0.0,
      "learning_rate": 3.641524932354078e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5624
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.59375,
      "epoch": 0.27179165056049476,
      "grad_norm": 0.0,
      "learning_rate": 3.641283339775802e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5625
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.78125,
      "epoch": 0.27183996907615,
      "grad_norm": 0.0,
      "learning_rate": 3.641041747197526e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5626
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.5,
      "epoch": 0.27188828759180517,
      "grad_norm": 0.0,
      "learning_rate": 3.6408001546192504e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5627
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.625,
      "epoch": 0.2719366061074604,
      "grad_norm": 0.0,
      "learning_rate": 3.640558562040974e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5628
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.96875,
      "epoch": 0.2719849246231156,
      "grad_norm": 0.0,
      "learning_rate": 3.6403169694626976e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5629
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.15625,
      "epoch": 0.27203324313877075,
      "grad_norm": 0.0,
      "learning_rate": 3.640075376884422e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5630
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.5,
      "epoch": 0.272081561654426,
      "grad_norm": 0.0,
      "learning_rate": 3.639833784306146e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5631
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.28125,
      "epoch": 0.27212988017008116,
      "grad_norm": 0.0,
      "learning_rate": 3.63959219172787e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5632
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.28125,
      "epoch": 0.2721781986857364,
      "grad_norm": 0.0,
      "learning_rate": 3.6393505991495936e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5633
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.875,
      "epoch": 0.27222651720139157,
      "grad_norm": 0.0,
      "learning_rate": 3.6391090065713183e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5634
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.75,
      "epoch": 0.27227483571704675,
      "grad_norm": 0.0,
      "learning_rate": 3.638867413993042e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5635
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.5625,
      "epoch": 0.272323154232702,
      "grad_norm": 8.253857400495054,
      "learning_rate": 3.638625821414766e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5636
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.59375,
      "epoch": 0.27237147274835716,
      "grad_norm": 0.0,
      "learning_rate": 3.63838422883649e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5637
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.1875,
      "epoch": 0.2724197912640124,
      "grad_norm": 0.0,
      "learning_rate": 3.6381426362582143e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5638
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.09375,
      "epoch": 0.27246810977966757,
      "grad_norm": 0.0,
      "learning_rate": 3.637901043679938e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5639
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.0625,
      "epoch": 0.27251642829532274,
      "grad_norm": 0.0,
      "learning_rate": 3.6376594511016615e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5640
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.15625,
      "epoch": 0.272564746810978,
      "grad_norm": 0.0,
      "learning_rate": 3.637417858523386e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5641
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.0625,
      "epoch": 0.27261306532663315,
      "grad_norm": 0.0,
      "learning_rate": 3.63717626594511e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5642
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.25,
      "epoch": 0.2726613838422884,
      "grad_norm": 0.0,
      "learning_rate": 3.636934673366834e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5643
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.875,
      "epoch": 0.27270970235794356,
      "grad_norm": 0.0,
      "learning_rate": 3.636693080788558e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5644
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.28125,
      "epoch": 0.27275802087359874,
      "grad_norm": 0.0,
      "learning_rate": 3.636451488210282e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5645
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.3125,
      "epoch": 0.27280633938925397,
      "grad_norm": 18.084203167285974,
      "learning_rate": 3.636209895632006e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5646
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.875,
      "epoch": 0.27285465790490915,
      "grad_norm": 0.0,
      "learning_rate": 3.6359683030537305e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5647
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.90625,
      "epoch": 0.2729029764205644,
      "grad_norm": 0.0,
      "learning_rate": 3.635726710475454e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5648
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.84375,
      "epoch": 0.27295129493621956,
      "grad_norm": 0.0,
      "learning_rate": 3.6354851178971777e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5649
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.40625,
      "epoch": 0.27299961345187473,
      "grad_norm": 0.0,
      "learning_rate": 3.635243525318902e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5650
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.65625,
      "epoch": 0.27304793196752997,
      "grad_norm": 0.0,
      "learning_rate": 3.635001932740626e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5651
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.78125,
      "epoch": 0.27309625048318514,
      "grad_norm": 0.0,
      "learning_rate": 3.63476034016235e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5652
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.46875,
      "epoch": 0.2731445689988404,
      "grad_norm": 0.0,
      "learning_rate": 3.6345187475840737e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5653
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.375,
      "epoch": 0.27319288751449555,
      "grad_norm": 0.0,
      "learning_rate": 3.6342771550057984e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5654
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 96.25,
      "epoch": 0.27324120603015073,
      "grad_norm": 0.0,
      "learning_rate": 3.634035562427522e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 5655
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.03125,
      "epoch": 0.27328952454580596,
      "grad_norm": 0.0,
      "learning_rate": 3.633793969849246e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5656
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.21875,
      "epoch": 0.27333784306146114,
      "grad_norm": 0.0,
      "learning_rate": 3.63355237727097e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5657
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.8125,
      "epoch": 0.27338616157711637,
      "grad_norm": 0.0,
      "learning_rate": 3.6333107846926944e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5658
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.8125,
      "epoch": 0.27343448009277155,
      "grad_norm": 0.0,
      "learning_rate": 3.633069192114418e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5659
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 96.5625,
      "epoch": 0.2734827986084267,
      "grad_norm": 334.4490143262005,
      "learning_rate": 3.6328275995361416e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5660
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.25,
      "epoch": 0.27353111712408196,
      "grad_norm": 0.0,
      "learning_rate": 3.6325860069578663e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 5661
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.15625,
      "epoch": 0.27357943563973713,
      "grad_norm": 0.0,
      "learning_rate": 3.63234441437959e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5662
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.15625,
      "epoch": 0.27362775415539237,
      "grad_norm": 0.0,
      "learning_rate": 3.632102821801314e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5663
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.625,
      "epoch": 0.27367607267104754,
      "grad_norm": 0.0,
      "learning_rate": 3.631861229223038e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5664
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.4375,
      "epoch": 0.2737243911867027,
      "grad_norm": 0.0,
      "learning_rate": 3.6316196366447623e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5665
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.59375,
      "epoch": 0.27377270970235795,
      "grad_norm": 0.0,
      "learning_rate": 3.631378044066486e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5666
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.8125,
      "epoch": 0.27382102821801313,
      "grad_norm": 8.388569356144913,
      "learning_rate": 3.6311364514882106e-07,
      "loss": -0.0,
      "reward": 2.265625,
      "reward_std": 0.32878512144088745,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5667
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.0,
      "epoch": 0.27386934673366836,
      "grad_norm": 0.0,
      "learning_rate": 3.630894858909934e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5668
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.78125,
      "epoch": 0.27391766524932354,
      "grad_norm": 6.57084639225868,
      "learning_rate": 3.630653266331658e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5669
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.9375,
      "epoch": 0.2739659837649787,
      "grad_norm": 0.0,
      "learning_rate": 3.630411673753382e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5670
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.875,
      "epoch": 0.27401430228063395,
      "grad_norm": 0.0,
      "learning_rate": 3.630170081175106e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5671
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.6875,
      "epoch": 0.2740626207962891,
      "grad_norm": 0.0,
      "learning_rate": 3.62992848859683e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5672
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.03125,
      "epoch": 0.27411093931194436,
      "grad_norm": 0.0,
      "learning_rate": 3.629686896018554e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5673
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.28125,
      "epoch": 0.27415925782759953,
      "grad_norm": 0.0,
      "learning_rate": 3.6294453034402785e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5674
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.78125,
      "epoch": 0.2742075763432547,
      "grad_norm": 0.0,
      "learning_rate": 3.629203710862002e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5675
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.4375,
      "epoch": 0.27425589485890994,
      "grad_norm": 0.0,
      "learning_rate": 3.628962118283726e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5676
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.09375,
      "epoch": 0.2743042133745651,
      "grad_norm": 0.0,
      "learning_rate": 3.6287205257054503e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5677
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.4375,
      "epoch": 0.27435253189022035,
      "grad_norm": 0.0,
      "learning_rate": 3.6284789331271745e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 5678
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.34375,
      "epoch": 0.27440085040587553,
      "grad_norm": 0.0,
      "learning_rate": 3.628237340548898e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5679
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.65625,
      "epoch": 0.2744491689215307,
      "grad_norm": 0.0,
      "learning_rate": 3.6279957479706217e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5680
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.25,
      "epoch": 0.27449748743718594,
      "grad_norm": 0.0,
      "learning_rate": 3.6277541553923463e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5681
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.40625,
      "epoch": 0.2745458059528411,
      "grad_norm": 0.0,
      "learning_rate": 3.62751256281407e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5682
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.46875,
      "epoch": 0.27459412446849635,
      "grad_norm": 0.0,
      "learning_rate": 3.627270970235794e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5683
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.96875,
      "epoch": 0.2746424429841515,
      "grad_norm": 0.0,
      "learning_rate": 3.627029377657518e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5684
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.78125,
      "epoch": 0.2746907614998067,
      "grad_norm": 0.0,
      "learning_rate": 3.6267877850792424e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5685
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.28125,
      "epoch": 0.27473908001546193,
      "grad_norm": 0.0,
      "learning_rate": 3.626546192500966e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5686
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.625,
      "epoch": 0.2747873985311171,
      "grad_norm": 12.7944296165538,
      "learning_rate": 3.6263045999226906e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.2920827269554138,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5687
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.78125,
      "epoch": 0.27483571704677234,
      "grad_norm": 0.0,
      "learning_rate": 3.626063007344414e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5688
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 89.65625,
      "epoch": 0.2748840355624275,
      "grad_norm": 25.862673009713042,
      "learning_rate": 3.6258214147661384e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5689
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.96875,
      "epoch": 0.2749323540780827,
      "grad_norm": 0.0,
      "learning_rate": 3.625579822187862e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5690
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.1875,
      "epoch": 0.27498067259373793,
      "grad_norm": 0.0,
      "learning_rate": 3.625338229609586e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5691
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.5625,
      "epoch": 0.2750289911093931,
      "grad_norm": 0.0,
      "learning_rate": 3.62509663703131e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5692
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.8125,
      "epoch": 0.27507730962504834,
      "grad_norm": 0.0,
      "learning_rate": 3.624855044453034e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5693
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 70.65625,
      "epoch": 0.2751256281407035,
      "grad_norm": 26.966591127187385,
      "learning_rate": 3.6246134518747585e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.319751501083374,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5694
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.9375,
      "epoch": 0.2751739466563587,
      "grad_norm": 0.0,
      "learning_rate": 3.624371859296482e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5695
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.4375,
      "epoch": 0.2752222651720139,
      "grad_norm": 0.0,
      "learning_rate": 3.6241302667182063e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5696
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.90625,
      "epoch": 0.2752705836876691,
      "grad_norm": 0.0,
      "learning_rate": 3.6238886741399304e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5697
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.5625,
      "epoch": 0.27531890220332433,
      "grad_norm": 0.0,
      "learning_rate": 3.6236470815616545e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5698
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.8125,
      "epoch": 0.2753672207189795,
      "grad_norm": 0.0,
      "learning_rate": 3.623405488983378e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5699
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.9375,
      "epoch": 0.2754155392346347,
      "grad_norm": 0.0,
      "learning_rate": 3.623163896405102e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5700
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.46875,
      "epoch": 0.2754638577502899,
      "grad_norm": 0.0,
      "learning_rate": 3.6229223038268264e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5701
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.90625,
      "epoch": 0.2755121762659451,
      "grad_norm": 0.0,
      "learning_rate": 3.62268071124855e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 5702
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.3125,
      "epoch": 0.27556049478160033,
      "grad_norm": 0.0,
      "learning_rate": 3.622439118670274e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5703
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.09375,
      "epoch": 0.2756088132972555,
      "grad_norm": 0.0,
      "learning_rate": 3.6221975260919983e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5704
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.34375,
      "epoch": 0.2756571318129107,
      "grad_norm": 0.0,
      "learning_rate": 3.6219559335137224e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5705
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.90625,
      "epoch": 0.2757054503285659,
      "grad_norm": 24.46146381547606,
      "learning_rate": 3.621714340935446e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.25069957971572876,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5706
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 85.71875,
      "epoch": 0.2757537688442211,
      "grad_norm": 16.24314776062899,
      "learning_rate": 3.6214727483571707e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5707
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.59375,
      "epoch": 0.2758020873598763,
      "grad_norm": 0.0,
      "learning_rate": 3.6212311557788943e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5708
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.03125,
      "epoch": 0.2758504058755315,
      "grad_norm": 0.0,
      "learning_rate": 3.6209895632006185e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5709
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.25,
      "epoch": 0.2758987243911867,
      "grad_norm": 0.0,
      "learning_rate": 3.620747970622342e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5710
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.03125,
      "epoch": 0.2759470429068419,
      "grad_norm": 0.0,
      "learning_rate": 3.620506378044066e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5711
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.75,
      "epoch": 0.2759953614224971,
      "grad_norm": 0.0,
      "learning_rate": 3.6202647854657903e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5712
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.0,
      "epoch": 0.2760436799381523,
      "grad_norm": 0.0,
      "learning_rate": 3.620023192887514e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5713
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.34375,
      "epoch": 0.2760919984538075,
      "grad_norm": 0.0,
      "learning_rate": 3.6197816003092386e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5714
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.0625,
      "epoch": 0.2761403169694627,
      "grad_norm": 0.0,
      "learning_rate": 3.619540007730962e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5715
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.125,
      "epoch": 0.2761886354851179,
      "grad_norm": 0.0,
      "learning_rate": 3.6192984151526864e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5716
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 85.09375,
      "epoch": 0.2762369540007731,
      "grad_norm": 57.38454462395573,
      "learning_rate": 3.6190568225744105e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5717
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.1875,
      "epoch": 0.2762852725164283,
      "grad_norm": 0.0,
      "learning_rate": 3.6188152299961346e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5718
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 88.625,
      "epoch": 0.2763335910320835,
      "grad_norm": 0.0,
      "learning_rate": 3.618573637417858e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 0.96875,
      "step": 5719
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.5,
      "epoch": 0.27638190954773867,
      "grad_norm": 0.0,
      "learning_rate": 3.618332044839582e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5720
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.28125,
      "epoch": 0.2764302280633939,
      "grad_norm": 0.0,
      "learning_rate": 3.6180904522613065e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5721
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.375,
      "epoch": 0.2764785465790491,
      "grad_norm": 0.0,
      "learning_rate": 3.61784885968303e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5722
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.0,
      "epoch": 0.2765268650947043,
      "grad_norm": 0.0,
      "learning_rate": 3.617607267104754e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5723
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.625,
      "epoch": 0.2765751836103595,
      "grad_norm": 0.0,
      "learning_rate": 3.6173656745264784e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5724
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.0,
      "epoch": 0.27662350212601466,
      "grad_norm": 0.0,
      "learning_rate": 3.6171240819482025e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5725
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.875,
      "epoch": 0.2766718206416699,
      "grad_norm": 0.0,
      "learning_rate": 3.616882489369926e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5726
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.34375,
      "epoch": 0.2767201391573251,
      "grad_norm": 0.0,
      "learning_rate": 3.616640896791651e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5727
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.6875,
      "epoch": 0.2767684576729803,
      "grad_norm": 0.0,
      "learning_rate": 3.6163993042133744e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5728
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.0625,
      "epoch": 0.2768167761886355,
      "grad_norm": 0.0,
      "learning_rate": 3.6161577116350985e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 5729
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.75,
      "epoch": 0.27686509470429066,
      "grad_norm": 0.0,
      "learning_rate": 3.615916119056822e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5730
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.0,
      "epoch": 0.2769134132199459,
      "grad_norm": 0.0,
      "learning_rate": 3.6156745264785463e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5731
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.6875,
      "epoch": 0.27696173173560107,
      "grad_norm": 0.0,
      "learning_rate": 3.6154329339002704e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5732
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.75,
      "epoch": 0.2770100502512563,
      "grad_norm": 0.0,
      "learning_rate": 3.615191341321994e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5733
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.9375,
      "epoch": 0.2770583687669115,
      "grad_norm": 0.0,
      "learning_rate": 3.6149497487437187e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 5734
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.53125,
      "epoch": 0.27710668728256665,
      "grad_norm": 0.0,
      "learning_rate": 3.6147081561654423e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5735
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.15625,
      "epoch": 0.2771550057982219,
      "grad_norm": 0.0,
      "learning_rate": 3.6144665635871664e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5736
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.28125,
      "epoch": 0.27720332431387706,
      "grad_norm": 0.0,
      "learning_rate": 3.6142249710088906e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5737
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.1875,
      "epoch": 0.2772516428295323,
      "grad_norm": 0.0,
      "learning_rate": 3.6139833784306147e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5738
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.15625,
      "epoch": 0.2772999613451875,
      "grad_norm": 0.0,
      "learning_rate": 3.6137417858523383e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5739
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.5,
      "epoch": 0.27734827986084265,
      "grad_norm": 0.0,
      "learning_rate": 3.6135001932740624e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5740
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.625,
      "epoch": 0.2773965983764979,
      "grad_norm": 0.0,
      "learning_rate": 3.6132586006957866e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5741
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.78125,
      "epoch": 0.27744491689215306,
      "grad_norm": 0.0,
      "learning_rate": 3.61301700811751e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5742
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.96875,
      "epoch": 0.2774932354078083,
      "grad_norm": 0.0,
      "learning_rate": 3.6127754155392343e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5743
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.3125,
      "epoch": 0.27754155392346347,
      "grad_norm": 0.0,
      "learning_rate": 3.6125338229609585e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5744
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.5625,
      "epoch": 0.27758987243911865,
      "grad_norm": 0.0,
      "learning_rate": 3.6122922303826826e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5745
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.25,
      "epoch": 0.2776381909547739,
      "grad_norm": 0.0,
      "learning_rate": 3.612050637804406e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5746
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.9375,
      "epoch": 0.27768650947042905,
      "grad_norm": 0.0,
      "learning_rate": 3.611809045226131e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5747
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.8125,
      "epoch": 0.2777348279860843,
      "grad_norm": 0.0,
      "learning_rate": 3.6115674526478545e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5748
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.03125,
      "epoch": 0.27778314650173946,
      "grad_norm": 0.0,
      "learning_rate": 3.6113258600695786e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5749
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.1875,
      "epoch": 0.27783146501739464,
      "grad_norm": 0.0,
      "learning_rate": 3.611084267491302e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5750
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.8125,
      "epoch": 0.2778797835330499,
      "grad_norm": 0.0,
      "learning_rate": 3.610842674913027e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 5751
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 92.8125,
      "epoch": 0.27792810204870505,
      "grad_norm": 8.240018257955311,
      "learning_rate": 3.6106010823347505e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5752
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.65625,
      "epoch": 0.2779764205643603,
      "grad_norm": 0.0,
      "learning_rate": 3.610359489756474e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5753
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.875,
      "epoch": 0.27802473908001546,
      "grad_norm": 0.0,
      "learning_rate": 3.610117897178199e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5754
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.25,
      "epoch": 0.27807305759567064,
      "grad_norm": 0.0,
      "learning_rate": 3.6098763045999224e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5755
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.1875,
      "epoch": 0.27812137611132587,
      "grad_norm": 0.0,
      "learning_rate": 3.6096347120216465e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5756
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.46875,
      "epoch": 0.27816969462698105,
      "grad_norm": 0.0,
      "learning_rate": 3.6093931194433706e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5757
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.15625,
      "epoch": 0.2782180131426363,
      "grad_norm": 0.0,
      "learning_rate": 3.609151526865095e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5758
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.65625,
      "epoch": 0.27826633165829145,
      "grad_norm": 0.0,
      "learning_rate": 3.6089099342868184e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 5759
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.25,
      "epoch": 0.27831465017394663,
      "grad_norm": 0.0,
      "learning_rate": 3.6086683417085425e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5760
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.03125,
      "epoch": 0.27836296868960186,
      "grad_norm": 0.0,
      "learning_rate": 3.6084267491302667e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5761
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.59375,
      "epoch": 0.27841128720525704,
      "grad_norm": 0.0,
      "learning_rate": 3.6081851565519903e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5762
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.46875,
      "epoch": 0.2784596057209123,
      "grad_norm": 0.0,
      "learning_rate": 3.6079435639737144e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5763
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.96875,
      "epoch": 0.27850792423656745,
      "grad_norm": 0.0,
      "learning_rate": 3.6077019713954385e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5764
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.625,
      "epoch": 0.2785562427522226,
      "grad_norm": 0.0,
      "learning_rate": 3.6074603788171627e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5765
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.09375,
      "epoch": 0.27860456126787786,
      "grad_norm": 0.0,
      "learning_rate": 3.6072187862388863e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5766
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.4375,
      "epoch": 0.27865287978353304,
      "grad_norm": 0.0,
      "learning_rate": 3.606977193660611e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5767
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.875,
      "epoch": 0.27870119829918827,
      "grad_norm": 0.0,
      "learning_rate": 3.6067356010823346e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 5768
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.84375,
      "epoch": 0.27874951681484345,
      "grad_norm": 0.0,
      "learning_rate": 3.6064940085040587e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5769
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.84375,
      "epoch": 0.2787978353304986,
      "grad_norm": 0.0,
      "learning_rate": 3.6062524159257823e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5770
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.34375,
      "epoch": 0.27884615384615385,
      "grad_norm": 0.0,
      "learning_rate": 3.606010823347507e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5771
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.6875,
      "epoch": 0.27889447236180903,
      "grad_norm": 0.0,
      "learning_rate": 3.6057692307692306e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5772
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.53125,
      "epoch": 0.27894279087746426,
      "grad_norm": 0.0,
      "learning_rate": 3.605527638190954e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5773
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.125,
      "epoch": 0.27899110939311944,
      "grad_norm": 0.0,
      "learning_rate": 3.605286045612679e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5774
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.34375,
      "epoch": 0.2790394279087746,
      "grad_norm": 0.0,
      "learning_rate": 3.6050444530344025e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5775
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.78125,
      "epoch": 0.27908774642442985,
      "grad_norm": 0.0,
      "learning_rate": 3.6048028604561266e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5776
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.96875,
      "epoch": 0.279136064940085,
      "grad_norm": 0.0,
      "learning_rate": 3.6045612678778507e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5777
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.875,
      "epoch": 0.27918438345574026,
      "grad_norm": 0.0,
      "learning_rate": 3.604319675299575e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5778
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.25,
      "epoch": 0.27923270197139544,
      "grad_norm": 0.0,
      "learning_rate": 3.6040780827212985e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5779
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.59375,
      "epoch": 0.2792810204870506,
      "grad_norm": 0.0,
      "learning_rate": 3.6038364901430226e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5780
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.90625,
      "epoch": 0.27932933900270585,
      "grad_norm": 0.0,
      "learning_rate": 3.603594897564747e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5781
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.75,
      "epoch": 0.279377657518361,
      "grad_norm": 0.0,
      "learning_rate": 3.6033533049864704e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 5782
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 92.78125,
      "epoch": 0.27942597603401625,
      "grad_norm": 8.9641665661394,
      "learning_rate": 3.6031117124081945e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5783
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.1875,
      "epoch": 0.27947429454967143,
      "grad_norm": 0.0,
      "learning_rate": 3.6028701198299186e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5784
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.90625,
      "epoch": 0.2795226130653266,
      "grad_norm": 0.0,
      "learning_rate": 3.602628527251643e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5785
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.34375,
      "epoch": 0.27957093158098184,
      "grad_norm": 15.723622623039633,
      "learning_rate": 3.6023869346733664e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5786
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.5,
      "epoch": 0.279619250096637,
      "grad_norm": 0.0,
      "learning_rate": 3.602145342095091e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5787
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.71875,
      "epoch": 0.27966756861229225,
      "grad_norm": 0.0,
      "learning_rate": 3.6019037495168146e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5788
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.6875,
      "epoch": 0.2797158871279474,
      "grad_norm": 0.0,
      "learning_rate": 3.601662156938539e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 5789
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.46875,
      "epoch": 0.2797642056436026,
      "grad_norm": 0.0,
      "learning_rate": 3.6014205643602624e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5790
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.40625,
      "epoch": 0.27981252415925784,
      "grad_norm": 0.0,
      "learning_rate": 3.601178971781987e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5791
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.1875,
      "epoch": 0.279860842674913,
      "grad_norm": 0.0,
      "learning_rate": 3.6009373792037107e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5792
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.3125,
      "epoch": 0.27990916119056825,
      "grad_norm": 0.0,
      "learning_rate": 3.600695786625434e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5793
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.90625,
      "epoch": 0.2799574797062234,
      "grad_norm": 25.119941854032085,
      "learning_rate": 3.600454194047159e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.323003351688385,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5794
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.875,
      "epoch": 0.2800057982218786,
      "grad_norm": 0.0,
      "learning_rate": 3.6002126014688825e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5795
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.25,
      "epoch": 0.28005411673753383,
      "grad_norm": 0.0,
      "learning_rate": 3.5999710088906067e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5796
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.21875,
      "epoch": 0.280102435253189,
      "grad_norm": 0.0,
      "learning_rate": 3.599729416312331e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5797
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.375,
      "epoch": 0.28015075376884424,
      "grad_norm": 0.0,
      "learning_rate": 3.599487823734055e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5798
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.0,
      "epoch": 0.2801990722844994,
      "grad_norm": 0.0,
      "learning_rate": 3.5992462311557785e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5799
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.53125,
      "epoch": 0.2802473908001546,
      "grad_norm": 0.0,
      "learning_rate": 3.5990046385775027e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5800
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.9375,
      "epoch": 0.2802957093158098,
      "grad_norm": 0.0,
      "learning_rate": 3.598763045999227e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5801
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.0625,
      "epoch": 0.280344027831465,
      "grad_norm": 0.0,
      "learning_rate": 3.598521453420951e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5802
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.3125,
      "epoch": 0.28039234634712024,
      "grad_norm": 0.0,
      "learning_rate": 3.5982798608426746e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5803
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.25,
      "epoch": 0.2804406648627754,
      "grad_norm": 0.0,
      "learning_rate": 3.5980382682643987e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5804
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.71875,
      "epoch": 0.2804889833784306,
      "grad_norm": 0.0,
      "learning_rate": 3.597796675686123e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5805
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.90625,
      "epoch": 0.2805373018940858,
      "grad_norm": 71.67940225315888,
      "learning_rate": 3.5975550831078464e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5806
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.84375,
      "epoch": 0.280585620409741,
      "grad_norm": 0.0,
      "learning_rate": 3.597313490529571e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5807
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.53125,
      "epoch": 0.28063393892539623,
      "grad_norm": 0.0,
      "learning_rate": 3.5970718979512947e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 5808
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.125,
      "epoch": 0.2806822574410514,
      "grad_norm": 0.0,
      "learning_rate": 3.596830305373019e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5809
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 97.5,
      "epoch": 0.2807305759567066,
      "grad_norm": 0.0,
      "learning_rate": 3.5965887127947425e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 5810
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.875,
      "epoch": 0.2807788944723618,
      "grad_norm": 0.0,
      "learning_rate": 3.596347120216467e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5811
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.34375,
      "epoch": 0.280827212988017,
      "grad_norm": 55.889578979302335,
      "learning_rate": 3.5961055276381907e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.26726123690605164,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5812
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.375,
      "epoch": 0.2808755315036722,
      "grad_norm": 9.188134382857113,
      "learning_rate": 3.5958639350599143e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.24831002950668335,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5813
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.65625,
      "epoch": 0.2809238500193274,
      "grad_norm": 0.0,
      "learning_rate": 3.595622342481639e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5814
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.53125,
      "epoch": 0.2809721685349826,
      "grad_norm": 0.0,
      "learning_rate": 3.5953807499033626e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5815
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.875,
      "epoch": 0.2810204870506378,
      "grad_norm": 0.0,
      "learning_rate": 3.595139157325087e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5816
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.5625,
      "epoch": 0.281068805566293,
      "grad_norm": 0.0,
      "learning_rate": 3.594897564746811e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5817
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.59375,
      "epoch": 0.2811171240819482,
      "grad_norm": 0.0,
      "learning_rate": 3.594655972168535e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5818
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.8125,
      "epoch": 0.2811654425976034,
      "grad_norm": 0.0,
      "learning_rate": 3.5944143795902586e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 5819
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.96875,
      "epoch": 0.2812137611132586,
      "grad_norm": 0.0,
      "learning_rate": 3.594172787011983e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5820
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.8125,
      "epoch": 0.2812620796289138,
      "grad_norm": 0.0,
      "learning_rate": 3.593931194433707e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5821
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.46875,
      "epoch": 0.281310398144569,
      "grad_norm": 0.0,
      "learning_rate": 3.593689601855431e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 5822
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.5625,
      "epoch": 0.2813587166602242,
      "grad_norm": 0.0,
      "learning_rate": 3.5934480092771546e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5823
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.5625,
      "epoch": 0.2814070351758794,
      "grad_norm": 0.0,
      "learning_rate": 3.593206416698879e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5824
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.84375,
      "epoch": 0.28145535369153457,
      "grad_norm": 0.0,
      "learning_rate": 3.592964824120603e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5825
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.53125,
      "epoch": 0.2815036722071898,
      "grad_norm": 0.0,
      "learning_rate": 3.5927232315423265e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5826
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.3125,
      "epoch": 0.281551990722845,
      "grad_norm": 0.0,
      "learning_rate": 3.592481638964051e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5827
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.625,
      "epoch": 0.2816003092385002,
      "grad_norm": 0.0,
      "learning_rate": 3.592240046385775e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5828
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.125,
      "epoch": 0.2816486277541554,
      "grad_norm": 0.0,
      "learning_rate": 3.591998453807499e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5829
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.0625,
      "epoch": 0.28169694626981057,
      "grad_norm": 0.0,
      "learning_rate": 3.5917568612292225e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5830
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.6875,
      "epoch": 0.2817452647854658,
      "grad_norm": 0.0,
      "learning_rate": 3.591515268650947e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5831
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 100.875,
      "epoch": 0.281793583301121,
      "grad_norm": 15.137584457036171,
      "learning_rate": 3.591273676072671e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5832
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.875,
      "epoch": 0.2818419018167762,
      "grad_norm": 0.0,
      "learning_rate": 3.5910320834943944e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5833
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.34375,
      "epoch": 0.2818902203324314,
      "grad_norm": 0.0,
      "learning_rate": 3.590790490916119e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5834
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.46875,
      "epoch": 0.28193853884808656,
      "grad_norm": 0.0,
      "learning_rate": 3.5905488983378427e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5835
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 105.875,
      "epoch": 0.2819868573637418,
      "grad_norm": 28.954766744648882,
      "learning_rate": 3.590307305759567e-07,
      "loss": -0.0,
      "reward": 1.546875,
      "reward_std": 0.36403757333755493,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 5836
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.15625,
      "epoch": 0.28203517587939697,
      "grad_norm": 0.0,
      "learning_rate": 3.590065713181291e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5837
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.1875,
      "epoch": 0.2820834943950522,
      "grad_norm": 0.0,
      "learning_rate": 3.589824120603015e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5838
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.59375,
      "epoch": 0.2821318129107074,
      "grad_norm": 0.0,
      "learning_rate": 3.5895825280247387e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5839
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.34375,
      "epoch": 0.28218013142636256,
      "grad_norm": 0.0,
      "learning_rate": 3.589340935446463e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5840
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 89.40625,
      "epoch": 0.2822284499420178,
      "grad_norm": 11.96555674382888,
      "learning_rate": 3.589099342868187e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.17598573863506317,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5841
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.375,
      "epoch": 0.28227676845767297,
      "grad_norm": 0.0,
      "learning_rate": 3.588857750289911e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5842
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 125.5625,
      "epoch": 0.2823250869733282,
      "grad_norm": 2.7523926482313437,
      "learning_rate": 3.5886161577116347e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.35247981548309326,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 0.96875,
      "step": 5843
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.21875,
      "epoch": 0.2823734054889834,
      "grad_norm": 0.0,
      "learning_rate": 3.588374565133359e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5844
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.0625,
      "epoch": 0.28242172400463855,
      "grad_norm": 0.0,
      "learning_rate": 3.588132972555083e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5845
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.1875,
      "epoch": 0.2824700425202938,
      "grad_norm": 0.0,
      "learning_rate": 3.5878913799768066e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 5846
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.875,
      "epoch": 0.28251836103594896,
      "grad_norm": 0.0,
      "learning_rate": 3.5876497873985313e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5847
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.53125,
      "epoch": 0.2825666795516042,
      "grad_norm": 0.0,
      "learning_rate": 3.587408194820255e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5848
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.15625,
      "epoch": 0.28261499806725937,
      "grad_norm": 0.0,
      "learning_rate": 3.587166602241979e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5849
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 107.78125,
      "epoch": 0.28266331658291455,
      "grad_norm": 0.0,
      "learning_rate": 3.5869250096637026e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5850
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.09375,
      "epoch": 0.2827116350985698,
      "grad_norm": 0.0,
      "learning_rate": 3.5866834170854273e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5851
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.59375,
      "epoch": 0.28275995361422496,
      "grad_norm": 0.0,
      "learning_rate": 3.586441824507151e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5852
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.40625,
      "epoch": 0.2828082721298802,
      "grad_norm": 0.0,
      "learning_rate": 3.586200231928875e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5853
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.625,
      "epoch": 0.28285659064553537,
      "grad_norm": 0.0,
      "learning_rate": 3.585958639350599e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5854
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 89.5,
      "epoch": 0.28290490916119054,
      "grad_norm": 36.96123620447894,
      "learning_rate": 3.585717046772323e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.3298332095146179,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5855
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 104.71875,
      "epoch": 0.2829532276768458,
      "grad_norm": 7.280739374557648,
      "learning_rate": 3.585475454194047e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.3359074294567108,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5856
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 113.59375,
      "epoch": 0.28300154619250095,
      "grad_norm": 0.0,
      "learning_rate": 3.585233861615771e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 5857
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.0,
      "epoch": 0.2830498647081562,
      "grad_norm": 0.0,
      "learning_rate": 3.584992269037495e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5858
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.34375,
      "epoch": 0.28309818322381136,
      "grad_norm": 0.0,
      "learning_rate": 3.584750676459219e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5859
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 107.625,
      "epoch": 0.28314650173946654,
      "grad_norm": 0.0,
      "learning_rate": 3.584509083880943e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 5860
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.375,
      "epoch": 0.28319482025512177,
      "grad_norm": 0.0,
      "learning_rate": 3.584267491302667e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5861
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.0625,
      "epoch": 0.28324313877077695,
      "grad_norm": 0.0,
      "learning_rate": 3.584025898724391e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5862
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 118.5625,
      "epoch": 0.2832914572864322,
      "grad_norm": 0.0,
      "learning_rate": 3.583784306146115e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 5863
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.625,
      "epoch": 0.28333977580208736,
      "grad_norm": 0.0,
      "learning_rate": 3.5835427135678395e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5864
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.03125,
      "epoch": 0.28338809431774253,
      "grad_norm": 0.0,
      "learning_rate": 3.583301120989563e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5865
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.4375,
      "epoch": 0.28343641283339777,
      "grad_norm": 0.0,
      "learning_rate": 3.5830595284112867e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5866
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.84375,
      "epoch": 0.28348473134905294,
      "grad_norm": 0.0,
      "learning_rate": 3.5828179358330113e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5867
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.0,
      "epoch": 0.2835330498647082,
      "grad_norm": 0.0,
      "learning_rate": 3.582576343254735e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5868
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.59375,
      "epoch": 0.28358136838036335,
      "grad_norm": 0.0,
      "learning_rate": 3.582334750676459e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5869
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.5625,
      "epoch": 0.28362968689601853,
      "grad_norm": 0.0,
      "learning_rate": 3.5820931580981827e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5870
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 112.125,
      "epoch": 0.28367800541167376,
      "grad_norm": 35.40834245568123,
      "learning_rate": 3.5818515655199074e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5871
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.25,
      "epoch": 0.28372632392732894,
      "grad_norm": 0.0,
      "learning_rate": 3.581609972941631e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5872
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.0,
      "epoch": 0.28377464244298417,
      "grad_norm": 0.0,
      "learning_rate": 3.581368380363355e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5873
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 115.0625,
      "epoch": 0.28382296095863935,
      "grad_norm": 6.94404346405804,
      "learning_rate": 3.581126787785079e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5874
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.1875,
      "epoch": 0.2838712794742945,
      "grad_norm": 0.0,
      "learning_rate": 3.580885195206803e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5875
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.65625,
      "epoch": 0.28391959798994976,
      "grad_norm": 0.0,
      "learning_rate": 3.580643602628527e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5876
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.9375,
      "epoch": 0.28396791650560493,
      "grad_norm": 0.0,
      "learning_rate": 3.580402010050251e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5877
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.84375,
      "epoch": 0.28401623502126017,
      "grad_norm": 0.0,
      "learning_rate": 3.580160417471975e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5878
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.875,
      "epoch": 0.28406455353691534,
      "grad_norm": 0.0,
      "learning_rate": 3.579918824893699e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5879
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.65625,
      "epoch": 0.2841128720525705,
      "grad_norm": 0.0,
      "learning_rate": 3.579677232315423e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5880
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.84375,
      "epoch": 0.28416119056822575,
      "grad_norm": 0.0,
      "learning_rate": 3.579435639737147e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5881
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.71875,
      "epoch": 0.28420950908388093,
      "grad_norm": 0.0,
      "learning_rate": 3.5791940471588713e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5882
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.6875,
      "epoch": 0.28425782759953616,
      "grad_norm": 0.0,
      "learning_rate": 3.578952454580595e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5883
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 109.90625,
      "epoch": 0.28430614611519134,
      "grad_norm": 0.0,
      "learning_rate": 3.5787108620023195e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 5884
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.375,
      "epoch": 0.2843544646308465,
      "grad_norm": 0.0,
      "learning_rate": 3.578469269424043e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5885
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.09375,
      "epoch": 0.28440278314650175,
      "grad_norm": 0.0,
      "learning_rate": 3.578227676845767e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5886
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.46875,
      "epoch": 0.2844511016621569,
      "grad_norm": 0.0,
      "learning_rate": 3.5779860842674914e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5887
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.90625,
      "epoch": 0.28449942017781216,
      "grad_norm": 0.0,
      "learning_rate": 3.577744491689215e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5888
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.28125,
      "epoch": 0.28454773869346733,
      "grad_norm": 0.0,
      "learning_rate": 3.577502899110939e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5889
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 104.96875,
      "epoch": 0.2845960572091225,
      "grad_norm": 14.701749022989636,
      "learning_rate": 3.577261306532663e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5890
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.96875,
      "epoch": 0.28464437572477774,
      "grad_norm": 0.0,
      "learning_rate": 3.5770197139543874e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5891
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.25,
      "epoch": 0.2846926942404329,
      "grad_norm": 0.0,
      "learning_rate": 3.576778121376111e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5892
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.03125,
      "epoch": 0.28474101275608815,
      "grad_norm": 0.0,
      "learning_rate": 3.576536528797835e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5893
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 115.46875,
      "epoch": 0.28478933127174333,
      "grad_norm": 14.894452786819985,
      "learning_rate": 3.5762949362195593e-07,
      "loss": -0.0,
      "reward": 1.625,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5894
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.0,
      "epoch": 0.2848376497873985,
      "grad_norm": 0.0,
      "learning_rate": 3.5760533436412835e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5895
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.8125,
      "epoch": 0.28488596830305374,
      "grad_norm": 0.0,
      "learning_rate": 3.575811751063007e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5896
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 114.15625,
      "epoch": 0.2849342868187089,
      "grad_norm": 26.06931004616874,
      "learning_rate": 3.575570158484731e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 5897
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 108.0,
      "epoch": 0.28498260533436415,
      "grad_norm": 0.0,
      "learning_rate": 3.5753285659064553e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 5898
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.28125,
      "epoch": 0.2850309238500193,
      "grad_norm": 0.0,
      "learning_rate": 3.575086973328179e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5899
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.46875,
      "epoch": 0.2850792423656745,
      "grad_norm": 0.0,
      "learning_rate": 3.574845380749903e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5900
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.21875,
      "epoch": 0.28512756088132973,
      "grad_norm": 0.0,
      "learning_rate": 3.574603788171627e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5901
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.53125,
      "epoch": 0.2851758793969849,
      "grad_norm": 0.0,
      "learning_rate": 3.5743621955933514e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5902
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.03125,
      "epoch": 0.28522419791264014,
      "grad_norm": 0.0,
      "learning_rate": 3.574120603015075e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5903
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.0,
      "epoch": 0.2852725164282953,
      "grad_norm": 0.0,
      "learning_rate": 3.5738790104367996e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5904
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.6875,
      "epoch": 0.2853208349439505,
      "grad_norm": 0.0,
      "learning_rate": 3.573637417858523e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5905
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 122.5625,
      "epoch": 0.28536915345960573,
      "grad_norm": 0.0,
      "learning_rate": 3.573395825280247e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 5906
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.4375,
      "epoch": 0.2854174719752609,
      "grad_norm": 0.0,
      "learning_rate": 3.5731542327019715e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 5907
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.46875,
      "epoch": 0.28546579049091614,
      "grad_norm": 0.0,
      "learning_rate": 3.572912640123695e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5908
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.15625,
      "epoch": 0.2855141090065713,
      "grad_norm": 0.0,
      "learning_rate": 3.572671047545419e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5909
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 100.59375,
      "epoch": 0.2855624275222265,
      "grad_norm": 0.0,
      "learning_rate": 3.572429454967143e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 5910
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.53125,
      "epoch": 0.2856107460378817,
      "grad_norm": 0.0,
      "learning_rate": 3.5721878623888675e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5911
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.3125,
      "epoch": 0.2856590645535369,
      "grad_norm": 0.0,
      "learning_rate": 3.571946269810591e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5912
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.8125,
      "epoch": 0.28570738306919213,
      "grad_norm": 0.0,
      "learning_rate": 3.571704677232315e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5913
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.09375,
      "epoch": 0.2857557015848473,
      "grad_norm": 0.0,
      "learning_rate": 3.5714630846540394e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5914
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.40625,
      "epoch": 0.2858040201005025,
      "grad_norm": 0.0,
      "learning_rate": 3.5712214920757635e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 5915
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.375,
      "epoch": 0.2858523386161577,
      "grad_norm": 0.0,
      "learning_rate": 3.570979899497487e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5916
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.96875,
      "epoch": 0.2859006571318129,
      "grad_norm": 0.0,
      "learning_rate": 3.5707383069192113e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5917
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.8125,
      "epoch": 0.28594897564746813,
      "grad_norm": 0.0,
      "learning_rate": 3.5704967143409354e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5918
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.125,
      "epoch": 0.2859972941631233,
      "grad_norm": 0.0,
      "learning_rate": 3.570255121762659e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5919
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 119.09375,
      "epoch": 0.2860456126787785,
      "grad_norm": 47.28172465756864,
      "learning_rate": 3.570013529184383e-07,
      "loss": -0.0,
      "reward": 1.09375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.0625,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 5920
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.8125,
      "epoch": 0.2860939311944337,
      "grad_norm": 0.0,
      "learning_rate": 3.5697719366061073e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5921
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.6875,
      "epoch": 0.2861422497100889,
      "grad_norm": 0.0,
      "learning_rate": 3.5695303440278314e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 5922
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.59375,
      "epoch": 0.2861905682257441,
      "grad_norm": 0.0,
      "learning_rate": 3.569288751449555e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5923
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.5,
      "epoch": 0.2862388867413993,
      "grad_norm": 0.0,
      "learning_rate": 3.5690471588712797e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5924
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 119.0,
      "epoch": 0.2862872052570545,
      "grad_norm": 0.0,
      "learning_rate": 3.5688055662930033e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 5925
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.84375,
      "epoch": 0.2863355237727097,
      "grad_norm": 0.0,
      "learning_rate": 3.568563973714727e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5926
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.03125,
      "epoch": 0.2863838422883649,
      "grad_norm": 0.0,
      "learning_rate": 3.5683223811364516e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5927
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.90625,
      "epoch": 0.2864321608040201,
      "grad_norm": 0.0,
      "learning_rate": 3.568080788558175e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 5928
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.96875,
      "epoch": 0.2864804793196753,
      "grad_norm": 0.0,
      "learning_rate": 3.5678391959798993e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5929
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 115.9375,
      "epoch": 0.2865287978353305,
      "grad_norm": 11.451596886635901,
      "learning_rate": 3.567597603401623e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5930
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.125,
      "epoch": 0.2865771163509857,
      "grad_norm": 0.0,
      "learning_rate": 3.5673560108233476e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5931
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.9375,
      "epoch": 0.2866254348666409,
      "grad_norm": 0.0,
      "learning_rate": 3.567114418245071e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5932
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.6875,
      "epoch": 0.2866737533822961,
      "grad_norm": 0.0,
      "learning_rate": 3.5668728256667953e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5933
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.125,
      "epoch": 0.2867220718979513,
      "grad_norm": 0.0,
      "learning_rate": 3.5666312330885195e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5934
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.59375,
      "epoch": 0.28677039041360647,
      "grad_norm": 0.0,
      "learning_rate": 3.5663896405102436e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 5935
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.84375,
      "epoch": 0.2868187089292617,
      "grad_norm": 0.0,
      "learning_rate": 3.566148047931967e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5936
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 109.9375,
      "epoch": 0.2868670274449169,
      "grad_norm": 0.0,
      "learning_rate": 3.5659064553536914e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5937
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 112.0,
      "epoch": 0.2869153459605721,
      "grad_norm": 10.499879291029679,
      "learning_rate": 3.5656648627754155e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5938
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.71875,
      "epoch": 0.2869636644762273,
      "grad_norm": 0.0,
      "learning_rate": 3.565423270197139e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5939
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.25,
      "epoch": 0.28701198299188246,
      "grad_norm": 0.0,
      "learning_rate": 3.565181677618863e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5940
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.96875,
      "epoch": 0.2870603015075377,
      "grad_norm": 0.0,
      "learning_rate": 3.5649400850405874e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5941
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.59375,
      "epoch": 0.2871086200231929,
      "grad_norm": 0.0,
      "learning_rate": 3.5646984924623115e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5942
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.25,
      "epoch": 0.2871569385388481,
      "grad_norm": 0.0,
      "learning_rate": 3.564456899884035e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5943
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.59375,
      "epoch": 0.2872052570545033,
      "grad_norm": 0.0,
      "learning_rate": 3.56421530730576e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5944
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.59375,
      "epoch": 0.28725357557015846,
      "grad_norm": 0.0,
      "learning_rate": 3.5639737147274834e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 5945
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.6875,
      "epoch": 0.2873018940858137,
      "grad_norm": 0.0,
      "learning_rate": 3.5637321221492075e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5946
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 94.34375,
      "epoch": 0.28735021260146887,
      "grad_norm": 23.12367397101218,
      "learning_rate": 3.5634905295709317e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.22152717411518097,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5947
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.71875,
      "epoch": 0.2873985311171241,
      "grad_norm": 0.0,
      "learning_rate": 3.5632489369926553e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5948
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.65625,
      "epoch": 0.2874468496327793,
      "grad_norm": 0.0,
      "learning_rate": 3.5630073444143794e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5949
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.53125,
      "epoch": 0.28749516814843445,
      "grad_norm": 0.0,
      "learning_rate": 3.562765751836103e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 5950
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.78125,
      "epoch": 0.2875434866640897,
      "grad_norm": 0.0,
      "learning_rate": 3.5625241592578277e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5951
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.6875,
      "epoch": 0.28759180517974486,
      "grad_norm": 0.0,
      "learning_rate": 3.5622825666795513e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5952
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.4375,
      "epoch": 0.2876401236954001,
      "grad_norm": 0.0,
      "learning_rate": 3.5620409741012754e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5953
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.875,
      "epoch": 0.2876884422110553,
      "grad_norm": 0.0,
      "learning_rate": 3.5617993815229996e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5954
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.625,
      "epoch": 0.28773676072671045,
      "grad_norm": 0.0,
      "learning_rate": 3.5615577889447237e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 5955
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.0,
      "epoch": 0.2877850792423657,
      "grad_norm": 0.0,
      "learning_rate": 3.5613161963664473e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5956
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.03125,
      "epoch": 0.28783339775802086,
      "grad_norm": 0.0,
      "learning_rate": 3.561074603788172e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5957
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 108.34375,
      "epoch": 0.2878817162736761,
      "grad_norm": 0.0,
      "learning_rate": 3.5608330112098956e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5958
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.15625,
      "epoch": 0.28793003478933127,
      "grad_norm": 0.0,
      "learning_rate": 3.560591418631619e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 5959
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.1875,
      "epoch": 0.28797835330498645,
      "grad_norm": 0.0,
      "learning_rate": 3.5603498260533433e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5960
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 102.1875,
      "epoch": 0.2880266718206417,
      "grad_norm": 33.37990982746681,
      "learning_rate": 3.5601082334750675e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5961
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.15625,
      "epoch": 0.28807499033629685,
      "grad_norm": 0.0,
      "learning_rate": 3.5598666408967916e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5962
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.25,
      "epoch": 0.2881233088519521,
      "grad_norm": 0.0,
      "learning_rate": 3.559625048318515e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 5963
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.0,
      "epoch": 0.28817162736760726,
      "grad_norm": 0.0,
      "learning_rate": 3.55938345574024e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5964
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.71875,
      "epoch": 0.28821994588326244,
      "grad_norm": 0.0,
      "learning_rate": 3.5591418631619635e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5965
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.75,
      "epoch": 0.2882682643989177,
      "grad_norm": 0.0,
      "learning_rate": 3.5589002705836876e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5966
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.53125,
      "epoch": 0.28831658291457285,
      "grad_norm": 0.0,
      "learning_rate": 3.558658678005412e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5967
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 103.8125,
      "epoch": 0.2883649014302281,
      "grad_norm": 0.0,
      "learning_rate": 3.5584170854271353e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5968
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.25,
      "epoch": 0.28841321994588326,
      "grad_norm": 0.0,
      "learning_rate": 3.5581754928488595e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5969
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.0,
      "epoch": 0.28846153846153844,
      "grad_norm": 0.0,
      "learning_rate": 3.557933900270583e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5970
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.53125,
      "epoch": 0.28850985697719367,
      "grad_norm": 0.0,
      "learning_rate": 3.557692307692308e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5971
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.6875,
      "epoch": 0.28855817549284885,
      "grad_norm": 0.0,
      "learning_rate": 3.5574507151140314e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 5972
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.09375,
      "epoch": 0.2886064940085041,
      "grad_norm": 0.0,
      "learning_rate": 3.5572091225357555e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5973
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.96875,
      "epoch": 0.28865481252415925,
      "grad_norm": 0.0,
      "learning_rate": 3.5569675299574796e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5974
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.53125,
      "epoch": 0.28870313103981443,
      "grad_norm": 0.0,
      "learning_rate": 3.556725937379204e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5975
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.5625,
      "epoch": 0.28875144955546966,
      "grad_norm": 0.0,
      "learning_rate": 3.5564843448009274e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5976
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 111.34375,
      "epoch": 0.28879976807112484,
      "grad_norm": 0.0,
      "learning_rate": 3.556242752222652e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5977
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.03125,
      "epoch": 0.2888480865867801,
      "grad_norm": 0.0,
      "learning_rate": 3.5560011596443757e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5978
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.34375,
      "epoch": 0.28889640510243525,
      "grad_norm": 0.0,
      "learning_rate": 3.555759567066099e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 5979
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.03125,
      "epoch": 0.2889447236180904,
      "grad_norm": 0.0,
      "learning_rate": 3.5555179744878234e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5980
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.4375,
      "epoch": 0.28899304213374566,
      "grad_norm": 0.0,
      "learning_rate": 3.5552763819095475e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5981
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.1875,
      "epoch": 0.28904136064940084,
      "grad_norm": 0.0,
      "learning_rate": 3.5550347893312717e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5982
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.5,
      "epoch": 0.28908967916505607,
      "grad_norm": 0.0,
      "learning_rate": 3.5547931967529953e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 5983
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.5625,
      "epoch": 0.28913799768071125,
      "grad_norm": 63.10793831058633,
      "learning_rate": 3.55455160417472e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.17598573863506317,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 5984
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.65625,
      "epoch": 0.2891863161963664,
      "grad_norm": 0.0,
      "learning_rate": 3.5543100115964435e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 5985
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.15625,
      "epoch": 0.28923463471202165,
      "grad_norm": 0.0,
      "learning_rate": 3.5540684190181677e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5986
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.46875,
      "epoch": 0.28928295322767683,
      "grad_norm": 0.0,
      "learning_rate": 3.553826826439892e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5987
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.0625,
      "epoch": 0.28933127174333206,
      "grad_norm": 0.0,
      "learning_rate": 3.5535852338616154e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5988
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.84375,
      "epoch": 0.28937959025898724,
      "grad_norm": 0.0,
      "learning_rate": 3.5533436412833396e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 5989
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 111.96875,
      "epoch": 0.2894279087746424,
      "grad_norm": 0.0,
      "learning_rate": 3.553102048705063e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 5990
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.75,
      "epoch": 0.28947622729029765,
      "grad_norm": 0.0,
      "learning_rate": 3.552860456126788e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 5991
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 95.875,
      "epoch": 0.2895245458059528,
      "grad_norm": 13.353823727811333,
      "learning_rate": 3.5526188635485114e-07,
      "loss": -0.0,
      "reward": 2.03125,
      "reward_std": 0.3061639666557312,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 5992
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.125,
      "epoch": 0.28957286432160806,
      "grad_norm": 0.0,
      "learning_rate": 3.5523772709702356e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5993
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.46875,
      "epoch": 0.28962118283726324,
      "grad_norm": 0.0,
      "learning_rate": 3.5521356783919597e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5994
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.75,
      "epoch": 0.2896695013529184,
      "grad_norm": 0.0,
      "learning_rate": 3.551894085813684e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5995
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.25,
      "epoch": 0.28971781986857364,
      "grad_norm": 0.0,
      "learning_rate": 3.5516524932354075e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 5996
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.78125,
      "epoch": 0.2897661383842288,
      "grad_norm": 0.0,
      "learning_rate": 3.551410900657132e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 5997
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.03125,
      "epoch": 0.28981445689988405,
      "grad_norm": 0.0,
      "learning_rate": 3.5511693080788557e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 5998
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.65625,
      "epoch": 0.28986277541553923,
      "grad_norm": 0.0,
      "learning_rate": 3.5509277155005793e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 5999
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.75,
      "epoch": 0.2899110939311944,
      "grad_norm": 0.0,
      "learning_rate": 3.5506861229223035e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6000
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.09375,
      "epoch": 0.28995941244684964,
      "grad_norm": 0.0,
      "learning_rate": 3.5504445303440276e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6001
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.96875,
      "epoch": 0.2900077309625048,
      "grad_norm": 0.0,
      "learning_rate": 3.550202937765752e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6002
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.0625,
      "epoch": 0.29005604947816005,
      "grad_norm": 0.0,
      "learning_rate": 3.5499613451874754e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6003
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.625,
      "epoch": 0.2901043679938152,
      "grad_norm": 0.0,
      "learning_rate": 3.5497197526092e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6004
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.34375,
      "epoch": 0.2901526865094704,
      "grad_norm": 0.0,
      "learning_rate": 3.5494781600309236e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6005
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.3125,
      "epoch": 0.29020100502512564,
      "grad_norm": 0.0,
      "learning_rate": 3.549236567452648e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6006
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.71875,
      "epoch": 0.2902493235407808,
      "grad_norm": 0.0,
      "learning_rate": 3.548994974874372e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6007
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.53125,
      "epoch": 0.29029764205643604,
      "grad_norm": 7.942956398669668,
      "learning_rate": 3.548753382296096e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6008
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.4375,
      "epoch": 0.2903459605720912,
      "grad_norm": 0.0,
      "learning_rate": 3.5485117897178196e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6009
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.625,
      "epoch": 0.2903942790877464,
      "grad_norm": 0.0,
      "learning_rate": 3.548270197139543e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6010
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.25,
      "epoch": 0.29044259760340163,
      "grad_norm": 0.0,
      "learning_rate": 3.548028604561268e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6011
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.5625,
      "epoch": 0.2904909161190568,
      "grad_norm": 0.0,
      "learning_rate": 3.5477870119829915e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6012
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.96875,
      "epoch": 0.29053923463471204,
      "grad_norm": 0.0,
      "learning_rate": 3.5475454194047157e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6013
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.3125,
      "epoch": 0.2905875531503672,
      "grad_norm": 0.0,
      "learning_rate": 3.54730382682644e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6014
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.75,
      "epoch": 0.2906358716660224,
      "grad_norm": 0.0,
      "learning_rate": 3.547062234248164e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6015
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.0,
      "epoch": 0.2906841901816776,
      "grad_norm": 0.0,
      "learning_rate": 3.5468206416698875e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6016
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 113.90625,
      "epoch": 0.2907325086973328,
      "grad_norm": 0.0,
      "learning_rate": 3.546579049091612e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 6017
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.59375,
      "epoch": 0.29078082721298804,
      "grad_norm": 0.0,
      "learning_rate": 3.546337456513336e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6018
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.6875,
      "epoch": 0.2908291457286432,
      "grad_norm": 0.0,
      "learning_rate": 3.5460958639350594e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6019
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.0625,
      "epoch": 0.2908774642442984,
      "grad_norm": 0.0,
      "learning_rate": 3.5458542713567836e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6020
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.125,
      "epoch": 0.2909257827599536,
      "grad_norm": 0.0,
      "learning_rate": 3.5456126787785077e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6021
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.25,
      "epoch": 0.2909741012756088,
      "grad_norm": 0.0,
      "learning_rate": 3.545371086200232e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6022
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.1875,
      "epoch": 0.29102241979126403,
      "grad_norm": 0.0,
      "learning_rate": 3.5451294936219554e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6023
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.375,
      "epoch": 0.2910707383069192,
      "grad_norm": 0.0,
      "learning_rate": 3.54488790104368e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6024
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.28125,
      "epoch": 0.2911190568225744,
      "grad_norm": 0.0,
      "learning_rate": 3.5446463084654037e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6025
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.96875,
      "epoch": 0.2911673753382296,
      "grad_norm": 0.0,
      "learning_rate": 3.544404715887128e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6026
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.34375,
      "epoch": 0.2912156938538848,
      "grad_norm": 0.0,
      "learning_rate": 3.544163123308852e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6027
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.78125,
      "epoch": 0.29126401236954,
      "grad_norm": 0.0,
      "learning_rate": 3.543921530730576e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6028
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.375,
      "epoch": 0.2913123308851952,
      "grad_norm": 0.0,
      "learning_rate": 3.5436799381522997e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6029
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.5,
      "epoch": 0.2913606494008504,
      "grad_norm": 0.0,
      "learning_rate": 3.5434383455740233e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6030
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.9375,
      "epoch": 0.2914089679165056,
      "grad_norm": 0.0,
      "learning_rate": 3.543196752995748e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6031
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.4375,
      "epoch": 0.2914572864321608,
      "grad_norm": 0.0,
      "learning_rate": 3.5429551604174716e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6032
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.84375,
      "epoch": 0.291505604947816,
      "grad_norm": 0.0,
      "learning_rate": 3.5427135678391957e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6033
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.0,
      "epoch": 0.2915539234634712,
      "grad_norm": 0.0,
      "learning_rate": 3.54247197526092e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6034
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.25,
      "epoch": 0.2916022419791264,
      "grad_norm": 0.0,
      "learning_rate": 3.542230382682644e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6035
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.59375,
      "epoch": 0.2916505604947816,
      "grad_norm": 0.0,
      "learning_rate": 3.5419887901043676e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6036
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 103.9375,
      "epoch": 0.2916988790104368,
      "grad_norm": 0.0,
      "learning_rate": 3.5417471975260923e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 6037
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.6875,
      "epoch": 0.291747197526092,
      "grad_norm": 0.0,
      "learning_rate": 3.541505604947816e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6038
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.96875,
      "epoch": 0.2917955160417472,
      "grad_norm": 0.0,
      "learning_rate": 3.5412640123695395e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6039
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.5625,
      "epoch": 0.29184383455740237,
      "grad_norm": 7.479797156232251,
      "learning_rate": 3.5410224197912636e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6040
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.5625,
      "epoch": 0.2918921530730576,
      "grad_norm": 0.0,
      "learning_rate": 3.540780827212988e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6041
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.4375,
      "epoch": 0.2919404715887128,
      "grad_norm": 0.0,
      "learning_rate": 3.540539234634712e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6042
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.0625,
      "epoch": 0.291988790104368,
      "grad_norm": 0.0,
      "learning_rate": 3.5402976420564355e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6043
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.375,
      "epoch": 0.2920371086200232,
      "grad_norm": 0.0,
      "learning_rate": 3.54005604947816e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6044
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.78125,
      "epoch": 0.29208542713567837,
      "grad_norm": 0.0,
      "learning_rate": 3.539814456899884e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6045
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.5,
      "epoch": 0.2921337456513336,
      "grad_norm": 0.0,
      "learning_rate": 3.539572864321608e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6046
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.0625,
      "epoch": 0.2921820641669888,
      "grad_norm": 0.0,
      "learning_rate": 3.539331271743332e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6047
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.8125,
      "epoch": 0.292230382682644,
      "grad_norm": 0.0,
      "learning_rate": 3.539089679165056e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6048
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.46875,
      "epoch": 0.2922787011982992,
      "grad_norm": 0.0,
      "learning_rate": 3.53884808658678e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6049
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.4375,
      "epoch": 0.29232701971395436,
      "grad_norm": 0.0,
      "learning_rate": 3.5386064940085034e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6050
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.53125,
      "epoch": 0.2923753382296096,
      "grad_norm": 0.0,
      "learning_rate": 3.538364901430228e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6051
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.625,
      "epoch": 0.29242365674526477,
      "grad_norm": 0.0,
      "learning_rate": 3.5381233088519517e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6052
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.34375,
      "epoch": 0.29247197526092,
      "grad_norm": 0.0,
      "learning_rate": 3.537881716273676e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6053
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.21875,
      "epoch": 0.2925202937765752,
      "grad_norm": 0.0,
      "learning_rate": 3.5376401236954e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6054
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.125,
      "epoch": 0.29256861229223036,
      "grad_norm": 0.0,
      "learning_rate": 3.537398531117124e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6055
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.46875,
      "epoch": 0.2926169308078856,
      "grad_norm": 0.0,
      "learning_rate": 3.5371569385388477e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6056
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.4375,
      "epoch": 0.29266524932354077,
      "grad_norm": 0.0,
      "learning_rate": 3.5369153459605724e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6057
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.78125,
      "epoch": 0.292713567839196,
      "grad_norm": 0.0,
      "learning_rate": 3.536673753382296e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6058
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.8125,
      "epoch": 0.2927618863548512,
      "grad_norm": 0.0,
      "learning_rate": 3.53643216080402e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6059
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.875,
      "epoch": 0.29281020487050635,
      "grad_norm": 0.0,
      "learning_rate": 3.5361905682257437e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6060
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.125,
      "epoch": 0.2928585233861616,
      "grad_norm": 0.0,
      "learning_rate": 3.535948975647468e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6061
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.59375,
      "epoch": 0.29290684190181676,
      "grad_norm": 0.0,
      "learning_rate": 3.535707383069192e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6062
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.84375,
      "epoch": 0.292955160417472,
      "grad_norm": 0.0,
      "learning_rate": 3.5354657904909156e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6063
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.375,
      "epoch": 0.29300347893312717,
      "grad_norm": 0.0,
      "learning_rate": 3.53522419791264e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6064
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.84375,
      "epoch": 0.29305179744878235,
      "grad_norm": 0.0,
      "learning_rate": 3.534982605334364e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 6065
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.21875,
      "epoch": 0.2931001159644376,
      "grad_norm": 0.0,
      "learning_rate": 3.534741012756088e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6066
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.5,
      "epoch": 0.29314843448009276,
      "grad_norm": 0.0,
      "learning_rate": 3.534499420177812e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6067
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.5625,
      "epoch": 0.293196752995748,
      "grad_norm": 0.0,
      "learning_rate": 3.5342578275995363e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6068
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.65625,
      "epoch": 0.29324507151140317,
      "grad_norm": 0.0,
      "learning_rate": 3.53401623502126e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 6069
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.375,
      "epoch": 0.29329339002705834,
      "grad_norm": 0.0,
      "learning_rate": 3.5337746424429835e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6070
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.1875,
      "epoch": 0.2933417085427136,
      "grad_norm": 0.0,
      "learning_rate": 3.533533049864708e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6071
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.34375,
      "epoch": 0.29339002705836875,
      "grad_norm": 0.0,
      "learning_rate": 3.533291457286432e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6072
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.71875,
      "epoch": 0.293438345574024,
      "grad_norm": 0.0,
      "learning_rate": 3.533049864708156e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6073
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 97.09375,
      "epoch": 0.29348666408967916,
      "grad_norm": 0.0,
      "learning_rate": 3.53280827212988e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 6074
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.1875,
      "epoch": 0.29353498260533434,
      "grad_norm": 0.0,
      "learning_rate": 3.532566679551604e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6075
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.09375,
      "epoch": 0.29358330112098957,
      "grad_norm": 0.0,
      "learning_rate": 3.532325086973328e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6076
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.09375,
      "epoch": 0.29363161963664475,
      "grad_norm": 0.0,
      "learning_rate": 3.5320834943950524e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6077
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.8125,
      "epoch": 0.2936799381523,
      "grad_norm": 0.0,
      "learning_rate": 3.531841901816776e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6078
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.25,
      "epoch": 0.29372825666795516,
      "grad_norm": 0.0,
      "learning_rate": 3.5316003092385e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 6079
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.0625,
      "epoch": 0.29377657518361033,
      "grad_norm": 0.0,
      "learning_rate": 3.531358716660224e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6080
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.65625,
      "epoch": 0.29382489369926557,
      "grad_norm": 0.0,
      "learning_rate": 3.531117124081948e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6081
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.9375,
      "epoch": 0.29387321221492074,
      "grad_norm": 0.0,
      "learning_rate": 3.530875531503672e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6082
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.40625,
      "epoch": 0.293921530730576,
      "grad_norm": 0.0,
      "learning_rate": 3.5306339389253957e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6083
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.8125,
      "epoch": 0.29396984924623115,
      "grad_norm": 0.0,
      "learning_rate": 3.5303923463471203e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6084
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 91.96875,
      "epoch": 0.29401816776188633,
      "grad_norm": 4.066117666609117,
      "learning_rate": 3.530150753768844e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6085
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.375,
      "epoch": 0.29406648627754156,
      "grad_norm": 0.0,
      "learning_rate": 3.529909161190568e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6086
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.53125,
      "epoch": 0.29411480479319674,
      "grad_norm": 0.0,
      "learning_rate": 3.529667568612292e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6087
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.9375,
      "epoch": 0.29416312330885197,
      "grad_norm": 0.0,
      "learning_rate": 3.5294259760340163e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6088
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.84375,
      "epoch": 0.29421144182450715,
      "grad_norm": 0.0,
      "learning_rate": 3.52918438345574e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6089
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.28125,
      "epoch": 0.2942597603401623,
      "grad_norm": 0.0,
      "learning_rate": 3.528942790877464e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6090
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.25,
      "epoch": 0.29430807885581756,
      "grad_norm": 0.0,
      "learning_rate": 3.528701198299188e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6091
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.125,
      "epoch": 0.29435639737147273,
      "grad_norm": 0.0,
      "learning_rate": 3.528459605720912e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 6092
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.84375,
      "epoch": 0.29440471588712797,
      "grad_norm": 0.0,
      "learning_rate": 3.528218013142636e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6093
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.84375,
      "epoch": 0.29445303440278314,
      "grad_norm": 0.0,
      "learning_rate": 3.52797642056436e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6094
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.5625,
      "epoch": 0.2945013529184383,
      "grad_norm": 0.0,
      "learning_rate": 3.527734827986084e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6095
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.53125,
      "epoch": 0.29454967143409355,
      "grad_norm": 0.0,
      "learning_rate": 3.527493235407808e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6096
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.75,
      "epoch": 0.29459798994974873,
      "grad_norm": 0.0,
      "learning_rate": 3.5272516428295325e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6097
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.875,
      "epoch": 0.29464630846540396,
      "grad_norm": 0.0,
      "learning_rate": 3.527010050251256e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6098
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.53125,
      "epoch": 0.29469462698105914,
      "grad_norm": 0.0,
      "learning_rate": 3.52676845767298e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 6099
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.9375,
      "epoch": 0.2947429454967143,
      "grad_norm": 0.0,
      "learning_rate": 3.526526865094704e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6100
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.0625,
      "epoch": 0.29479126401236955,
      "grad_norm": 0.0,
      "learning_rate": 3.5262852725164285e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6101
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.40625,
      "epoch": 0.2948395825280247,
      "grad_norm": 0.0,
      "learning_rate": 3.526043679938152e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6102
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.90625,
      "epoch": 0.29488790104367996,
      "grad_norm": 0.0,
      "learning_rate": 3.525802087359876e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6103
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.0625,
      "epoch": 0.29493621955933513,
      "grad_norm": 0.0,
      "learning_rate": 3.5255604947816004e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6104
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.78125,
      "epoch": 0.2949845380749903,
      "grad_norm": 0.0,
      "learning_rate": 3.525318902203324e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6105
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.78125,
      "epoch": 0.29503285659064554,
      "grad_norm": 0.0,
      "learning_rate": 3.525077309625048e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6106
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.34375,
      "epoch": 0.2950811751063007,
      "grad_norm": 0.0,
      "learning_rate": 3.5248357170467723e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6107
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.71875,
      "epoch": 0.29512949362195595,
      "grad_norm": 0.0,
      "learning_rate": 3.5245941244684964e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6108
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.40625,
      "epoch": 0.29517781213761113,
      "grad_norm": 0.0,
      "learning_rate": 3.52435253189022e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6109
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 98.9375,
      "epoch": 0.2952261306532663,
      "grad_norm": 21.137169297911957,
      "learning_rate": 3.524110939311944e-07,
      "loss": -0.0,
      "reward": 1.734375,
      "reward_std": 0.3051002025604248,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6110
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.0625,
      "epoch": 0.29527444916892154,
      "grad_norm": 0.0,
      "learning_rate": 3.5238693467336683e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6111
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 111.09375,
      "epoch": 0.2953227676845767,
      "grad_norm": 14.978886012890975,
      "learning_rate": 3.523627754155392e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.21366669237613678,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6112
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.09375,
      "epoch": 0.29537108620023195,
      "grad_norm": 0.0,
      "learning_rate": 3.523386161577116e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6113
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.625,
      "epoch": 0.2954194047158871,
      "grad_norm": 0.0,
      "learning_rate": 3.52314456899884e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6114
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.625,
      "epoch": 0.2954677232315423,
      "grad_norm": 0.0,
      "learning_rate": 3.5229029764205643e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6115
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.5625,
      "epoch": 0.29551604174719753,
      "grad_norm": 0.0,
      "learning_rate": 3.522661383842288e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6116
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.125,
      "epoch": 0.2955643602628527,
      "grad_norm": 0.0,
      "learning_rate": 3.5224197912640126e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6117
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.6875,
      "epoch": 0.29561267877850794,
      "grad_norm": 0.0,
      "learning_rate": 3.522178198685736e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6118
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.15625,
      "epoch": 0.2956609972941631,
      "grad_norm": 0.0,
      "learning_rate": 3.5219366061074603e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6119
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.90625,
      "epoch": 0.2957093158098183,
      "grad_norm": 0.0,
      "learning_rate": 3.521695013529184e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6120
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.1875,
      "epoch": 0.29575763432547353,
      "grad_norm": 0.0,
      "learning_rate": 3.5214534209509086e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6121
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.5,
      "epoch": 0.2958059528411287,
      "grad_norm": 0.0,
      "learning_rate": 3.521211828372632e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 6122
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.375,
      "epoch": 0.29585427135678394,
      "grad_norm": 0.0,
      "learning_rate": 3.520970235794356e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6123
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.78125,
      "epoch": 0.2959025898724391,
      "grad_norm": 0.0,
      "learning_rate": 3.5207286432160805e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6124
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.09375,
      "epoch": 0.2959509083880943,
      "grad_norm": 0.0,
      "learning_rate": 3.520487050637804e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6125
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.65625,
      "epoch": 0.2959992269037495,
      "grad_norm": 0.0,
      "learning_rate": 3.520245458059528e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6126
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.53125,
      "epoch": 0.2960475454194047,
      "grad_norm": 0.0,
      "learning_rate": 3.5200038654812524e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6127
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.125,
      "epoch": 0.29609586393505993,
      "grad_norm": 0.0,
      "learning_rate": 3.5197622729029765e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6128
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.0625,
      "epoch": 0.2961441824507151,
      "grad_norm": 0.0,
      "learning_rate": 3.5195206803247e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6129
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.8125,
      "epoch": 0.2961925009663703,
      "grad_norm": 0.0,
      "learning_rate": 3.519279087746424e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6130
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.46875,
      "epoch": 0.2962408194820255,
      "grad_norm": 0.0,
      "learning_rate": 3.5190374951681484e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6131
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.65625,
      "epoch": 0.2962891379976807,
      "grad_norm": 0.0,
      "learning_rate": 3.518795902589872e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6132
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.0625,
      "epoch": 0.29633745651333593,
      "grad_norm": 0.0,
      "learning_rate": 3.518554310011596e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6133
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 100.75,
      "epoch": 0.2963857750289911,
      "grad_norm": 16.363615963545595,
      "learning_rate": 3.51831271743332e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 6134
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.9375,
      "epoch": 0.2964340935446463,
      "grad_norm": 0.0,
      "learning_rate": 3.5180711248550444e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6135
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.15625,
      "epoch": 0.2964824120603015,
      "grad_norm": 0.0,
      "learning_rate": 3.517829532276768e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6136
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.3125,
      "epoch": 0.2965307305759567,
      "grad_norm": 0.0,
      "learning_rate": 3.5175879396984927e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6137
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 108.59375,
      "epoch": 0.2965790490916119,
      "grad_norm": 17.677505387787384,
      "learning_rate": 3.5173463471202163e-07,
      "loss": 0.0,
      "reward": 1.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.03125,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 6138
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.9375,
      "epoch": 0.2966273676072671,
      "grad_norm": 0.0,
      "learning_rate": 3.5171047545419404e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6139
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.375,
      "epoch": 0.2966756861229223,
      "grad_norm": 30.31230329125376,
      "learning_rate": 3.516863161963664e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6140
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.25,
      "epoch": 0.2967240046385775,
      "grad_norm": 0.0,
      "learning_rate": 3.5166215693853887e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6141
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.78125,
      "epoch": 0.2967723231542327,
      "grad_norm": 0.0,
      "learning_rate": 3.5163799768071123e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6142
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 108.8125,
      "epoch": 0.2968206416698879,
      "grad_norm": 24.012560691171917,
      "learning_rate": 3.516138384228836e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.2799758315086365,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6143
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.25,
      "epoch": 0.2968689601855431,
      "grad_norm": 0.0,
      "learning_rate": 3.5158967916505606e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6144
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.8125,
      "epoch": 0.2969172787011983,
      "grad_norm": 0.0,
      "learning_rate": 3.515655199072284e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6145
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 111.5,
      "epoch": 0.2969655972168535,
      "grad_norm": 0.0,
      "learning_rate": 3.5154136064940083e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 6146
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.125,
      "epoch": 0.2970139157325087,
      "grad_norm": 0.0,
      "learning_rate": 3.5151720139157324e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6147
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.5,
      "epoch": 0.2970622342481639,
      "grad_norm": 0.0,
      "learning_rate": 3.5149304213374566e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6148
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.21875,
      "epoch": 0.2971105527638191,
      "grad_norm": 0.0,
      "learning_rate": 3.51468882875918e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6149
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.125,
      "epoch": 0.29715887127947427,
      "grad_norm": 0.0,
      "learning_rate": 3.5144472361809043e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6150
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.46875,
      "epoch": 0.2972071897951295,
      "grad_norm": 0.0,
      "learning_rate": 3.5142056436026285e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6151
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.5625,
      "epoch": 0.2972555083107847,
      "grad_norm": 0.0,
      "learning_rate": 3.5139640510243526e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6152
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.84375,
      "epoch": 0.2973038268264399,
      "grad_norm": 0.0,
      "learning_rate": 3.513722458446076e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6153
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.09375,
      "epoch": 0.2973521453420951,
      "grad_norm": 0.0,
      "learning_rate": 3.5134808658678003e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6154
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.375,
      "epoch": 0.29740046385775026,
      "grad_norm": 0.0,
      "learning_rate": 3.5132392732895245e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6155
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 116.90625,
      "epoch": 0.2974487823734055,
      "grad_norm": 0.0,
      "learning_rate": 3.512997680711248e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6156
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 110.84375,
      "epoch": 0.2974971008890607,
      "grad_norm": 0.0,
      "learning_rate": 3.512756088132973e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 6157
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.34375,
      "epoch": 0.2975454194047159,
      "grad_norm": 0.0,
      "learning_rate": 3.5125144955546964e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6158
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.6875,
      "epoch": 0.2975937379203711,
      "grad_norm": 0.0,
      "learning_rate": 3.5122729029764205e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6159
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.65625,
      "epoch": 0.29764205643602626,
      "grad_norm": 0.0,
      "learning_rate": 3.512031310398144e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6160
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.8125,
      "epoch": 0.2976903749516815,
      "grad_norm": 0.0,
      "learning_rate": 3.511789717819869e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 6161
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.5625,
      "epoch": 0.29773869346733667,
      "grad_norm": 0.0,
      "learning_rate": 3.5115481252415924e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6162
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.96875,
      "epoch": 0.2977870119829919,
      "grad_norm": 0.0,
      "learning_rate": 3.511306532663316e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6163
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.0,
      "epoch": 0.2978353304986471,
      "grad_norm": 0.0,
      "learning_rate": 3.5110649400850406e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6164
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 97.75,
      "epoch": 0.29788364901430225,
      "grad_norm": 11.087885524116437,
      "learning_rate": 3.510823347506764e-07,
      "loss": -0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 6165
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 89.0625,
      "epoch": 0.2979319675299575,
      "grad_norm": 6.503090642252366,
      "learning_rate": 3.5105817549284884e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6166
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.0,
      "epoch": 0.29798028604561266,
      "grad_norm": 0.0,
      "learning_rate": 3.5103401623502125e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6167
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.53125,
      "epoch": 0.2980286045612679,
      "grad_norm": 0.0,
      "learning_rate": 3.5100985697719367e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6168
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 92.65625,
      "epoch": 0.2980769230769231,
      "grad_norm": 6.754946127412035,
      "learning_rate": 3.5098569771936603e-07,
      "loss": -0.0,
      "reward": 1.328125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 6169
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.40625,
      "epoch": 0.29812524159257825,
      "grad_norm": 0.0,
      "learning_rate": 3.5096153846153844e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6170
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 113.59375,
      "epoch": 0.2981735601082335,
      "grad_norm": 9.902922452190845,
      "learning_rate": 3.5093737920371085e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6171
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.125,
      "epoch": 0.29822187862388866,
      "grad_norm": 0.0,
      "learning_rate": 3.5091321994588327e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6172
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.5,
      "epoch": 0.2982701971395439,
      "grad_norm": 0.0,
      "learning_rate": 3.5088906068805563e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6173
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.0625,
      "epoch": 0.29831851565519907,
      "grad_norm": 0.0,
      "learning_rate": 3.5086490143022804e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6174
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.46875,
      "epoch": 0.29836683417085424,
      "grad_norm": 0.0,
      "learning_rate": 3.5084074217240046e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6175
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.96875,
      "epoch": 0.2984151526865095,
      "grad_norm": 0.0,
      "learning_rate": 3.508165829145728e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6176
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.90625,
      "epoch": 0.29846347120216465,
      "grad_norm": 0.0,
      "learning_rate": 3.507924236567453e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6177
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.0625,
      "epoch": 0.2985117897178199,
      "grad_norm": 0.0,
      "learning_rate": 3.5076826439891764e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6178
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.46875,
      "epoch": 0.29856010823347506,
      "grad_norm": 0.0,
      "learning_rate": 3.5074410514109006e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6179
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.53125,
      "epoch": 0.29860842674913024,
      "grad_norm": 0.0,
      "learning_rate": 3.507199458832624e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6180
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.375,
      "epoch": 0.2986567452647855,
      "grad_norm": 0.0,
      "learning_rate": 3.506957866254349e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6181
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.09375,
      "epoch": 0.29870506378044065,
      "grad_norm": 0.0,
      "learning_rate": 3.5067162736760725e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6182
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.6875,
      "epoch": 0.2987533822960959,
      "grad_norm": 0.0,
      "learning_rate": 3.506474681097796e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6183
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.625,
      "epoch": 0.29880170081175106,
      "grad_norm": 0.0,
      "learning_rate": 3.5062330885195207e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6184
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.5,
      "epoch": 0.29885001932740624,
      "grad_norm": 0.0,
      "learning_rate": 3.5059914959412443e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6185
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.4375,
      "epoch": 0.29889833784306147,
      "grad_norm": 0.0,
      "learning_rate": 3.5057499033629685e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6186
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.09375,
      "epoch": 0.29894665635871664,
      "grad_norm": 0.0,
      "learning_rate": 3.5055083107846926e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6187
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.75,
      "epoch": 0.2989949748743719,
      "grad_norm": 0.0,
      "learning_rate": 3.505266718206417e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6188
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 116.59375,
      "epoch": 0.29904329339002705,
      "grad_norm": 0.0,
      "learning_rate": 3.5050251256281403e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6189
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.53125,
      "epoch": 0.29909161190568223,
      "grad_norm": 0.0,
      "learning_rate": 3.5047835330498645e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6190
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.3125,
      "epoch": 0.29913993042133746,
      "grad_norm": 0.0,
      "learning_rate": 3.5045419404715886e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6191
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.03125,
      "epoch": 0.29918824893699264,
      "grad_norm": 0.0,
      "learning_rate": 3.504300347893313e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6192
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.65625,
      "epoch": 0.2992365674526479,
      "grad_norm": 0.0,
      "learning_rate": 3.5040587553150364e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6193
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 134.34375,
      "epoch": 0.29928488596830305,
      "grad_norm": 0.0,
      "learning_rate": 3.5038171627367605e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6194
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.40625,
      "epoch": 0.2993332044839582,
      "grad_norm": 0.0,
      "learning_rate": 3.5035755701584846e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6195
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 124.125,
      "epoch": 0.29938152299961346,
      "grad_norm": 0.0,
      "learning_rate": 3.503333977580208e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6196
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.3125,
      "epoch": 0.29942984151526864,
      "grad_norm": 0.0,
      "learning_rate": 3.503092385001933e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6197
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 111.5625,
      "epoch": 0.29947816003092387,
      "grad_norm": 18.19281166299649,
      "learning_rate": 3.5028507924236565e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.34564992785453796,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6198
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.59375,
      "epoch": 0.29952647854657904,
      "grad_norm": 0.0,
      "learning_rate": 3.5026091998453807e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6199
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 117.78125,
      "epoch": 0.2995747970622342,
      "grad_norm": 0.0,
      "learning_rate": 3.502367607267104e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6200
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 122.5625,
      "epoch": 0.29962311557788945,
      "grad_norm": 0.0,
      "learning_rate": 3.502126014688829e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6201
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.0,
      "epoch": 0.29967143409354463,
      "grad_norm": 0.0,
      "learning_rate": 3.5018844221105525e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6202
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 119.46875,
      "epoch": 0.29971975260919986,
      "grad_norm": 38.264115236532376,
      "learning_rate": 3.5016428295322767e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6203
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.90625,
      "epoch": 0.29976807112485504,
      "grad_norm": 0.0,
      "learning_rate": 3.501401236954001e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6204
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.375,
      "epoch": 0.2998163896405102,
      "grad_norm": 0.0,
      "learning_rate": 3.5011596443757244e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6205
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.0,
      "epoch": 0.29986470815616545,
      "grad_norm": 0.0,
      "learning_rate": 3.5009180517974485e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6206
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.71875,
      "epoch": 0.2999130266718206,
      "grad_norm": 0.0,
      "learning_rate": 3.5006764592191727e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6207
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.0,
      "epoch": 0.29996134518747586,
      "grad_norm": 0.0,
      "learning_rate": 3.500434866640897e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6208
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.4375,
      "epoch": 0.30000966370313104,
      "grad_norm": 0.0,
      "learning_rate": 3.5001932740626204e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6209
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.09375,
      "epoch": 0.3000579822187862,
      "grad_norm": 0.0,
      "learning_rate": 3.4999516814843446e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6210
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.625,
      "epoch": 0.30010630073444144,
      "grad_norm": 0.0,
      "learning_rate": 3.4997100889060687e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6211
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.65625,
      "epoch": 0.3001546192500966,
      "grad_norm": 0.0,
      "learning_rate": 3.499468496327793e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6212
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 124.09375,
      "epoch": 0.30020293776575185,
      "grad_norm": 0.0,
      "learning_rate": 3.4992269037495164e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6213
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.96875,
      "epoch": 0.30025125628140703,
      "grad_norm": 0.0,
      "learning_rate": 3.498985311171241e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6214
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.46875,
      "epoch": 0.3002995747970622,
      "grad_norm": 0.0,
      "learning_rate": 3.4987437185929647e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6215
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.6875,
      "epoch": 0.30034789331271744,
      "grad_norm": 0.0,
      "learning_rate": 3.4985021260146883e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6216
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.90625,
      "epoch": 0.3003962118283726,
      "grad_norm": 0.0,
      "learning_rate": 3.498260533436413e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6217
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.28125,
      "epoch": 0.30044453034402785,
      "grad_norm": 0.0,
      "learning_rate": 3.4980189408581366e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6218
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 126.59375,
      "epoch": 0.300492848859683,
      "grad_norm": 0.0,
      "learning_rate": 3.4977773482798607e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 6219
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.125,
      "epoch": 0.3005411673753382,
      "grad_norm": 0.0,
      "learning_rate": 3.4975357557015843e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6220
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 120.75,
      "epoch": 0.30058948589099344,
      "grad_norm": 0.0,
      "learning_rate": 3.497294163123309e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 6221
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.3125,
      "epoch": 0.3006378044066486,
      "grad_norm": 0.0,
      "learning_rate": 3.4970525705450326e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6222
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.5625,
      "epoch": 0.30068612292230384,
      "grad_norm": 0.0,
      "learning_rate": 3.496810977966757e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6223
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.5,
      "epoch": 0.300734441437959,
      "grad_norm": 0.0,
      "learning_rate": 3.496569385388481e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 6224
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.125,
      "epoch": 0.3007827599536142,
      "grad_norm": 0.0,
      "learning_rate": 3.4963277928102045e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6225
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 118.34375,
      "epoch": 0.30083107846926943,
      "grad_norm": 0.0,
      "learning_rate": 3.4960862002319286e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6226
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.375,
      "epoch": 0.3008793969849246,
      "grad_norm": 0.0,
      "learning_rate": 3.495844607653653e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6227
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.625,
      "epoch": 0.30092771550057984,
      "grad_norm": 0.0,
      "learning_rate": 3.495603015075377e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6228
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 109.21875,
      "epoch": 0.300976034016235,
      "grad_norm": 0.0,
      "learning_rate": 3.4953614224971005e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 6229
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.15625,
      "epoch": 0.3010243525318902,
      "grad_norm": 0.0,
      "learning_rate": 3.4951198299188246e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6230
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.84375,
      "epoch": 0.3010726710475454,
      "grad_norm": 0.0,
      "learning_rate": 3.494878237340549e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6231
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 108.78125,
      "epoch": 0.3011209895632006,
      "grad_norm": 4.403989699366016,
      "learning_rate": 3.494636644762273e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 6232
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.03125,
      "epoch": 0.30116930807885584,
      "grad_norm": 0.0,
      "learning_rate": 3.4943950521839965e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6233
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.4375,
      "epoch": 0.301217626594511,
      "grad_norm": 0.0,
      "learning_rate": 3.494153459605721e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6234
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.3125,
      "epoch": 0.3012659451101662,
      "grad_norm": 0.0,
      "learning_rate": 3.493911867027445e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6235
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.6875,
      "epoch": 0.3013142636258214,
      "grad_norm": 0.0,
      "learning_rate": 3.4936702744491684e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6236
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.8125,
      "epoch": 0.3013625821414766,
      "grad_norm": 0.0,
      "learning_rate": 3.493428681870893e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6237
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.90625,
      "epoch": 0.30141090065713183,
      "grad_norm": 0.0,
      "learning_rate": 3.4931870892926167e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6238
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.625,
      "epoch": 0.301459219172787,
      "grad_norm": 0.0,
      "learning_rate": 3.492945496714341e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6239
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.21875,
      "epoch": 0.3015075376884422,
      "grad_norm": 0.0,
      "learning_rate": 3.4927039041360644e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6240
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.375,
      "epoch": 0.3015558562040974,
      "grad_norm": 0.0,
      "learning_rate": 3.492462311557789e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6241
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.5,
      "epoch": 0.3016041747197526,
      "grad_norm": 0.0,
      "learning_rate": 3.4922207189795127e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6242
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.75,
      "epoch": 0.3016524932354078,
      "grad_norm": 0.0,
      "learning_rate": 3.491979126401237e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6243
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.375,
      "epoch": 0.301700811751063,
      "grad_norm": 0.0,
      "learning_rate": 3.491737533822961e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6244
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 108.53125,
      "epoch": 0.3017491302667182,
      "grad_norm": 58.31167331512515,
      "learning_rate": 3.4914959412446846e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2961486577987671,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6245
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.53125,
      "epoch": 0.3017974487823734,
      "grad_norm": 0.0,
      "learning_rate": 3.4912543486664087e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6246
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.15625,
      "epoch": 0.3018457672980286,
      "grad_norm": 0.0,
      "learning_rate": 3.491012756088133e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6247
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.40625,
      "epoch": 0.3018940858136838,
      "grad_norm": 0.0,
      "learning_rate": 3.490771163509857e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6248
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.65625,
      "epoch": 0.301942404329339,
      "grad_norm": 0.0,
      "learning_rate": 3.4905295709315806e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6249
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.0,
      "epoch": 0.3019907228449942,
      "grad_norm": 0.0,
      "learning_rate": 3.4902879783533047e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6250
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.6875,
      "epoch": 0.3020390413606494,
      "grad_norm": 0.0,
      "learning_rate": 3.490046385775029e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6251
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.84375,
      "epoch": 0.3020873598763046,
      "grad_norm": 0.0,
      "learning_rate": 3.489804793196753e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6252
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.0625,
      "epoch": 0.3021356783919598,
      "grad_norm": 0.0,
      "learning_rate": 3.4895632006184766e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6253
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.125,
      "epoch": 0.302183996907615,
      "grad_norm": 0.0,
      "learning_rate": 3.4893216080402013e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6254
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.8125,
      "epoch": 0.30223231542327017,
      "grad_norm": 0.0,
      "learning_rate": 3.489080015461925e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6255
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.375,
      "epoch": 0.3022806339389254,
      "grad_norm": 0.0,
      "learning_rate": 3.4888384228836485e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6256
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.21875,
      "epoch": 0.3023289524545806,
      "grad_norm": 0.0,
      "learning_rate": 3.488596830305373e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6257
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.78125,
      "epoch": 0.3023772709702358,
      "grad_norm": 0.0,
      "learning_rate": 3.488355237727097e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6258
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.59375,
      "epoch": 0.302425589485891,
      "grad_norm": 0.0,
      "learning_rate": 3.488113645148821e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6259
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 120.375,
      "epoch": 0.30247390800154617,
      "grad_norm": 0.0,
      "learning_rate": 3.4878720525705445e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 6260
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.09375,
      "epoch": 0.3025222265172014,
      "grad_norm": 0.0,
      "learning_rate": 3.487630459992269e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6261
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.0625,
      "epoch": 0.3025705450328566,
      "grad_norm": 0.0,
      "learning_rate": 3.487388867413993e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6262
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.3125,
      "epoch": 0.3026188635485118,
      "grad_norm": 0.0,
      "learning_rate": 3.487147274835717e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6263
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.71875,
      "epoch": 0.302667182064167,
      "grad_norm": 0.0,
      "learning_rate": 3.486905682257441e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6264
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.3125,
      "epoch": 0.30271550057982216,
      "grad_norm": 0.0,
      "learning_rate": 3.486664089679165e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6265
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 116.28125,
      "epoch": 0.3027638190954774,
      "grad_norm": 0.0,
      "learning_rate": 3.486422497100889e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6266
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 102.21875,
      "epoch": 0.30281213761113257,
      "grad_norm": 11.93485617991847,
      "learning_rate": 3.486180904522613e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.23047873377799988,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6267
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.09375,
      "epoch": 0.3028604561267878,
      "grad_norm": 0.0,
      "learning_rate": 3.485939311944337e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6268
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.9375,
      "epoch": 0.302908774642443,
      "grad_norm": 0.0,
      "learning_rate": 3.4856977193660607e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6269
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.96875,
      "epoch": 0.30295709315809816,
      "grad_norm": 0.0,
      "learning_rate": 3.485456126787785e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6270
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.53125,
      "epoch": 0.3030054116737534,
      "grad_norm": 0.0,
      "learning_rate": 3.485214534209509e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6271
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.90625,
      "epoch": 0.30305373018940857,
      "grad_norm": 0.0,
      "learning_rate": 3.484972941631233e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6272
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 99.0,
      "epoch": 0.3031020487050638,
      "grad_norm": 12.99633936986788,
      "learning_rate": 3.4847313490529567e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 6273
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.0,
      "epoch": 0.303150367220719,
      "grad_norm": 0.0,
      "learning_rate": 3.4844897564746813e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6274
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.96875,
      "epoch": 0.30319868573637415,
      "grad_norm": 0.0,
      "learning_rate": 3.484248163896405e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6275
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.65625,
      "epoch": 0.3032470042520294,
      "grad_norm": 0.0,
      "learning_rate": 3.4840065713181286e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6276
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.84375,
      "epoch": 0.30329532276768456,
      "grad_norm": 0.0,
      "learning_rate": 3.483764978739853e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6277
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 103.1875,
      "epoch": 0.3033436412833398,
      "grad_norm": 15.580873156463989,
      "learning_rate": 3.483523386161577e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.3902835249900818,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6278
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.90625,
      "epoch": 0.30339195979899497,
      "grad_norm": 0.0,
      "learning_rate": 3.483281793583301e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 6279
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.21875,
      "epoch": 0.30344027831465015,
      "grad_norm": 0.0,
      "learning_rate": 3.4830402010050246e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6280
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.0625,
      "epoch": 0.3034885968303054,
      "grad_norm": 0.0,
      "learning_rate": 3.482798608426749e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6281
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.09375,
      "epoch": 0.30353691534596056,
      "grad_norm": 0.0,
      "learning_rate": 3.482557015848473e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6282
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.71875,
      "epoch": 0.3035852338616158,
      "grad_norm": 0.0,
      "learning_rate": 3.482315423270197e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6283
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.375,
      "epoch": 0.30363355237727097,
      "grad_norm": 0.0,
      "learning_rate": 3.482073830691921e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6284
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.65625,
      "epoch": 0.30368187089292614,
      "grad_norm": 0.0,
      "learning_rate": 3.481832238113645e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6285
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.71875,
      "epoch": 0.3037301894085814,
      "grad_norm": 0.0,
      "learning_rate": 3.481590645535369e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6286
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.53125,
      "epoch": 0.30377850792423655,
      "grad_norm": 0.0,
      "learning_rate": 3.481349052957093e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6287
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.1875,
      "epoch": 0.3038268264398918,
      "grad_norm": 0.0,
      "learning_rate": 3.481107460378817e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6288
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.6875,
      "epoch": 0.30387514495554696,
      "grad_norm": 0.0,
      "learning_rate": 3.480865867800541e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 6289
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.40625,
      "epoch": 0.30392346347120214,
      "grad_norm": 0.0,
      "learning_rate": 3.480624275222265e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6290
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.65625,
      "epoch": 0.30397178198685737,
      "grad_norm": 0.0,
      "learning_rate": 3.480382682643989e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6291
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.46875,
      "epoch": 0.30402010050251255,
      "grad_norm": 0.0,
      "learning_rate": 3.480141090065713e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6292
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.09375,
      "epoch": 0.3040684190181678,
      "grad_norm": 0.0,
      "learning_rate": 3.479899497487437e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6293
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.0625,
      "epoch": 0.30411673753382296,
      "grad_norm": 0.0,
      "learning_rate": 3.4796579049091614e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6294
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 127.90625,
      "epoch": 0.30416505604947813,
      "grad_norm": 0.0,
      "learning_rate": 3.479416312330885e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6295
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.84375,
      "epoch": 0.30421337456513337,
      "grad_norm": 0.0,
      "learning_rate": 3.4791747197526086e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6296
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 108.8125,
      "epoch": 0.30426169308078854,
      "grad_norm": 0.0,
      "learning_rate": 3.4789331271743333e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 6297
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.1875,
      "epoch": 0.3043100115964438,
      "grad_norm": 0.0,
      "learning_rate": 3.478691534596057e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6298
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.59375,
      "epoch": 0.30435833011209895,
      "grad_norm": 0.0,
      "learning_rate": 3.478449942017781e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6299
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 111.4375,
      "epoch": 0.30440664862775413,
      "grad_norm": 17.741331034107468,
      "learning_rate": 3.4782083494395047e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22738909721374512,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6300
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.5625,
      "epoch": 0.30445496714340936,
      "grad_norm": 0.0,
      "learning_rate": 3.4779667568612293e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6301
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.375,
      "epoch": 0.30450328565906454,
      "grad_norm": 0.0,
      "learning_rate": 3.477725164282953e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6302
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.0,
      "epoch": 0.30455160417471977,
      "grad_norm": 0.0,
      "learning_rate": 3.477483571704677e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6303
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.25,
      "epoch": 0.30459992269037495,
      "grad_norm": 0.0,
      "learning_rate": 3.477241979126401e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6304
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.78125,
      "epoch": 0.3046482412060301,
      "grad_norm": 0.0,
      "learning_rate": 3.4770003865481253e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6305
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.25,
      "epoch": 0.30469655972168536,
      "grad_norm": 0.0,
      "learning_rate": 3.476758793969849e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6306
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 118.40625,
      "epoch": 0.30474487823734053,
      "grad_norm": 0.0,
      "learning_rate": 3.476517201391573e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 6307
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.40625,
      "epoch": 0.30479319675299577,
      "grad_norm": 0.0,
      "learning_rate": 3.476275608813297e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 6308
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 112.28125,
      "epoch": 0.30484151526865094,
      "grad_norm": 0.0,
      "learning_rate": 3.476034016235021e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6309
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.84375,
      "epoch": 0.3048898337843061,
      "grad_norm": 0.0,
      "learning_rate": 3.475792423656745e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6310
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 122.65625,
      "epoch": 0.30493815229996135,
      "grad_norm": 0.0,
      "learning_rate": 3.475550831078469e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6311
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.96875,
      "epoch": 0.30498647081561653,
      "grad_norm": 0.0,
      "learning_rate": 3.475309238500193e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6312
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.96875,
      "epoch": 0.30503478933127176,
      "grad_norm": 0.0,
      "learning_rate": 3.475067645921917e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6313
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 108.59375,
      "epoch": 0.30508310784692694,
      "grad_norm": 0.0,
      "learning_rate": 3.4748260533436415e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 6314
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.21875,
      "epoch": 0.3051314263625821,
      "grad_norm": 0.0,
      "learning_rate": 3.474584460765365e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6315
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.59375,
      "epoch": 0.30517974487823735,
      "grad_norm": 0.0,
      "learning_rate": 3.474342868187089e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6316
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.21875,
      "epoch": 0.3052280633938925,
      "grad_norm": 0.0,
      "learning_rate": 3.4741012756088134e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6317
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.0625,
      "epoch": 0.30527638190954776,
      "grad_norm": 0.0,
      "learning_rate": 3.473859683030537e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6318
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.09375,
      "epoch": 0.30532470042520293,
      "grad_norm": 0.0,
      "learning_rate": 3.473618090452261e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6319
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.4375,
      "epoch": 0.3053730189408581,
      "grad_norm": 0.0,
      "learning_rate": 3.4733764978739847e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6320
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.40625,
      "epoch": 0.30542133745651334,
      "grad_norm": 0.0,
      "learning_rate": 3.4731349052957094e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6321
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.78125,
      "epoch": 0.3054696559721685,
      "grad_norm": 0.0,
      "learning_rate": 3.472893312717433e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6322
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.78125,
      "epoch": 0.30551797448782375,
      "grad_norm": 0.0,
      "learning_rate": 3.472651720139157e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6323
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.09375,
      "epoch": 0.30556629300347893,
      "grad_norm": 0.0,
      "learning_rate": 3.4724101275608813e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6324
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 125.875,
      "epoch": 0.3056146115191341,
      "grad_norm": 0.0,
      "learning_rate": 3.4721685349826054e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6325
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.65625,
      "epoch": 0.30566293003478934,
      "grad_norm": 0.0,
      "learning_rate": 3.471926942404329e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6326
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.75,
      "epoch": 0.3057112485504445,
      "grad_norm": 0.0,
      "learning_rate": 3.4716853498260537e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6327
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.59375,
      "epoch": 0.30575956706609975,
      "grad_norm": 0.0,
      "learning_rate": 3.4714437572477773e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6328
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.8125,
      "epoch": 0.3058078855817549,
      "grad_norm": 0.0,
      "learning_rate": 3.471202164669501e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6329
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 130.65625,
      "epoch": 0.3058562040974101,
      "grad_norm": 0.0,
      "learning_rate": 3.470960572091225e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 6330
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.1875,
      "epoch": 0.30590452261306533,
      "grad_norm": 0.0,
      "learning_rate": 3.470718979512949e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6331
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.59375,
      "epoch": 0.3059528411287205,
      "grad_norm": 0.0,
      "learning_rate": 3.4704773869346733e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6332
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.4375,
      "epoch": 0.30600115964437574,
      "grad_norm": 0.0,
      "learning_rate": 3.470235794356397e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6333
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 109.5625,
      "epoch": 0.3060494781600309,
      "grad_norm": 0.0,
      "learning_rate": 3.4699942017781216e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6334
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.75,
      "epoch": 0.3060977966756861,
      "grad_norm": 0.0,
      "learning_rate": 3.469752609199845e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6335
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.84375,
      "epoch": 0.30614611519134133,
      "grad_norm": 0.0,
      "learning_rate": 3.4695110166215693e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6336
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.34375,
      "epoch": 0.3061944337069965,
      "grad_norm": 0.0,
      "learning_rate": 3.4692694240432935e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6337
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.375,
      "epoch": 0.30624275222265174,
      "grad_norm": 0.0,
      "learning_rate": 3.469027831465017e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6338
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.75,
      "epoch": 0.3062910707383069,
      "grad_norm": 0.0,
      "learning_rate": 3.468786238886741e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6339
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 127.78125,
      "epoch": 0.3063393892539621,
      "grad_norm": 14.160985383523458,
      "learning_rate": 3.468544646308465e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.2646907567977905,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6340
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.53125,
      "epoch": 0.3063877077696173,
      "grad_norm": 0.0,
      "learning_rate": 3.4683030537301895e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6341
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.65625,
      "epoch": 0.3064360262852725,
      "grad_norm": 0.0,
      "learning_rate": 3.468061461151913e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6342
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.5,
      "epoch": 0.30648434480092773,
      "grad_norm": 0.0,
      "learning_rate": 3.467819868573637e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6343
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 118.75,
      "epoch": 0.3065326633165829,
      "grad_norm": 17.67279947222342,
      "learning_rate": 3.4675782759953614e-07,
      "loss": 0.0,
      "reward": 1.515625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6344
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 120.0,
      "epoch": 0.3065809818322381,
      "grad_norm": 19.333473521402073,
      "learning_rate": 3.4673366834170855e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6345
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.25,
      "epoch": 0.3066293003478933,
      "grad_norm": 0.0,
      "learning_rate": 3.467095090838809e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6346
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 112.78125,
      "epoch": 0.3066776188635485,
      "grad_norm": 0.0,
      "learning_rate": 3.466853498260534e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6347
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.75,
      "epoch": 0.30672593737920373,
      "grad_norm": 0.0,
      "learning_rate": 3.4666119056822574e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6348
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.25,
      "epoch": 0.3067742558948589,
      "grad_norm": 0.0,
      "learning_rate": 3.466370313103981e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6349
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.90625,
      "epoch": 0.3068225744105141,
      "grad_norm": 0.0,
      "learning_rate": 3.466128720525705e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6350
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 112.75,
      "epoch": 0.3068708929261693,
      "grad_norm": 9.020189836824027,
      "learning_rate": 3.465887127947429e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6351
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.1875,
      "epoch": 0.3069192114418245,
      "grad_norm": 0.0,
      "learning_rate": 3.4656455353691534e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6352
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.28125,
      "epoch": 0.3069675299574797,
      "grad_norm": 0.0,
      "learning_rate": 3.465403942790877e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6353
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.5625,
      "epoch": 0.3070158484731349,
      "grad_norm": 0.0,
      "learning_rate": 3.4651623502126017e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6354
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.0,
      "epoch": 0.3070641669887901,
      "grad_norm": 0.0,
      "learning_rate": 3.4649207576343253e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6355
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 105.65625,
      "epoch": 0.3071124855044453,
      "grad_norm": 10.761353526996837,
      "learning_rate": 3.4646791650560494e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6356
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.125,
      "epoch": 0.3071608040201005,
      "grad_norm": 0.0,
      "learning_rate": 3.4644375724777735e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6357
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.40625,
      "epoch": 0.3072091225357557,
      "grad_norm": 0.0,
      "learning_rate": 3.4641959798994977e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6358
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 128.21875,
      "epoch": 0.3072574410514109,
      "grad_norm": 0.0,
      "learning_rate": 3.4639543873212213e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6359
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.0,
      "epoch": 0.3073057595670661,
      "grad_norm": 0.0,
      "learning_rate": 3.463712794742945e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6360
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.34375,
      "epoch": 0.3073540780827213,
      "grad_norm": 0.0,
      "learning_rate": 3.4634712021646696e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6361
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 108.875,
      "epoch": 0.3074023965983765,
      "grad_norm": 20.80958238928901,
      "learning_rate": 3.463229609586393e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2961486577987671,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6362
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.84375,
      "epoch": 0.3074507151140317,
      "grad_norm": 0.0,
      "learning_rate": 3.4629880170081173e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6363
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.09375,
      "epoch": 0.3074990336296869,
      "grad_norm": 0.0,
      "learning_rate": 3.4627464244298414e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6364
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.59375,
      "epoch": 0.30754735214534207,
      "grad_norm": 0.0,
      "learning_rate": 3.4625048318515656e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6365
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.8125,
      "epoch": 0.3075956706609973,
      "grad_norm": 0.0,
      "learning_rate": 3.462263239273289e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6366
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.6875,
      "epoch": 0.3076439891766525,
      "grad_norm": 0.0,
      "learning_rate": 3.462021646695014e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6367
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.5625,
      "epoch": 0.3076923076923077,
      "grad_norm": 0.0,
      "learning_rate": 3.4617800541167374e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6368
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.28125,
      "epoch": 0.3077406262079629,
      "grad_norm": 0.0,
      "learning_rate": 3.461538461538461e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6369
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.90625,
      "epoch": 0.30778894472361806,
      "grad_norm": 0.0,
      "learning_rate": 3.461296868960185e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6370
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 125.625,
      "epoch": 0.3078372632392733,
      "grad_norm": 0.0,
      "learning_rate": 3.4610552763819093e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6371
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 129.25,
      "epoch": 0.3078855817549285,
      "grad_norm": 15.32740047149697,
      "learning_rate": 3.4608136838036335e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6372
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.90625,
      "epoch": 0.3079339002705837,
      "grad_norm": 0.0,
      "learning_rate": 3.460572091225357e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6373
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.0625,
      "epoch": 0.3079822187862389,
      "grad_norm": 0.0,
      "learning_rate": 3.460330498647082e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6374
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.9375,
      "epoch": 0.30803053730189406,
      "grad_norm": 0.0,
      "learning_rate": 3.4600889060688053e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6375
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 122.46875,
      "epoch": 0.3080788558175493,
      "grad_norm": 0.0,
      "learning_rate": 3.4598473134905295e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 6376
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 132.0,
      "epoch": 0.30812717433320447,
      "grad_norm": 0.0,
      "learning_rate": 3.4596057209122536e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6377
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.59375,
      "epoch": 0.3081754928488597,
      "grad_norm": 0.0,
      "learning_rate": 3.459364128333978e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6378
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 124.78125,
      "epoch": 0.3082238113645149,
      "grad_norm": 0.0,
      "learning_rate": 3.4591225357557014e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6379
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.28125,
      "epoch": 0.30827212988017005,
      "grad_norm": 0.0,
      "learning_rate": 3.458880943177425e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6380
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 120.875,
      "epoch": 0.3083204483958253,
      "grad_norm": 0.0,
      "learning_rate": 3.4586393505991496e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6381
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 114.8125,
      "epoch": 0.30836876691148046,
      "grad_norm": 0.0,
      "learning_rate": 3.458397758020873e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6382
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 124.6875,
      "epoch": 0.3084170854271357,
      "grad_norm": 0.0,
      "learning_rate": 3.4581561654425974e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6383
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.96875,
      "epoch": 0.3084654039427909,
      "grad_norm": 0.0,
      "learning_rate": 3.4579145728643215e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6384
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 131.75,
      "epoch": 0.30851372245844605,
      "grad_norm": 0.0,
      "learning_rate": 3.4576729802860456e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6385
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 117.6875,
      "epoch": 0.3085620409741013,
      "grad_norm": 34.94237089200833,
      "learning_rate": 3.457431387707769e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6386
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.84375,
      "epoch": 0.30861035948975646,
      "grad_norm": 0.0,
      "learning_rate": 3.457189795129494e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6387
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.25,
      "epoch": 0.3086586780054117,
      "grad_norm": 0.0,
      "learning_rate": 3.4569482025512175e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6388
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.3125,
      "epoch": 0.30870699652106687,
      "grad_norm": 0.0,
      "learning_rate": 3.456706609972941e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6389
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.09375,
      "epoch": 0.30875531503672204,
      "grad_norm": 0.0,
      "learning_rate": 3.4564650173946653e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 6390
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 140.25,
      "epoch": 0.3088036335523773,
      "grad_norm": 0.0,
      "learning_rate": 3.4562234248163894e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.3093337416648865,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 0.96875,
      "step": 6391
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.0625,
      "epoch": 0.30885195206803245,
      "grad_norm": 0.0,
      "learning_rate": 3.4559818322381135e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6392
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 129.3125,
      "epoch": 0.3089002705836877,
      "grad_norm": 0.0,
      "learning_rate": 3.455740239659837e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 6393
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.84375,
      "epoch": 0.30894858909934286,
      "grad_norm": 0.0,
      "learning_rate": 3.455498647081562e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6394
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 143.78125,
      "epoch": 0.30899690761499804,
      "grad_norm": 7.951689237416153,
      "learning_rate": 3.4552570545032854e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.2893187999725342,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6395
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.65625,
      "epoch": 0.30904522613065327,
      "grad_norm": 0.0,
      "learning_rate": 3.4550154619250096e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6396
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.9375,
      "epoch": 0.30909354464630845,
      "grad_norm": 0.0,
      "learning_rate": 3.4547738693467337e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6397
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.09375,
      "epoch": 0.3091418631619637,
      "grad_norm": 0.0,
      "learning_rate": 3.454532276768458e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6398
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.15625,
      "epoch": 0.30919018167761886,
      "grad_norm": 0.0,
      "learning_rate": 3.4542906841901814e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6399
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.84375,
      "epoch": 0.30923850019327404,
      "grad_norm": 0.0,
      "learning_rate": 3.454049091611905e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6400
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 114.625,
      "epoch": 0.30928681870892927,
      "grad_norm": 0.0,
      "learning_rate": 3.4538074990336297e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6401
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.09375,
      "epoch": 0.30933513722458444,
      "grad_norm": 0.0,
      "learning_rate": 3.4535659064553533e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6402
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.375,
      "epoch": 0.3093834557402397,
      "grad_norm": 0.0,
      "learning_rate": 3.4533243138770775e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6403
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.0,
      "epoch": 0.30943177425589485,
      "grad_norm": 0.0,
      "learning_rate": 3.4530827212988016e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6404
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.59375,
      "epoch": 0.30948009277155003,
      "grad_norm": 0.0,
      "learning_rate": 3.4528411287205257e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6405
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.125,
      "epoch": 0.30952841128720526,
      "grad_norm": 0.0,
      "learning_rate": 3.4525995361422493e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6406
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.125,
      "epoch": 0.30957672980286044,
      "grad_norm": 0.0,
      "learning_rate": 3.452357943563974e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6407
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 134.6875,
      "epoch": 0.30962504831851567,
      "grad_norm": 0.0,
      "learning_rate": 3.4521163509856976e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6408
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 125.15625,
      "epoch": 0.30967336683417085,
      "grad_norm": 0.0,
      "learning_rate": 3.451874758407422e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6409
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 131.8125,
      "epoch": 0.309721685349826,
      "grad_norm": 0.0,
      "learning_rate": 3.4516331658291453e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 6410
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.03125,
      "epoch": 0.30977000386548126,
      "grad_norm": 0.0,
      "learning_rate": 3.4513915732508695e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.22201895713806152,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6411
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.4375,
      "epoch": 0.30981832238113644,
      "grad_norm": 0.0,
      "learning_rate": 3.4511499806725936e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6412
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.1875,
      "epoch": 0.30986664089679167,
      "grad_norm": 0.0,
      "learning_rate": 3.450908388094317e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 6413
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 126.03125,
      "epoch": 0.30991495941244684,
      "grad_norm": 0.0,
      "learning_rate": 3.450666795516042e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 6414
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.5625,
      "epoch": 0.309963277928102,
      "grad_norm": 0.0,
      "learning_rate": 3.4504252029377655e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6415
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 121.75,
      "epoch": 0.31001159644375725,
      "grad_norm": 0.0,
      "learning_rate": 3.4501836103594896e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6416
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.5,
      "epoch": 0.31005991495941243,
      "grad_norm": 0.0,
      "learning_rate": 3.449942017781214e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6417
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.5,
      "epoch": 0.31010823347506766,
      "grad_norm": 0.0,
      "learning_rate": 3.449700425202938e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6418
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.15625,
      "epoch": 0.31015655199072284,
      "grad_norm": 0.0,
      "learning_rate": 3.4494588326246615e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6419
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 124.25,
      "epoch": 0.310204870506378,
      "grad_norm": 0.0,
      "learning_rate": 3.449217240046385e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6420
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 111.3125,
      "epoch": 0.31025318902203325,
      "grad_norm": 10.920041393653994,
      "learning_rate": 3.44897564746811e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2566300630569458,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6421
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.8125,
      "epoch": 0.3103015075376884,
      "grad_norm": 0.0,
      "learning_rate": 3.4487340548898334e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6422
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.28125,
      "epoch": 0.31034982605334366,
      "grad_norm": 0.0,
      "learning_rate": 3.4484924623115575e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6423
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 123.625,
      "epoch": 0.31039814456899883,
      "grad_norm": 0.0,
      "learning_rate": 3.4482508697332817e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6424
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.28125,
      "epoch": 0.310446463084654,
      "grad_norm": 0.0,
      "learning_rate": 3.448009277155006e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6425
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.46875,
      "epoch": 0.31049478160030924,
      "grad_norm": 0.0,
      "learning_rate": 3.4477676845767294e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6426
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.28125,
      "epoch": 0.3105431001159644,
      "grad_norm": 0.0,
      "learning_rate": 3.447526091998454e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6427
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.71875,
      "epoch": 0.31059141863161965,
      "grad_norm": 0.0,
      "learning_rate": 3.4472844994201777e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6428
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 120.28125,
      "epoch": 0.31063973714727483,
      "grad_norm": 13.530809858377019,
      "learning_rate": 3.447042906841902e-07,
      "loss": 0.0,
      "reward": 1.484375,
      "reward_std": 0.15580293536186218,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6429
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.625,
      "epoch": 0.31068805566293,
      "grad_norm": 0.0,
      "learning_rate": 3.4468013142636254e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6430
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.53125,
      "epoch": 0.31073637417858524,
      "grad_norm": 0.0,
      "learning_rate": 3.4465597216853496e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6431
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.90625,
      "epoch": 0.3107846926942404,
      "grad_norm": 0.0,
      "learning_rate": 3.4463181291070737e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6432
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.3125,
      "epoch": 0.31083301120989565,
      "grad_norm": 0.0,
      "learning_rate": 3.4460765365287973e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6433
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.34375,
      "epoch": 0.3108813297255508,
      "grad_norm": 0.0,
      "learning_rate": 3.445834943950522e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 6434
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.96875,
      "epoch": 0.310929648241206,
      "grad_norm": 0.0,
      "learning_rate": 3.4455933513722456e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6435
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 118.71875,
      "epoch": 0.31097796675686123,
      "grad_norm": 0.0,
      "learning_rate": 3.4453517587939697e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6436
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.0,
      "epoch": 0.3110262852725164,
      "grad_norm": 0.0,
      "learning_rate": 3.445110166215694e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6437
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.3125,
      "epoch": 0.31107460378817164,
      "grad_norm": 0.0,
      "learning_rate": 3.444868573637418e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6438
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 130.03125,
      "epoch": 0.3111229223038268,
      "grad_norm": 0.0,
      "learning_rate": 3.4446269810591416e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6439
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 123.09375,
      "epoch": 0.311171240819482,
      "grad_norm": 0.0,
      "learning_rate": 3.444385388480865e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6440
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.46875,
      "epoch": 0.31121955933513723,
      "grad_norm": 0.0,
      "learning_rate": 3.44414379590259e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6441
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.875,
      "epoch": 0.3112678778507924,
      "grad_norm": 0.0,
      "learning_rate": 3.4439022033243135e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6442
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.1875,
      "epoch": 0.31131619636644764,
      "grad_norm": 0.0,
      "learning_rate": 3.4436606107460376e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6443
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 115.875,
      "epoch": 0.3113645148821028,
      "grad_norm": 11.819541842768654,
      "learning_rate": 3.443419018167762e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.22835706174373627,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6444
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.9375,
      "epoch": 0.311412833397758,
      "grad_norm": 0.0,
      "learning_rate": 3.443177425589486e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 6445
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 129.5625,
      "epoch": 0.3114611519134132,
      "grad_norm": 0.0,
      "learning_rate": 3.4429358330112095e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6446
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.1875,
      "epoch": 0.3115094704290684,
      "grad_norm": 0.0,
      "learning_rate": 3.442694240432934e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6447
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.21875,
      "epoch": 0.31155778894472363,
      "grad_norm": 0.0,
      "learning_rate": 3.442452647854658e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6448
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.90625,
      "epoch": 0.3116061074603788,
      "grad_norm": 0.0,
      "learning_rate": 3.442211055276382e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6449
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 123.90625,
      "epoch": 0.311654425976034,
      "grad_norm": 15.564815020802945,
      "learning_rate": 3.4419694626981055e-07,
      "loss": 0.0,
      "reward": 1.53125,
      "reward_std": 0.24067938327789307,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6450
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 135.90625,
      "epoch": 0.3117027444916892,
      "grad_norm": 0.0,
      "learning_rate": 3.4417278701198296e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6451
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 133.78125,
      "epoch": 0.3117510630073444,
      "grad_norm": 0.0,
      "learning_rate": 3.441486277541554e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6452
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 112.6875,
      "epoch": 0.31179938152299963,
      "grad_norm": 0.0,
      "learning_rate": 3.4412446849632774e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 6453
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 117.34375,
      "epoch": 0.3118477000386548,
      "grad_norm": 5.860670244359919,
      "learning_rate": 3.441003092385002e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6454
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.5,
      "epoch": 0.31189601855431,
      "grad_norm": 0.0,
      "learning_rate": 3.4407614998067257e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6455
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.28125,
      "epoch": 0.3119443370699652,
      "grad_norm": 0.0,
      "learning_rate": 3.44051990722845e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6456
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 113.5625,
      "epoch": 0.3119926555856204,
      "grad_norm": 17.351974419645376,
      "learning_rate": 3.440278314650174e-07,
      "loss": 0.0,
      "reward": 1.484375,
      "reward_std": 0.15580293536186218,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6457
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.03125,
      "epoch": 0.3120409741012756,
      "grad_norm": 0.0,
      "learning_rate": 3.440036722071898e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6458
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 125.125,
      "epoch": 0.3120892926169308,
      "grad_norm": 6.713607556697349,
      "learning_rate": 3.4397951294936217e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6459
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.84375,
      "epoch": 0.312137611132586,
      "grad_norm": 0.0,
      "learning_rate": 3.439553536915346e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6460
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.4375,
      "epoch": 0.3121859296482412,
      "grad_norm": 0.0,
      "learning_rate": 3.43931194433707e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6461
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 133.75,
      "epoch": 0.3122342481638964,
      "grad_norm": 0.0,
      "learning_rate": 3.4390703517587936e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6462
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.90625,
      "epoch": 0.3122825666795516,
      "grad_norm": 0.0,
      "learning_rate": 3.4388287591805177e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6463
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 119.8125,
      "epoch": 0.3123308851952068,
      "grad_norm": 0.0,
      "learning_rate": 3.438587166602242e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6464
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 130.78125,
      "epoch": 0.312379203710862,
      "grad_norm": 0.0,
      "learning_rate": 3.438345574023966e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 6465
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.96875,
      "epoch": 0.3124275222265172,
      "grad_norm": 0.0,
      "learning_rate": 3.4381039814456896e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6466
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 141.6875,
      "epoch": 0.3124758407421724,
      "grad_norm": 10.181598326146752,
      "learning_rate": 3.437862388867414e-07,
      "loss": -0.0,
      "reward": 1.5,
      "reward_std": 0.2529362142086029,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6467
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.28125,
      "epoch": 0.3125241592578276,
      "grad_norm": 0.0,
      "learning_rate": 3.437620796289138e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6468
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 113.6875,
      "epoch": 0.3125724777734828,
      "grad_norm": 26.349759101008264,
      "learning_rate": 3.437379203710862e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6469
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.46875,
      "epoch": 0.312620796289138,
      "grad_norm": 0.0,
      "learning_rate": 3.4371376111325856e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6470
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 113.0625,
      "epoch": 0.3126691148047932,
      "grad_norm": 10.75865757988912,
      "learning_rate": 3.43689601855431e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6471
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.75,
      "epoch": 0.3127174333204484,
      "grad_norm": 0.0,
      "learning_rate": 3.436654425976034e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6472
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 116.25,
      "epoch": 0.3127657518361036,
      "grad_norm": 6.23645694499472,
      "learning_rate": 3.4364128333977575e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.2226182520389557,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6473
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.625,
      "epoch": 0.3128140703517588,
      "grad_norm": 0.0,
      "learning_rate": 3.436171240819482e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6474
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.5,
      "epoch": 0.312862388867414,
      "grad_norm": 0.0,
      "learning_rate": 3.435929648241206e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 6475
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.46875,
      "epoch": 0.3129107073830692,
      "grad_norm": 0.0,
      "learning_rate": 3.43568805566293e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6476
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.53125,
      "epoch": 0.3129590258987244,
      "grad_norm": 0.0,
      "learning_rate": 3.435446463084654e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6477
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.28125,
      "epoch": 0.3130073444143796,
      "grad_norm": 0.0,
      "learning_rate": 3.435204870506378e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6478
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 120.15625,
      "epoch": 0.3130556629300348,
      "grad_norm": 0.0,
      "learning_rate": 3.434963277928102e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6479
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 119.8125,
      "epoch": 0.31310398144569,
      "grad_norm": 0.0,
      "learning_rate": 3.434721685349826e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6480
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 118.875,
      "epoch": 0.3131522999613452,
      "grad_norm": 5.877070515952173,
      "learning_rate": 3.43448009277155e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.2951723635196686,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6481
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.34375,
      "epoch": 0.31320061847700037,
      "grad_norm": 0.0,
      "learning_rate": 3.4342385001932736e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6482
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.8125,
      "epoch": 0.3132489369926556,
      "grad_norm": 0.0,
      "learning_rate": 3.433996907614998e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6483
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.21875,
      "epoch": 0.3132972555083108,
      "grad_norm": 0.0,
      "learning_rate": 3.433755315036722e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6484
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.71875,
      "epoch": 0.313345574023966,
      "grad_norm": 0.0,
      "learning_rate": 3.433513722458446e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6485
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.53125,
      "epoch": 0.3133938925396212,
      "grad_norm": 0.0,
      "learning_rate": 3.4332721298801696e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6486
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 124.6875,
      "epoch": 0.31344221105527637,
      "grad_norm": 0.0,
      "learning_rate": 3.4330305373018943e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6487
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.375,
      "epoch": 0.3134905295709316,
      "grad_norm": 0.0,
      "learning_rate": 3.432788944723618e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6488
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 131.28125,
      "epoch": 0.3135388480865868,
      "grad_norm": 0.0,
      "learning_rate": 3.432547352145342e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6489
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.09375,
      "epoch": 0.313587166602242,
      "grad_norm": 0.0,
      "learning_rate": 3.4323057595670657e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6490
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.3125,
      "epoch": 0.3136354851178972,
      "grad_norm": 0.0,
      "learning_rate": 3.4320641669887903e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6491
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.28125,
      "epoch": 0.31368380363355236,
      "grad_norm": 0.0,
      "learning_rate": 3.431822574410514e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6492
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 181.53125,
      "epoch": 0.3137321221492076,
      "grad_norm": 34.156618972123454,
      "learning_rate": 3.4315809818322375e-07,
      "loss": -0.0,
      "reward": 1.4375,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6493
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.25,
      "epoch": 0.31378044066486277,
      "grad_norm": 0.0,
      "learning_rate": 3.431339389253962e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6494
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 128.5,
      "epoch": 0.313828759180518,
      "grad_norm": 0.0,
      "learning_rate": 3.431097796675686e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6495
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 121.21875,
      "epoch": 0.3138770776961732,
      "grad_norm": 0.0,
      "learning_rate": 3.43085620409741e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6496
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.6875,
      "epoch": 0.31392539621182836,
      "grad_norm": 0.0,
      "learning_rate": 3.430614611519134e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6497
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.25,
      "epoch": 0.3139737147274836,
      "grad_norm": 0.0,
      "learning_rate": 3.430373018940858e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6498
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 111.25,
      "epoch": 0.31402203324313877,
      "grad_norm": 0.0,
      "learning_rate": 3.430131426362582e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6499
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 136.5,
      "epoch": 0.314070351758794,
      "grad_norm": 0.0,
      "learning_rate": 3.429889833784306e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6500
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.125,
      "epoch": 0.3141186702744492,
      "grad_norm": 0.0,
      "learning_rate": 3.42964824120603e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6501
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 125.78125,
      "epoch": 0.31416698879010435,
      "grad_norm": 0.0,
      "learning_rate": 3.4294066486277537e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6502
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.6875,
      "epoch": 0.3142153073057596,
      "grad_norm": 0.0,
      "learning_rate": 3.429165056049478e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6503
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 125.9375,
      "epoch": 0.31426362582141476,
      "grad_norm": 0.0,
      "learning_rate": 3.428923463471202e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6504
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.75,
      "epoch": 0.31431194433707,
      "grad_norm": 0.0,
      "learning_rate": 3.428681870892926e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6505
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 116.25,
      "epoch": 0.31436026285272517,
      "grad_norm": 8.34345720313978,
      "learning_rate": 3.4284402783146497e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6506
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 132.78125,
      "epoch": 0.31440858136838035,
      "grad_norm": 0.0,
      "learning_rate": 3.4281986857363744e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6507
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.03125,
      "epoch": 0.3144568998840356,
      "grad_norm": 0.0,
      "learning_rate": 3.427957093158098e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6508
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.875,
      "epoch": 0.31450521839969076,
      "grad_norm": 0.0,
      "learning_rate": 3.427715500579822e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6509
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 124.1875,
      "epoch": 0.314553536915346,
      "grad_norm": 0.0,
      "learning_rate": 3.427473908001546e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6510
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.65625,
      "epoch": 0.31460185543100117,
      "grad_norm": 0.0,
      "learning_rate": 3.4272323154232704e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6511
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.78125,
      "epoch": 0.31465017394665634,
      "grad_norm": 0.0,
      "learning_rate": 3.426990722844994e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6512
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 129.25,
      "epoch": 0.3146984924623116,
      "grad_norm": 0.0,
      "learning_rate": 3.4267491302667176e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6513
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.53125,
      "epoch": 0.31474681097796675,
      "grad_norm": 0.0,
      "learning_rate": 3.4265075376884423e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6514
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.40625,
      "epoch": 0.314795129493622,
      "grad_norm": 0.0,
      "learning_rate": 3.426265945110166e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6515
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.25,
      "epoch": 0.31484344800927716,
      "grad_norm": 0.0,
      "learning_rate": 3.42602435253189e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6516
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.53125,
      "epoch": 0.31489176652493234,
      "grad_norm": 0.0,
      "learning_rate": 3.425782759953614e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6517
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 128.96875,
      "epoch": 0.31494008504058757,
      "grad_norm": 0.0,
      "learning_rate": 3.4255411673753383e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6518
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 117.875,
      "epoch": 0.31498840355624275,
      "grad_norm": 0.0,
      "learning_rate": 3.425299574797062e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6519
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 129.25,
      "epoch": 0.315036722071898,
      "grad_norm": 0.0,
      "learning_rate": 3.425057982218786e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6520
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.875,
      "epoch": 0.31508504058755316,
      "grad_norm": 0.0,
      "learning_rate": 3.42481638964051e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6521
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.53125,
      "epoch": 0.31513335910320833,
      "grad_norm": 0.0,
      "learning_rate": 3.4245747970622343e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6522
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 128.375,
      "epoch": 0.31518167761886356,
      "grad_norm": 0.0,
      "learning_rate": 3.424333204483958e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6523
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.0,
      "epoch": 0.31522999613451874,
      "grad_norm": 0.0,
      "learning_rate": 3.424091611905682e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6524
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 122.0,
      "epoch": 0.315278314650174,
      "grad_norm": 0.0,
      "learning_rate": 3.423850019327406e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6525
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 108.6875,
      "epoch": 0.31532663316582915,
      "grad_norm": 0.0,
      "learning_rate": 3.42360842674913e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6526
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.03125,
      "epoch": 0.31537495168148433,
      "grad_norm": 0.0,
      "learning_rate": 3.4233668341708545e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6527
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 135.21875,
      "epoch": 0.31542327019713956,
      "grad_norm": 0.0,
      "learning_rate": 3.423125241592578e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6528
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.78125,
      "epoch": 0.31547158871279474,
      "grad_norm": 0.0,
      "learning_rate": 3.422883649014302e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6529
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 118.375,
      "epoch": 0.31551990722844997,
      "grad_norm": 0.0,
      "learning_rate": 3.422642056436026e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6530
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 123.5,
      "epoch": 0.31556822574410515,
      "grad_norm": 0.0,
      "learning_rate": 3.4224004638577505e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6531
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.8125,
      "epoch": 0.3156165442597603,
      "grad_norm": 0.0,
      "learning_rate": 3.422158871279474e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6532
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.34375,
      "epoch": 0.31566486277541556,
      "grad_norm": 0.0,
      "learning_rate": 3.4219172787011977e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6533
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 111.03125,
      "epoch": 0.31571318129107073,
      "grad_norm": 0.0,
      "learning_rate": 3.4216756861229224e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6534
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.09375,
      "epoch": 0.31576149980672596,
      "grad_norm": 0.0,
      "learning_rate": 3.421434093544646e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6535
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.875,
      "epoch": 0.31580981832238114,
      "grad_norm": 0.0,
      "learning_rate": 3.42119250096637e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6536
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.6875,
      "epoch": 0.3158581368380363,
      "grad_norm": 0.0,
      "learning_rate": 3.420950908388094e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6537
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.84375,
      "epoch": 0.31590645535369155,
      "grad_norm": 0.0,
      "learning_rate": 3.4207093158098184e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6538
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.96875,
      "epoch": 0.31595477386934673,
      "grad_norm": 0.0,
      "learning_rate": 3.420467723231542e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6539
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 127.4375,
      "epoch": 0.31600309238500196,
      "grad_norm": 0.0,
      "learning_rate": 3.420226130653266e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6540
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 122.375,
      "epoch": 0.31605141090065714,
      "grad_norm": 0.0,
      "learning_rate": 3.41998453807499e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 6541
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 113.5625,
      "epoch": 0.3160997294163123,
      "grad_norm": 35.683592637627285,
      "learning_rate": 3.4197429454967144e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6542
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.8125,
      "epoch": 0.31614804793196755,
      "grad_norm": 0.0,
      "learning_rate": 3.419501352918438e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6543
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.375,
      "epoch": 0.3161963664476227,
      "grad_norm": 0.0,
      "learning_rate": 3.419259760340162e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6544
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 135.15625,
      "epoch": 0.31624468496327796,
      "grad_norm": 0.0,
      "learning_rate": 3.4190181677618863e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6545
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 126.46875,
      "epoch": 0.31629300347893313,
      "grad_norm": 0.0,
      "learning_rate": 3.41877657518361e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6546
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 124.90625,
      "epoch": 0.3163413219945883,
      "grad_norm": 0.0,
      "learning_rate": 3.4185349826053346e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 6547
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.21875,
      "epoch": 0.31638964051024354,
      "grad_norm": 0.0,
      "learning_rate": 3.418293390027058e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6548
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.8125,
      "epoch": 0.3164379590258987,
      "grad_norm": 0.0,
      "learning_rate": 3.4180517974487823e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6549
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.40625,
      "epoch": 0.31648627754155395,
      "grad_norm": 0.0,
      "learning_rate": 3.417810204870506e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6550
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.59375,
      "epoch": 0.3165345960572091,
      "grad_norm": 0.0,
      "learning_rate": 3.4175686122922306e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6551
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 114.65625,
      "epoch": 0.3165829145728643,
      "grad_norm": 17.99835473642806,
      "learning_rate": 3.417327019713954e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6552
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 114.0625,
      "epoch": 0.31663123308851954,
      "grad_norm": 0.0,
      "learning_rate": 3.4170854271356783e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6553
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.5625,
      "epoch": 0.3166795516041747,
      "grad_norm": 0.0,
      "learning_rate": 3.4168438345574024e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6554
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 138.0,
      "epoch": 0.31672787011982995,
      "grad_norm": 25.174264464970364,
      "learning_rate": 3.416602241979126e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.30935919284820557,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6555
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.46875,
      "epoch": 0.3167761886354851,
      "grad_norm": 0.0,
      "learning_rate": 3.41636064940085e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6556
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 127.90625,
      "epoch": 0.3168245071511403,
      "grad_norm": 0.0,
      "learning_rate": 3.4161190568225743e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6557
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 128.21875,
      "epoch": 0.31687282566679553,
      "grad_norm": 0.0,
      "learning_rate": 3.4158774642442985e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6558
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.15625,
      "epoch": 0.3169211441824507,
      "grad_norm": 0.0,
      "learning_rate": 3.415635871666022e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 6559
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.78125,
      "epoch": 0.31696946269810594,
      "grad_norm": 0.0,
      "learning_rate": 3.415394279087746e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6560
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.46875,
      "epoch": 0.3170177812137611,
      "grad_norm": 0.0,
      "learning_rate": 3.4151526865094703e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6561
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 131.1875,
      "epoch": 0.3170660997294163,
      "grad_norm": 0.0,
      "learning_rate": 3.4149110939311945e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6562
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.46875,
      "epoch": 0.3171144182450715,
      "grad_norm": 0.0,
      "learning_rate": 3.414669501352918e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6563
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 114.90625,
      "epoch": 0.3171627367607267,
      "grad_norm": 6.804036292089738,
      "learning_rate": 3.414427908774643e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6564
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 124.9375,
      "epoch": 0.31721105527638194,
      "grad_norm": 0.0,
      "learning_rate": 3.4141863161963664e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6565
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.09375,
      "epoch": 0.3172593737920371,
      "grad_norm": 0.0,
      "learning_rate": 3.41394472361809e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6566
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.34375,
      "epoch": 0.3173076923076923,
      "grad_norm": 0.0,
      "learning_rate": 3.4137031310398146e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6567
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.65625,
      "epoch": 0.3173560108233475,
      "grad_norm": 0.0,
      "learning_rate": 3.413461538461538e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6568
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 125.03125,
      "epoch": 0.3174043293390027,
      "grad_norm": 0.0,
      "learning_rate": 3.4132199458832624e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6569
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 136.09375,
      "epoch": 0.31745264785465793,
      "grad_norm": 0.0,
      "learning_rate": 3.412978353304986e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6570
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.125,
      "epoch": 0.3175009663703131,
      "grad_norm": 0.0,
      "learning_rate": 3.4127367607267106e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6571
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 133.03125,
      "epoch": 0.3175492848859683,
      "grad_norm": 0.0,
      "learning_rate": 3.412495168148434e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6572
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 139.125,
      "epoch": 0.3175976034016235,
      "grad_norm": 46.0400958146941,
      "learning_rate": 3.4122535755701584e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6573
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 135.4375,
      "epoch": 0.3176459219172787,
      "grad_norm": 0.0,
      "learning_rate": 3.4120119829918825e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6574
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.5,
      "epoch": 0.3176942404329339,
      "grad_norm": 0.0,
      "learning_rate": 3.411770390413606e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6575
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 146.8125,
      "epoch": 0.3177425589485891,
      "grad_norm": 0.0,
      "learning_rate": 3.4115287978353303e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6576
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.71875,
      "epoch": 0.3177908774642443,
      "grad_norm": 0.0,
      "learning_rate": 3.4112872052570544e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6577
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.34375,
      "epoch": 0.3178391959798995,
      "grad_norm": 0.0,
      "learning_rate": 3.4110456126787785e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6578
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.03125,
      "epoch": 0.3178875144955547,
      "grad_norm": 0.0,
      "learning_rate": 3.410804020100502e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6579
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.1875,
      "epoch": 0.3179358330112099,
      "grad_norm": 0.0,
      "learning_rate": 3.4105624275222263e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6580
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 121.0,
      "epoch": 0.3179841515268651,
      "grad_norm": 36.605553350049085,
      "learning_rate": 3.4103208349439504e-07,
      "loss": -0.0,
      "reward": 2.4375,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 6581
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.6875,
      "epoch": 0.3180324700425203,
      "grad_norm": 0.0,
      "learning_rate": 3.4100792423656746e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6582
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 131.9375,
      "epoch": 0.3180807885581755,
      "grad_norm": 5.358187955553591,
      "learning_rate": 3.409837649787398e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 6583
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 122.875,
      "epoch": 0.3181291070738307,
      "grad_norm": 0.0,
      "learning_rate": 3.409596057209123e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6584
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 130.0,
      "epoch": 0.3181774255894859,
      "grad_norm": 0.0,
      "learning_rate": 3.4093544646308464e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6585
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 125.90625,
      "epoch": 0.3182257441051411,
      "grad_norm": 0.0,
      "learning_rate": 3.40911287205257e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6586
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 128.8125,
      "epoch": 0.31827406262079627,
      "grad_norm": 0.0,
      "learning_rate": 3.4088712794742947e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6587
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 134.71875,
      "epoch": 0.3183223811364515,
      "grad_norm": 0.0,
      "learning_rate": 3.4086296868960183e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6588
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.6875,
      "epoch": 0.3183706996521067,
      "grad_norm": 0.0,
      "learning_rate": 3.4083880943177425e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6589
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 131.21875,
      "epoch": 0.3184190181677619,
      "grad_norm": 18.957751447027324,
      "learning_rate": 3.408146501739466e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.3902835249900818,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6590
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 134.25,
      "epoch": 0.3184673366834171,
      "grad_norm": 0.0,
      "learning_rate": 3.4079049091611907e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6591
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 142.59375,
      "epoch": 0.31851565519907227,
      "grad_norm": 0.0,
      "learning_rate": 3.4076633165829143e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6592
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 123.4375,
      "epoch": 0.3185639737147275,
      "grad_norm": 0.0,
      "learning_rate": 3.4074217240046385e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6593
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.9375,
      "epoch": 0.3186122922303827,
      "grad_norm": 0.0,
      "learning_rate": 3.4071801314263626e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6594
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 130.21875,
      "epoch": 0.3186606107460379,
      "grad_norm": 0.0,
      "learning_rate": 3.406938538848086e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 6595
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 116.84375,
      "epoch": 0.3187089292616931,
      "grad_norm": 0.0,
      "learning_rate": 3.4066969462698103e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6596
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 138.09375,
      "epoch": 0.31875724777734826,
      "grad_norm": 0.0,
      "learning_rate": 3.4064553536915345e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6597
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 225.65625,
      "epoch": 0.3188055662930035,
      "grad_norm": 0.0,
      "learning_rate": 3.4062137611132586e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 0.96875,
      "step": 6598
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 132.125,
      "epoch": 0.31885388480865867,
      "grad_norm": 0.0,
      "learning_rate": 3.405972168534982e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6599
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 138.5625,
      "epoch": 0.3189022033243139,
      "grad_norm": 0.0,
      "learning_rate": 3.4057305759567064e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 6600
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 136.6875,
      "epoch": 0.3189505218399691,
      "grad_norm": 0.0,
      "learning_rate": 3.4054889833784305e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6601
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 151.53125,
      "epoch": 0.31899884035562426,
      "grad_norm": 8.158993437887887,
      "learning_rate": 3.4052473908001546e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.27564918994903564,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6602
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 128.59375,
      "epoch": 0.3190471588712795,
      "grad_norm": 0.0,
      "learning_rate": 3.405005798221878e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 6603
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.4375,
      "epoch": 0.31909547738693467,
      "grad_norm": 0.0,
      "learning_rate": 3.404764205643603e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6604
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 130.0,
      "epoch": 0.3191437959025899,
      "grad_norm": 0.0,
      "learning_rate": 3.4045226130653265e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6605
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 147.65625,
      "epoch": 0.3191921144182451,
      "grad_norm": 6.3876109488289865,
      "learning_rate": 3.40428102048705e-07,
      "loss": -0.0,
      "reward": 2.25,
      "reward_std": 0.2529362142086029,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6606
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 135.59375,
      "epoch": 0.31924043293390025,
      "grad_norm": 0.0,
      "learning_rate": 3.404039427908775e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6607
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 139.625,
      "epoch": 0.3192887514495555,
      "grad_norm": 0.0,
      "learning_rate": 3.4037978353304984e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6608
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 139.375,
      "epoch": 0.31933706996521066,
      "grad_norm": 0.0,
      "learning_rate": 3.4035562427522225e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6609
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 197.40625,
      "epoch": 0.3193853884808659,
      "grad_norm": 1.7544445276251872,
      "learning_rate": 3.403314650173946e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.2158610224723816,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 0.96875,
      "step": 6610
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 142.78125,
      "epoch": 0.31943370699652107,
      "grad_norm": 0.0,
      "learning_rate": 3.403073057595671e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6611
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.125,
      "epoch": 0.31948202551217625,
      "grad_norm": 0.0,
      "learning_rate": 3.4028314650173944e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6612
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 155.71875,
      "epoch": 0.3195303440278315,
      "grad_norm": 0.0,
      "learning_rate": 3.4025898724391185e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6613
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 142.90625,
      "epoch": 0.31957866254348666,
      "grad_norm": 0.0,
      "learning_rate": 3.4023482798608427e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6614
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 197.75,
      "epoch": 0.3196269810591419,
      "grad_norm": 0.0,
      "learning_rate": 3.402106687282567e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 0.96875,
      "step": 6615
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 151.4375,
      "epoch": 0.31967529957479707,
      "grad_norm": 0.0,
      "learning_rate": 3.4018650947042904e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6616
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 143.25,
      "epoch": 0.31972361809045224,
      "grad_norm": 0.0,
      "learning_rate": 3.4016235021260146e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6617
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 130.6875,
      "epoch": 0.3197719366061075,
      "grad_norm": 0.0,
      "learning_rate": 3.4013819095477387e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6618
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 196.78125,
      "epoch": 0.31982025512176265,
      "grad_norm": 3.755804588821209,
      "learning_rate": 3.4011403169694623e-07,
      "loss": -0.0,
      "reward": 2.3125,
      "reward_std": 0.3384184241294861,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 0.96875,
      "step": 6619
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 156.03125,
      "epoch": 0.3198685736374179,
      "grad_norm": 0.0,
      "learning_rate": 3.4008987243911864e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6620
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 138.1875,
      "epoch": 0.31991689215307306,
      "grad_norm": 12.25123115366758,
      "learning_rate": 3.4006571318129106e-07,
      "loss": -0.0,
      "reward": 2.09375,
      "reward_std": 0.3608423173427582,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6621
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 192.3125,
      "epoch": 0.31996521066872824,
      "grad_norm": 9.502567702896572,
      "learning_rate": 3.4004155392346347e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.22738909721374512,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6622
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 138.1875,
      "epoch": 0.32001352918438347,
      "grad_norm": 0.0,
      "learning_rate": 3.4001739466563583e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6623
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 157.90625,
      "epoch": 0.32006184770003865,
      "grad_norm": 8.887635318774258,
      "learning_rate": 3.399932354078083e-07,
      "loss": -0.0,
      "reward": 1.171875,
      "reward_std": 0.188242569565773,
      "rewards/accuracy_reward": 0.125,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 6624
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 147.9375,
      "epoch": 0.3201101662156939,
      "grad_norm": 0.0,
      "learning_rate": 3.3996907614998066e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6625
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 153.59375,
      "epoch": 0.32015848473134906,
      "grad_norm": 0.0,
      "learning_rate": 3.39944916892153e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6626
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 183.78125,
      "epoch": 0.32020680324700423,
      "grad_norm": 0.0,
      "learning_rate": 3.399207576343255e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6627
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 143.34375,
      "epoch": 0.32025512176265947,
      "grad_norm": 0.0,
      "learning_rate": 3.3989659837649785e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6628
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 147.4375,
      "epoch": 0.32030344027831464,
      "grad_norm": 0.0,
      "learning_rate": 3.3987243911867026e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6629
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 142.90625,
      "epoch": 0.3203517587939699,
      "grad_norm": 0.0,
      "learning_rate": 3.398482798608426e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6630
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 144.71875,
      "epoch": 0.32040007730962505,
      "grad_norm": 0.0,
      "learning_rate": 3.398241206030151e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6631
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 158.46875,
      "epoch": 0.32044839582528023,
      "grad_norm": 0.0,
      "learning_rate": 3.3979996134518745e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6632
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 199.0,
      "epoch": 0.32049671434093546,
      "grad_norm": 1.8902451594858893,
      "learning_rate": 3.3977580208735986e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 0.96875,
      "step": 6633
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 139.59375,
      "epoch": 0.32054503285659064,
      "grad_norm": 0.0,
      "learning_rate": 3.397516428295323e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6634
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 187.8125,
      "epoch": 0.32059335137224587,
      "grad_norm": 0.0,
      "learning_rate": 3.397274835717047e-07,
      "loss": 0.0,
      "reward": 0.96875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 0.96875,
      "step": 6635
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 151.375,
      "epoch": 0.32064166988790105,
      "grad_norm": 0.0,
      "learning_rate": 3.3970332431387705e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6636
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 130.0,
      "epoch": 0.3206899884035562,
      "grad_norm": 0.0,
      "learning_rate": 3.3967916505604946e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6637
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 155.65625,
      "epoch": 0.32073830691921146,
      "grad_norm": 7.716679153820944,
      "learning_rate": 3.396550057982219e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.22152717411518097,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6638
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 148.9375,
      "epoch": 0.32078662543486663,
      "grad_norm": 0.0,
      "learning_rate": 3.3963084654039424e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6639
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 148.125,
      "epoch": 0.32083494395052187,
      "grad_norm": 0.0,
      "learning_rate": 3.3960668728256665e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6640
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 130.75,
      "epoch": 0.32088326246617704,
      "grad_norm": 0.0,
      "learning_rate": 3.3958252802473907e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6641
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 152.84375,
      "epoch": 0.3209315809818322,
      "grad_norm": 0.0,
      "learning_rate": 3.395583687669115e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6642
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 137.4375,
      "epoch": 0.32097989949748745,
      "grad_norm": 0.0,
      "learning_rate": 3.3953420950908384e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6643
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 141.75,
      "epoch": 0.32102821801314263,
      "grad_norm": 0.0,
      "learning_rate": 3.395100502512563e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6644
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 147.875,
      "epoch": 0.32107653652879786,
      "grad_norm": 0.0,
      "learning_rate": 3.3948589099342867e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 6645
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 130.1875,
      "epoch": 0.32112485504445304,
      "grad_norm": 0.0,
      "learning_rate": 3.3946173173560103e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6646
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 136.1875,
      "epoch": 0.3211731735601082,
      "grad_norm": 0.0,
      "learning_rate": 3.394375724777735e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6647
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 136.1875,
      "epoch": 0.32122149207576345,
      "grad_norm": 0.0,
      "learning_rate": 3.3941341321994586e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6648
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 134.0625,
      "epoch": 0.3212698105914186,
      "grad_norm": 0.0,
      "learning_rate": 3.3938925396211827e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6649
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 137.28125,
      "epoch": 0.32131812910707386,
      "grad_norm": 0.0,
      "learning_rate": 3.3936509470429063e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6650
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 159.34375,
      "epoch": 0.32136644762272903,
      "grad_norm": 0.0,
      "learning_rate": 3.393409354464631e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6651
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 148.59375,
      "epoch": 0.3214147661383842,
      "grad_norm": 0.0,
      "learning_rate": 3.3931677618863546e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6652
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 143.28125,
      "epoch": 0.32146308465403944,
      "grad_norm": 0.0,
      "learning_rate": 3.3929261693080787e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6653
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 133.4375,
      "epoch": 0.3215114031696946,
      "grad_norm": 0.0,
      "learning_rate": 3.392684576729803e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6654
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 134.03125,
      "epoch": 0.32155972168534985,
      "grad_norm": 0.0,
      "learning_rate": 3.392442984151527e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6655
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 136.90625,
      "epoch": 0.32160804020100503,
      "grad_norm": 0.0,
      "learning_rate": 3.3922013915732506e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6656
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 142.75,
      "epoch": 0.3216563587166602,
      "grad_norm": 0.0,
      "learning_rate": 3.3919597989949747e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6657
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 143.8125,
      "epoch": 0.32170467723231544,
      "grad_norm": 0.0,
      "learning_rate": 3.391718206416699e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6658
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 137.71875,
      "epoch": 0.3217529957479706,
      "grad_norm": 0.0,
      "learning_rate": 3.3914766138384225e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6659
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 131.34375,
      "epoch": 0.32180131426362585,
      "grad_norm": 0.0,
      "learning_rate": 3.3912350212601466e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6660
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 144.09375,
      "epoch": 0.321849632779281,
      "grad_norm": 17.886947718741858,
      "learning_rate": 3.3909934286818707e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.26726123690605164,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6661
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 131.625,
      "epoch": 0.3218979512949362,
      "grad_norm": 0.0,
      "learning_rate": 3.390751836103595e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6662
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 139.375,
      "epoch": 0.32194626981059143,
      "grad_norm": 0.0,
      "learning_rate": 3.3905102435253185e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6663
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 161.0625,
      "epoch": 0.3219945883262466,
      "grad_norm": 23.592103242348546,
      "learning_rate": 3.390268650947043e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.35564959049224854,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 0.96875,
      "step": 6664
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 142.96875,
      "epoch": 0.32204290684190184,
      "grad_norm": 0.0,
      "learning_rate": 3.390027058368767e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6665
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 135.5625,
      "epoch": 0.322091225357557,
      "grad_norm": 0.0,
      "learning_rate": 3.389785465790491e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 6666
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 141.125,
      "epoch": 0.3221395438732122,
      "grad_norm": 0.0,
      "learning_rate": 3.389543873212215e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6667
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 152.25,
      "epoch": 0.32218786238886743,
      "grad_norm": 0.0,
      "learning_rate": 3.3893022806339386e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 6668
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 135.625,
      "epoch": 0.3222361809045226,
      "grad_norm": 0.0,
      "learning_rate": 3.389060688055663e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6669
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 158.1875,
      "epoch": 0.32228449942017784,
      "grad_norm": 3.827698622716593,
      "learning_rate": 3.3888190954773864e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.26409149169921875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6670
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 137.3125,
      "epoch": 0.322332817935833,
      "grad_norm": 0.0,
      "learning_rate": 3.388577502899111e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6671
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 136.6875,
      "epoch": 0.3223811364514882,
      "grad_norm": 0.0,
      "learning_rate": 3.3883359103208346e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6672
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 126.125,
      "epoch": 0.3224294549671434,
      "grad_norm": 6.09843377137442,
      "learning_rate": 3.388094317742559e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6673
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 135.0,
      "epoch": 0.3224777734827986,
      "grad_norm": 0.0,
      "learning_rate": 3.387852725164283e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 6674
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 139.15625,
      "epoch": 0.32252609199845383,
      "grad_norm": 0.0,
      "learning_rate": 3.387611132586007e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6675
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 157.6875,
      "epoch": 0.322574410514109,
      "grad_norm": 0.0,
      "learning_rate": 3.3873695400077307e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6676
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.625,
      "epoch": 0.3226227290297642,
      "grad_norm": 0.0,
      "learning_rate": 3.3871279474294553e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6677
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 124.65625,
      "epoch": 0.3226710475454194,
      "grad_norm": 7.850385037879736,
      "learning_rate": 3.386886354851179e-07,
      "loss": 0.0,
      "reward": 1.453125,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6678
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 131.96875,
      "epoch": 0.3227193660610746,
      "grad_norm": 0.0,
      "learning_rate": 3.3866447622729025e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6679
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 130.65625,
      "epoch": 0.32276768457672983,
      "grad_norm": 0.0,
      "learning_rate": 3.3864031696946267e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6680
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 131.96875,
      "epoch": 0.322816003092385,
      "grad_norm": 0.0,
      "learning_rate": 3.386161577116351e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6681
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 131.59375,
      "epoch": 0.3228643216080402,
      "grad_norm": 0.0,
      "learning_rate": 3.385919984538075e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6682
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 136.15625,
      "epoch": 0.3229126401236954,
      "grad_norm": 0.0,
      "learning_rate": 3.3856783919597986e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6683
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 132.25,
      "epoch": 0.3229609586393506,
      "grad_norm": 0.0,
      "learning_rate": 3.385436799381523e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6684
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 126.65625,
      "epoch": 0.3230092771550058,
      "grad_norm": 0.0,
      "learning_rate": 3.385195206803247e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6685
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 129.25,
      "epoch": 0.323057595670661,
      "grad_norm": 3.7912647869338056,
      "learning_rate": 3.384953614224971e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6686
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 137.5,
      "epoch": 0.3231059141863162,
      "grad_norm": 0.0,
      "learning_rate": 3.384712021646695e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6687
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 132.75,
      "epoch": 0.3231542327019714,
      "grad_norm": 0.0,
      "learning_rate": 3.3844704290684187e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6688
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 123.9375,
      "epoch": 0.3232025512176266,
      "grad_norm": 0.0,
      "learning_rate": 3.384228836490143e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6689
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 143.75,
      "epoch": 0.3232508697332818,
      "grad_norm": 0.0,
      "learning_rate": 3.3839872439118665e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6690
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 131.9375,
      "epoch": 0.323299188248937,
      "grad_norm": 0.0,
      "learning_rate": 3.383745651333591e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6691
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 136.0,
      "epoch": 0.3233475067645922,
      "grad_norm": 0.0,
      "learning_rate": 3.3835040587553147e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6692
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 122.8125,
      "epoch": 0.3233958252802474,
      "grad_norm": 12.138058721017366,
      "learning_rate": 3.383262466177039e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6693
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 132.84375,
      "epoch": 0.3234441437959026,
      "grad_norm": 0.0,
      "learning_rate": 3.383020873598763e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6694
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 129.90625,
      "epoch": 0.3234924623115578,
      "grad_norm": 0.0,
      "learning_rate": 3.382779281020487e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6695
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 125.0625,
      "epoch": 0.323540780827213,
      "grad_norm": 6.829998567092092,
      "learning_rate": 3.382537688442211e-07,
      "loss": -0.0,
      "reward": 2.328125,
      "reward_std": 0.2962459325790405,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6696
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 135.125,
      "epoch": 0.32358909934286817,
      "grad_norm": 0.0,
      "learning_rate": 3.3822960958639354e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6697
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 126.125,
      "epoch": 0.3236374178585234,
      "grad_norm": 0.0,
      "learning_rate": 3.382054503285659e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6698
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.53125,
      "epoch": 0.3236857363741786,
      "grad_norm": 0.0,
      "learning_rate": 3.3818129107073826e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6699
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 139.0,
      "epoch": 0.3237340548898338,
      "grad_norm": 4.377615750571351,
      "learning_rate": 3.381571318129107e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6700
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 129.8125,
      "epoch": 0.323782373405489,
      "grad_norm": 12.41301689768289,
      "learning_rate": 3.381329725550831e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.30935919284820557,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6701
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 132.0,
      "epoch": 0.32383069192114416,
      "grad_norm": 0.0,
      "learning_rate": 3.381088132972555e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6702
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 135.6875,
      "epoch": 0.3238790104367994,
      "grad_norm": 0.0,
      "learning_rate": 3.3808465403942786e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6703
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 116.4375,
      "epoch": 0.3239273289524546,
      "grad_norm": 0.0,
      "learning_rate": 3.3806049478160033e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 6704
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 135.21875,
      "epoch": 0.3239756474681098,
      "grad_norm": 0.0,
      "learning_rate": 3.380363355237727e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6705
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 132.84375,
      "epoch": 0.324023965983765,
      "grad_norm": 0.0,
      "learning_rate": 3.380121762659451e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6706
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 127.96875,
      "epoch": 0.32407228449942016,
      "grad_norm": 0.0,
      "learning_rate": 3.379880170081175e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6707
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 149.15625,
      "epoch": 0.3241206030150754,
      "grad_norm": 0.0,
      "learning_rate": 3.379638577502899e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6708
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 138.71875,
      "epoch": 0.32416892153073057,
      "grad_norm": 0.0,
      "learning_rate": 3.379396984924623e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6709
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 119.03125,
      "epoch": 0.3242172400463858,
      "grad_norm": 0.0,
      "learning_rate": 3.3791553923463465e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 6710
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 144.65625,
      "epoch": 0.324265558562041,
      "grad_norm": 7.83321409041462,
      "learning_rate": 3.378913799768071e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6711
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 129.71875,
      "epoch": 0.32431387707769616,
      "grad_norm": 0.0,
      "learning_rate": 3.378672207189795e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6712
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 134.5625,
      "epoch": 0.3243621955933514,
      "grad_norm": 0.0,
      "learning_rate": 3.378430614611519e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6713
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 124.21875,
      "epoch": 0.32441051410900656,
      "grad_norm": 0.0,
      "learning_rate": 3.378189022033243e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6714
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 145.28125,
      "epoch": 0.3244588326246618,
      "grad_norm": 0.0,
      "learning_rate": 3.377947429454967e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6715
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 131.46875,
      "epoch": 0.324507151140317,
      "grad_norm": 0.0,
      "learning_rate": 3.377705836876691e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6716
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 134.78125,
      "epoch": 0.32455546965597215,
      "grad_norm": 0.0,
      "learning_rate": 3.3774642442984155e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6717
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 125.28125,
      "epoch": 0.3246037881716274,
      "grad_norm": 0.0,
      "learning_rate": 3.377222651720139e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6718
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.78125,
      "epoch": 0.32465210668728256,
      "grad_norm": 0.0,
      "learning_rate": 3.3769810591418627e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6719
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 124.5625,
      "epoch": 0.3247004252029378,
      "grad_norm": 0.0,
      "learning_rate": 3.376739466563587e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6720
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 148.625,
      "epoch": 0.32474874371859297,
      "grad_norm": 0.0,
      "learning_rate": 3.376497873985311e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6721
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 133.8125,
      "epoch": 0.32479706223424815,
      "grad_norm": 0.0,
      "learning_rate": 3.376256281407035e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 6722
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 144.78125,
      "epoch": 0.3248453807499034,
      "grad_norm": 0.0,
      "learning_rate": 3.3760146888287587e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6723
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 135.125,
      "epoch": 0.32489369926555856,
      "grad_norm": 0.0,
      "learning_rate": 3.3757730962504834e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6724
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 141.6875,
      "epoch": 0.3249420177812138,
      "grad_norm": 0.0,
      "learning_rate": 3.375531503672207e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6725
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 121.96875,
      "epoch": 0.32499033629686896,
      "grad_norm": 0.0,
      "learning_rate": 3.375289911093931e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6726
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 125.875,
      "epoch": 0.32503865481252414,
      "grad_norm": 0.0,
      "learning_rate": 3.375048318515655e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6727
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 113.5625,
      "epoch": 0.3250869733281794,
      "grad_norm": 0.0,
      "learning_rate": 3.3748067259373794e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6728
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 134.28125,
      "epoch": 0.32513529184383455,
      "grad_norm": 0.0,
      "learning_rate": 3.374565133359103e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6729
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 127.875,
      "epoch": 0.3251836103594898,
      "grad_norm": 0.0,
      "learning_rate": 3.3743235407808266e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6730
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 140.03125,
      "epoch": 0.32523192887514496,
      "grad_norm": 0.0,
      "learning_rate": 3.3740819482025513e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6731
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 136.125,
      "epoch": 0.32528024739080014,
      "grad_norm": 0.0,
      "learning_rate": 3.373840355624275e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6732
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 123.875,
      "epoch": 0.32532856590645537,
      "grad_norm": 0.0,
      "learning_rate": 3.373598763045999e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 6733
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.9375,
      "epoch": 0.32537688442211055,
      "grad_norm": 0.0,
      "learning_rate": 3.373357170467723e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6734
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 139.25,
      "epoch": 0.3254252029377658,
      "grad_norm": 0.0,
      "learning_rate": 3.3731155778894473e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6735
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 138.09375,
      "epoch": 0.32547352145342096,
      "grad_norm": 0.0,
      "learning_rate": 3.372873985311171e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6736
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.03125,
      "epoch": 0.32552183996907613,
      "grad_norm": 0.0,
      "learning_rate": 3.3726323927328956e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6737
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 129.875,
      "epoch": 0.32557015848473136,
      "grad_norm": 0.0,
      "learning_rate": 3.372390800154619e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6738
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 136.96875,
      "epoch": 0.32561847700038654,
      "grad_norm": 0.0,
      "learning_rate": 3.372149207576343e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6739
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 124.21875,
      "epoch": 0.3256667955160418,
      "grad_norm": 0.0,
      "learning_rate": 3.371907614998067e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6740
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 125.375,
      "epoch": 0.32571511403169695,
      "grad_norm": 0.0,
      "learning_rate": 3.371666022419791e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6741
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 138.96875,
      "epoch": 0.3257634325473521,
      "grad_norm": 0.0,
      "learning_rate": 3.371424429841515e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6742
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.40625,
      "epoch": 0.32581175106300736,
      "grad_norm": 0.0,
      "learning_rate": 3.371182837263239e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6743
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 178.65625,
      "epoch": 0.32586006957866254,
      "grad_norm": 12.016040581084532,
      "learning_rate": 3.3709412446849635e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.27598220109939575,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 0.96875,
      "step": 6744
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 149.78125,
      "epoch": 0.32590838809431777,
      "grad_norm": 0.0,
      "learning_rate": 3.370699652106687e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6745
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 140.5625,
      "epoch": 0.32595670660997295,
      "grad_norm": 0.0,
      "learning_rate": 3.370458059528411e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6746
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 128.34375,
      "epoch": 0.3260050251256281,
      "grad_norm": 0.0,
      "learning_rate": 3.3702164669501353e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6747
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 179.40625,
      "epoch": 0.32605334364128336,
      "grad_norm": 0.0,
      "learning_rate": 3.3699748743718595e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6748
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 135.0625,
      "epoch": 0.32610166215693853,
      "grad_norm": 0.0,
      "learning_rate": 3.369733281793583e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6749
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 125.78125,
      "epoch": 0.32614998067259376,
      "grad_norm": 8.825677185115433,
      "learning_rate": 3.3694916892153067e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6750
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 129.4375,
      "epoch": 0.32619829918824894,
      "grad_norm": 0.0,
      "learning_rate": 3.3692500966370314e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6751
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 140.5,
      "epoch": 0.3262466177039041,
      "grad_norm": 11.292230453992639,
      "learning_rate": 3.369008504058755e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.24251843988895416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6752
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 129.53125,
      "epoch": 0.32629493621955935,
      "grad_norm": 0.0,
      "learning_rate": 3.368766911480479e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6753
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 135.59375,
      "epoch": 0.3263432547352145,
      "grad_norm": 0.0,
      "learning_rate": 3.368525318902203e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6754
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 132.46875,
      "epoch": 0.32639157325086976,
      "grad_norm": 0.0,
      "learning_rate": 3.3682837263239274e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 6755
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 129.25,
      "epoch": 0.32643989176652494,
      "grad_norm": 0.0,
      "learning_rate": 3.368042133745651e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6756
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 130.25,
      "epoch": 0.3264882102821801,
      "grad_norm": 0.0,
      "learning_rate": 3.3678005411673756e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6757
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 127.09375,
      "epoch": 0.32653652879783535,
      "grad_norm": 0.0,
      "learning_rate": 3.367558948589099e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6758
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 139.6875,
      "epoch": 0.3265848473134905,
      "grad_norm": 0.0,
      "learning_rate": 3.367317356010823e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.260896235704422,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6759
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.4375,
      "epoch": 0.32663316582914576,
      "grad_norm": 0.0,
      "learning_rate": 3.367075763432547e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6760
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 130.21875,
      "epoch": 0.32668148434480093,
      "grad_norm": 0.0,
      "learning_rate": 3.366834170854271e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 6761
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 132.53125,
      "epoch": 0.3267298028604561,
      "grad_norm": 0.0,
      "learning_rate": 3.366592578275995e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6762
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 132.0,
      "epoch": 0.32677812137611134,
      "grad_norm": 0.0,
      "learning_rate": 3.366350985697719e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6763
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 148.53125,
      "epoch": 0.3268264398917665,
      "grad_norm": 0.0,
      "learning_rate": 3.3661093931194435e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6764
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 119.3125,
      "epoch": 0.32687475840742175,
      "grad_norm": 0.0,
      "learning_rate": 3.365867800541167e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6765
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 132.65625,
      "epoch": 0.3269230769230769,
      "grad_norm": 0.0,
      "learning_rate": 3.3656262079628913e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6766
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 128.0625,
      "epoch": 0.3269713954387321,
      "grad_norm": 0.0,
      "learning_rate": 3.3653846153846154e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6767
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 131.21875,
      "epoch": 0.32701971395438734,
      "grad_norm": 0.0,
      "learning_rate": 3.3651430228063396e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6768
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 118.28125,
      "epoch": 0.3270680324700425,
      "grad_norm": 0.0,
      "learning_rate": 3.364901430228063e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6769
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 130.53125,
      "epoch": 0.32711635098569775,
      "grad_norm": 0.0,
      "learning_rate": 3.364659837649787e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6770
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 128.96875,
      "epoch": 0.3271646695013529,
      "grad_norm": 0.0,
      "learning_rate": 3.3644182450715114e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6771
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 130.25,
      "epoch": 0.3272129880170081,
      "grad_norm": 0.0,
      "learning_rate": 3.364176652493235e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6772
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 131.90625,
      "epoch": 0.32726130653266333,
      "grad_norm": 0.0,
      "learning_rate": 3.363935059914959e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6773
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 127.65625,
      "epoch": 0.3273096250483185,
      "grad_norm": 0.0,
      "learning_rate": 3.3636934673366833e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6774
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 118.5625,
      "epoch": 0.32735794356397374,
      "grad_norm": 0.0,
      "learning_rate": 3.3634518747584074e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6775
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 129.28125,
      "epoch": 0.3274062620796289,
      "grad_norm": 0.0,
      "learning_rate": 3.363210282180131e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6776
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.78125,
      "epoch": 0.3274545805952841,
      "grad_norm": 0.0,
      "learning_rate": 3.3629686896018557e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6777
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 125.75,
      "epoch": 0.3275028991109393,
      "grad_norm": 0.0,
      "learning_rate": 3.3627270970235793e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6778
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 141.3125,
      "epoch": 0.3275512176265945,
      "grad_norm": 0.0,
      "learning_rate": 3.3624855044453035e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6779
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 124.46875,
      "epoch": 0.32759953614224974,
      "grad_norm": 0.0,
      "learning_rate": 3.362243911867027e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6780
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 129.78125,
      "epoch": 0.3276478546579049,
      "grad_norm": 12.252849122478134,
      "learning_rate": 3.362002319288751e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6781
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.75,
      "epoch": 0.3276961731735601,
      "grad_norm": 0.0,
      "learning_rate": 3.3617607267104753e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6782
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 132.4375,
      "epoch": 0.3277444916892153,
      "grad_norm": 0.0,
      "learning_rate": 3.361519134132199e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6783
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 196.59375,
      "epoch": 0.3277928102048705,
      "grad_norm": 2.2966064929804864,
      "learning_rate": 3.3612775415539236e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 0.96875,
      "step": 6784
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 118.3125,
      "epoch": 0.32784112872052573,
      "grad_norm": 0.0,
      "learning_rate": 3.361035948975647e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 6785
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 145.0625,
      "epoch": 0.3278894472361809,
      "grad_norm": 0.0,
      "learning_rate": 3.3607943563973714e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6786
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 112.1875,
      "epoch": 0.3279377657518361,
      "grad_norm": 57.7172168752194,
      "learning_rate": 3.3605527638190955e-07,
      "loss": -0.0,
      "reward": 2.328125,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6787
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 136.75,
      "epoch": 0.3279860842674913,
      "grad_norm": 0.0,
      "learning_rate": 3.3603111712408196e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6788
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 131.5625,
      "epoch": 0.3280344027831465,
      "grad_norm": 0.0,
      "learning_rate": 3.360069578662543e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6789
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 132.125,
      "epoch": 0.3280827212988017,
      "grad_norm": 0.0,
      "learning_rate": 3.359827986084267e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6790
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 128.6875,
      "epoch": 0.3281310398144569,
      "grad_norm": 10.764733851556512,
      "learning_rate": 3.3595863935059915e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6791
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 143.4375,
      "epoch": 0.3281793583301121,
      "grad_norm": 7.3628710184641735,
      "learning_rate": 3.359344800927715e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.30509042739868164,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6792
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 124.4375,
      "epoch": 0.3282276768457673,
      "grad_norm": 0.0,
      "learning_rate": 3.359103208349439e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6793
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 123.6875,
      "epoch": 0.3282759953614225,
      "grad_norm": 0.0,
      "learning_rate": 3.3588616157711634e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6794
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 118.25,
      "epoch": 0.3283243138770777,
      "grad_norm": 0.0,
      "learning_rate": 3.3586200231928875e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6795
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.15625,
      "epoch": 0.3283726323927329,
      "grad_norm": 0.0,
      "learning_rate": 3.358378430614611e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6796
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 118.21875,
      "epoch": 0.3284209509083881,
      "grad_norm": 0.0,
      "learning_rate": 3.358136838036336e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6797
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.90625,
      "epoch": 0.3284692694240433,
      "grad_norm": 0.0,
      "learning_rate": 3.3578952454580594e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6798
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 143.0625,
      "epoch": 0.3285175879396985,
      "grad_norm": 0.0,
      "learning_rate": 3.3576536528797835e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6799
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 130.9375,
      "epoch": 0.3285659064553537,
      "grad_norm": 0.0,
      "learning_rate": 3.357412060301507e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6800
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 130.0625,
      "epoch": 0.3286142249710089,
      "grad_norm": 0.0,
      "learning_rate": 3.3571704677232313e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6801
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 164.0625,
      "epoch": 0.32866254348666407,
      "grad_norm": 1.5639180492526668,
      "learning_rate": 3.3569288751449554e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.27883464097976685,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 0.96875,
      "step": 6802
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.59375,
      "epoch": 0.3287108620023193,
      "grad_norm": 0.0,
      "learning_rate": 3.356687282566679e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6803
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.5,
      "epoch": 0.3287591805179745,
      "grad_norm": 0.0,
      "learning_rate": 3.3564456899884037e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6804
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 145.90625,
      "epoch": 0.3288074990336297,
      "grad_norm": 1.497673946605784,
      "learning_rate": 3.3562040974101273e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.2651650309562683,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 0.96875,
      "step": 6805
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.15625,
      "epoch": 0.3288558175492849,
      "grad_norm": 0.0,
      "learning_rate": 3.3559625048318514e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6806
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 124.375,
      "epoch": 0.32890413606494007,
      "grad_norm": 0.0,
      "learning_rate": 3.3557209122535756e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6807
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 132.5,
      "epoch": 0.3289524545805953,
      "grad_norm": 0.0,
      "learning_rate": 3.3554793196752997e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6808
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.03125,
      "epoch": 0.3290007730962505,
      "grad_norm": 0.0,
      "learning_rate": 3.3552377270970233e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6809
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 129.53125,
      "epoch": 0.3290490916119057,
      "grad_norm": 0.0,
      "learning_rate": 3.3549961345187475e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6810
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.625,
      "epoch": 0.3290974101275609,
      "grad_norm": 0.0,
      "learning_rate": 3.3547545419404716e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6811
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 134.0625,
      "epoch": 0.32914572864321606,
      "grad_norm": 0.0,
      "learning_rate": 3.354512949362195e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 6812
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 129.71875,
      "epoch": 0.3291940471588713,
      "grad_norm": 0.0,
      "learning_rate": 3.3542713567839193e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6813
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.78125,
      "epoch": 0.32924236567452647,
      "grad_norm": 0.0,
      "learning_rate": 3.3540297642056435e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6814
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 131.21875,
      "epoch": 0.3292906841901817,
      "grad_norm": 10.278846339086222,
      "learning_rate": 3.3537881716273676e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.2461063265800476,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6815
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 120.09375,
      "epoch": 0.3293390027058369,
      "grad_norm": 0.0,
      "learning_rate": 3.353546579049091e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 6816
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.53125,
      "epoch": 0.32938732122149206,
      "grad_norm": 0.0,
      "learning_rate": 3.353304986470816e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6817
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 123.0,
      "epoch": 0.3294356397371473,
      "grad_norm": 0.0,
      "learning_rate": 3.3530633938925395e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6818
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.0,
      "epoch": 0.32948395825280247,
      "grad_norm": 0.0,
      "learning_rate": 3.3528218013142636e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6819
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.5625,
      "epoch": 0.3295322767684577,
      "grad_norm": 0.0,
      "learning_rate": 3.352580208735987e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6820
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 131.0625,
      "epoch": 0.3295805952841129,
      "grad_norm": 0.0,
      "learning_rate": 3.352338616157712e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6821
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.90625,
      "epoch": 0.32962891379976805,
      "grad_norm": 0.0,
      "learning_rate": 3.3520970235794355e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6822
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 139.9375,
      "epoch": 0.3296772323154233,
      "grad_norm": 0.0,
      "learning_rate": 3.351855431001159e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6823
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 122.4375,
      "epoch": 0.32972555083107846,
      "grad_norm": 0.0,
      "learning_rate": 3.351613838422884e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 6824
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 115.90625,
      "epoch": 0.3297738693467337,
      "grad_norm": 3.1261608968392123,
      "learning_rate": 3.3513722458446074e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6825
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 132.25,
      "epoch": 0.32982218786238887,
      "grad_norm": 17.030327755393756,
      "learning_rate": 3.3511306532663315e-07,
      "loss": -0.0,
      "reward": 1.734375,
      "reward_std": 0.24831002950668335,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6826
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.125,
      "epoch": 0.32987050637804405,
      "grad_norm": 0.0,
      "learning_rate": 3.3508890606880557e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.2130674123764038,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 0.96875,
      "step": 6827
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.78125,
      "epoch": 0.3299188248936993,
      "grad_norm": 0.0,
      "learning_rate": 3.35064746810978e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6828
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 120.9375,
      "epoch": 0.32996714340935446,
      "grad_norm": 0.0,
      "learning_rate": 3.3504058755315034e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6829
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 125.28125,
      "epoch": 0.3300154619250097,
      "grad_norm": 0.0,
      "learning_rate": 3.3501642829532275e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6830
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 130.34375,
      "epoch": 0.33006378044066487,
      "grad_norm": 0.0,
      "learning_rate": 3.3499226903749517e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6831
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.21875,
      "epoch": 0.33011209895632004,
      "grad_norm": 0.0,
      "learning_rate": 3.3496810977966753e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6832
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.4375,
      "epoch": 0.3301604174719753,
      "grad_norm": 0.0,
      "learning_rate": 3.3494395052183994e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6833
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.96875,
      "epoch": 0.33020873598763045,
      "grad_norm": 0.0,
      "learning_rate": 3.3491979126401235e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6834
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.84375,
      "epoch": 0.3302570545032857,
      "grad_norm": 0.0,
      "learning_rate": 3.3489563200618477e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6835
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.125,
      "epoch": 0.33030537301894086,
      "grad_norm": 0.0,
      "learning_rate": 3.3487147274835713e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6836
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 114.84375,
      "epoch": 0.33035369153459604,
      "grad_norm": 119.27340350539866,
      "learning_rate": 3.348473134905296e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.23827512562274933,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6837
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 129.0625,
      "epoch": 0.33040201005025127,
      "grad_norm": 12.67367168306734,
      "learning_rate": 3.3482315423270196e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6838
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.5,
      "epoch": 0.33045032856590645,
      "grad_norm": 0.0,
      "learning_rate": 3.3479899497487437e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6839
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.4375,
      "epoch": 0.3304986470815617,
      "grad_norm": 0.0,
      "learning_rate": 3.3477483571704673e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6840
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 127.34375,
      "epoch": 0.33054696559721686,
      "grad_norm": 0.0,
      "learning_rate": 3.347506764592192e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6841
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 115.28125,
      "epoch": 0.33059528411287203,
      "grad_norm": 4.82330427305433,
      "learning_rate": 3.3472651720139156e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6842
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 123.09375,
      "epoch": 0.33064360262852727,
      "grad_norm": 0.0,
      "learning_rate": 3.347023579435639e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6843
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.28125,
      "epoch": 0.33069192114418244,
      "grad_norm": 0.0,
      "learning_rate": 3.346781986857364e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6844
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.59375,
      "epoch": 0.3307402396598377,
      "grad_norm": 0.0,
      "learning_rate": 3.3465403942790875e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6845
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.28125,
      "epoch": 0.33078855817549285,
      "grad_norm": 0.0,
      "learning_rate": 3.3462988017008116e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6846
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.46875,
      "epoch": 0.33083687669114803,
      "grad_norm": 0.0,
      "learning_rate": 3.3460572091225357e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6847
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 111.84375,
      "epoch": 0.33088519520680326,
      "grad_norm": 0.0,
      "learning_rate": 3.34581561654426e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6848
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.15625,
      "epoch": 0.33093351372245844,
      "grad_norm": 0.0,
      "learning_rate": 3.3455740239659835e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6849
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.96875,
      "epoch": 0.33098183223811367,
      "grad_norm": 0.0,
      "learning_rate": 3.3453324313877076e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6850
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.75,
      "epoch": 0.33103015075376885,
      "grad_norm": 0.0,
      "learning_rate": 3.345090838809432e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6851
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.6875,
      "epoch": 0.331078469269424,
      "grad_norm": 0.0,
      "learning_rate": 3.3448492462311554e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6852
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.4375,
      "epoch": 0.33112678778507926,
      "grad_norm": 0.0,
      "learning_rate": 3.3446076536528795e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6853
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 99.375,
      "epoch": 0.33117510630073443,
      "grad_norm": 22.044100958984068,
      "learning_rate": 3.3443660610746036e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2619796097278595,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6854
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.15625,
      "epoch": 0.33122342481638967,
      "grad_norm": 0.0,
      "learning_rate": 3.344124468496328e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6855
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.125,
      "epoch": 0.33127174333204484,
      "grad_norm": 0.0,
      "learning_rate": 3.3438828759180514e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 6856
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.53125,
      "epoch": 0.3313200618477,
      "grad_norm": 0.0,
      "learning_rate": 3.343641283339776e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6857
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.78125,
      "epoch": 0.33136838036335525,
      "grad_norm": 0.0,
      "learning_rate": 3.3433996907614996e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6858
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.03125,
      "epoch": 0.33141669887901043,
      "grad_norm": 0.0,
      "learning_rate": 3.343158098183224e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6859
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.28125,
      "epoch": 0.33146501739466566,
      "grad_norm": 0.0,
      "learning_rate": 3.3429165056049474e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6860
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.34375,
      "epoch": 0.33151333591032084,
      "grad_norm": 0.0,
      "learning_rate": 3.342674913026672e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.19044628739356995,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6861
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 91.96875,
      "epoch": 0.331561654425976,
      "grad_norm": 15.467658194540912,
      "learning_rate": 3.3424333204483957e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6862
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.8125,
      "epoch": 0.33160997294163125,
      "grad_norm": 0.0,
      "learning_rate": 3.342191727870119e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6863
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.8125,
      "epoch": 0.3316582914572864,
      "grad_norm": 0.0,
      "learning_rate": 3.341950135291844e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6864
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.65625,
      "epoch": 0.33170660997294166,
      "grad_norm": 0.0,
      "learning_rate": 3.3417085427135675e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6865
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.21875,
      "epoch": 0.33175492848859683,
      "grad_norm": 0.0,
      "learning_rate": 3.3414669501352917e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6866
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.3125,
      "epoch": 0.331803247004252,
      "grad_norm": 0.0,
      "learning_rate": 3.341225357557016e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6867
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.75,
      "epoch": 0.33185156551990724,
      "grad_norm": 0.0,
      "learning_rate": 3.34098376497874e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6868
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.375,
      "epoch": 0.3318998840355624,
      "grad_norm": 0.0,
      "learning_rate": 3.3407421724004636e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6869
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.46875,
      "epoch": 0.33194820255121765,
      "grad_norm": 0.0,
      "learning_rate": 3.3405005798221877e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6870
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.09375,
      "epoch": 0.33199652106687283,
      "grad_norm": 0.0,
      "learning_rate": 3.340258987243912e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6871
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.4375,
      "epoch": 0.332044839582528,
      "grad_norm": 0.0,
      "learning_rate": 3.340017394665636e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6872
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.9375,
      "epoch": 0.33209315809818324,
      "grad_norm": 0.0,
      "learning_rate": 3.3397758020873596e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 6873
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.0,
      "epoch": 0.3321414766138384,
      "grad_norm": 0.0,
      "learning_rate": 3.3395342095090837e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6874
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.1875,
      "epoch": 0.33218979512949365,
      "grad_norm": 0.0,
      "learning_rate": 3.339292616930808e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6875
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.375,
      "epoch": 0.3322381136451488,
      "grad_norm": 0.0,
      "learning_rate": 3.3390510243525314e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6876
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.4375,
      "epoch": 0.332286432160804,
      "grad_norm": 0.0,
      "learning_rate": 3.338809431774256e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6877
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.65625,
      "epoch": 0.33233475067645923,
      "grad_norm": 0.0,
      "learning_rate": 3.3385678391959797e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6878
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.53125,
      "epoch": 0.3323830691921144,
      "grad_norm": 0.0,
      "learning_rate": 3.338326246617704e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6879
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.78125,
      "epoch": 0.33243138770776964,
      "grad_norm": 0.0,
      "learning_rate": 3.3380846540394275e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6880
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.40625,
      "epoch": 0.3324797062234248,
      "grad_norm": 0.0,
      "learning_rate": 3.337843061461152e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6881
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.15625,
      "epoch": 0.33252802473908,
      "grad_norm": 0.0,
      "learning_rate": 3.337601468882876e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6882
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.03125,
      "epoch": 0.33257634325473523,
      "grad_norm": 0.0,
      "learning_rate": 3.3373598763045993e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6883
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.21875,
      "epoch": 0.3326246617703904,
      "grad_norm": 0.0,
      "learning_rate": 3.337118283726324e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6884
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.6875,
      "epoch": 0.33267298028604564,
      "grad_norm": 0.0,
      "learning_rate": 3.3368766911480476e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6885
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.0,
      "epoch": 0.3327212988017008,
      "grad_norm": 0.0,
      "learning_rate": 3.336635098569772e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6886
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.5,
      "epoch": 0.332769617317356,
      "grad_norm": 0.0,
      "learning_rate": 3.336393505991496e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6887
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 100.90625,
      "epoch": 0.3328179358330112,
      "grad_norm": 9.415192777349008,
      "learning_rate": 3.33615191341322e-07,
      "loss": -0.0,
      "reward": 2.421875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6888
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.71875,
      "epoch": 0.3328662543486664,
      "grad_norm": 0.0,
      "learning_rate": 3.3359103208349436e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6889
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.875,
      "epoch": 0.33291457286432163,
      "grad_norm": 0.0,
      "learning_rate": 3.335668728256668e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6890
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.84375,
      "epoch": 0.3329628913799768,
      "grad_norm": 0.0,
      "learning_rate": 3.335427135678392e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6891
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.5,
      "epoch": 0.333011209895632,
      "grad_norm": 0.0,
      "learning_rate": 3.335185543100116e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6892
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.1875,
      "epoch": 0.3330595284112872,
      "grad_norm": 0.0,
      "learning_rate": 3.3349439505218396e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6893
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.3125,
      "epoch": 0.3331078469269424,
      "grad_norm": 0.0,
      "learning_rate": 3.334702357943564e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6894
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 98.28125,
      "epoch": 0.33315616544259763,
      "grad_norm": 0.0,
      "learning_rate": 3.334460765365288e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6895
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.53125,
      "epoch": 0.3332044839582528,
      "grad_norm": 0.0,
      "learning_rate": 3.3342191727870115e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6896
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.5625,
      "epoch": 0.333252802473908,
      "grad_norm": 0.0,
      "learning_rate": 3.333977580208736e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6897
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.9375,
      "epoch": 0.3333011209895632,
      "grad_norm": 0.0,
      "learning_rate": 3.33373598763046e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6898
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.3125,
      "epoch": 0.3333494395052184,
      "grad_norm": 0.0,
      "learning_rate": 3.333494395052184e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 6899
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.8125,
      "epoch": 0.3333977580208736,
      "grad_norm": 0.0,
      "learning_rate": 3.3332528024739075e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6900
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.71875,
      "epoch": 0.3334460765365288,
      "grad_norm": 0.0,
      "learning_rate": 3.333011209895632e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6901
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.6875,
      "epoch": 0.333494395052184,
      "grad_norm": 0.0,
      "learning_rate": 3.332769617317356e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6902
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.9375,
      "epoch": 0.3335427135678392,
      "grad_norm": 0.0,
      "learning_rate": 3.3325280247390794e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6903
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.1875,
      "epoch": 0.3335910320834944,
      "grad_norm": 5.729691297583013,
      "learning_rate": 3.332286432160804e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.21316051483154297,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6904
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.75,
      "epoch": 0.3336393505991496,
      "grad_norm": 0.0,
      "learning_rate": 3.3320448395825277e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6905
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.0,
      "epoch": 0.3336876691148048,
      "grad_norm": 0.0,
      "learning_rate": 3.331803247004252e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 6906
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.0625,
      "epoch": 0.33373598763046,
      "grad_norm": 0.0,
      "learning_rate": 3.331561654425976e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6907
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.40625,
      "epoch": 0.3337843061461152,
      "grad_norm": 0.0,
      "learning_rate": 3.3313200618477e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6908
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.28125,
      "epoch": 0.3338326246617704,
      "grad_norm": 0.0,
      "learning_rate": 3.3310784692694237e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6909
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.53125,
      "epoch": 0.3338809431774256,
      "grad_norm": 0.0,
      "learning_rate": 3.330836876691148e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6910
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.0,
      "epoch": 0.3339292616930808,
      "grad_norm": 0.0,
      "learning_rate": 3.330595284112872e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6911
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.625,
      "epoch": 0.33397758020873597,
      "grad_norm": 0.0,
      "learning_rate": 3.330353691534596e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6912
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.46875,
      "epoch": 0.3340258987243912,
      "grad_norm": 0.0,
      "learning_rate": 3.3301120989563197e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6913
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.78125,
      "epoch": 0.3340742172400464,
      "grad_norm": 0.0,
      "learning_rate": 3.329870506378044e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6914
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.125,
      "epoch": 0.3341225357557016,
      "grad_norm": 0.0,
      "learning_rate": 3.329628913799768e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6915
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.59375,
      "epoch": 0.3341708542713568,
      "grad_norm": 0.0,
      "learning_rate": 3.3293873212214916e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6916
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.84375,
      "epoch": 0.33421917278701196,
      "grad_norm": 0.0,
      "learning_rate": 3.3291457286432163e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6917
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.25,
      "epoch": 0.3342674913026672,
      "grad_norm": 0.0,
      "learning_rate": 3.32890413606494e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6918
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.53125,
      "epoch": 0.3343158098183224,
      "grad_norm": 0.0,
      "learning_rate": 3.328662543486664e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6919
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.5625,
      "epoch": 0.3343641283339776,
      "grad_norm": 0.0,
      "learning_rate": 3.3284209509083876e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6920
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 103.0,
      "epoch": 0.3344124468496328,
      "grad_norm": 0.0,
      "learning_rate": 3.3281793583301123e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 6921
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.53125,
      "epoch": 0.33446076536528796,
      "grad_norm": 0.0,
      "learning_rate": 3.327937765751836e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6922
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.03125,
      "epoch": 0.3345090838809432,
      "grad_norm": 0.0,
      "learning_rate": 3.32769617317356e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6923
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.71875,
      "epoch": 0.33455740239659837,
      "grad_norm": 0.0,
      "learning_rate": 3.327454580595284e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6924
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.9375,
      "epoch": 0.3346057209122536,
      "grad_norm": 0.0,
      "learning_rate": 3.327212988017008e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6925
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.375,
      "epoch": 0.3346540394279088,
      "grad_norm": 0.0,
      "learning_rate": 3.326971395438732e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6926
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.46875,
      "epoch": 0.33470235794356396,
      "grad_norm": 0.0,
      "learning_rate": 3.326729802860456e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6927
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.75,
      "epoch": 0.3347506764592192,
      "grad_norm": 0.0,
      "learning_rate": 3.32648821028218e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6928
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.625,
      "epoch": 0.33479899497487436,
      "grad_norm": 0.0,
      "learning_rate": 3.326246617703904e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6929
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.3125,
      "epoch": 0.3348473134905296,
      "grad_norm": 0.0,
      "learning_rate": 3.326005025125628e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6930
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.96875,
      "epoch": 0.3348956320061848,
      "grad_norm": 0.0,
      "learning_rate": 3.325763432547352e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6931
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.09375,
      "epoch": 0.33494395052183995,
      "grad_norm": 0.0,
      "learning_rate": 3.325521839969076e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6932
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 104.625,
      "epoch": 0.3349922690374952,
      "grad_norm": 7.509643029773759,
      "learning_rate": 3.3252802473908e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.23827512562274933,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6933
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 111.78125,
      "epoch": 0.33504058755315036,
      "grad_norm": 0.0,
      "learning_rate": 3.3250386548125245e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 6934
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.15625,
      "epoch": 0.3350889060688056,
      "grad_norm": 0.0,
      "learning_rate": 3.324797062234248e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6935
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.03125,
      "epoch": 0.33513722458446077,
      "grad_norm": 0.0,
      "learning_rate": 3.3245554696559717e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6936
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 103.59375,
      "epoch": 0.33518554310011595,
      "grad_norm": 0.0,
      "learning_rate": 3.3243138770776963e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6937
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.125,
      "epoch": 0.3352338616157712,
      "grad_norm": 0.0,
      "learning_rate": 3.32407228449942e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6938
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 99.3125,
      "epoch": 0.33528218013142636,
      "grad_norm": 4.7083345649673864,
      "learning_rate": 3.323830691921144e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6939
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.65625,
      "epoch": 0.3353304986470816,
      "grad_norm": 0.0,
      "learning_rate": 3.3235890993428677e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6940
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.3125,
      "epoch": 0.33537881716273676,
      "grad_norm": 0.0,
      "learning_rate": 3.3233475067645924e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6941
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.09375,
      "epoch": 0.33542713567839194,
      "grad_norm": 0.0,
      "learning_rate": 3.323105914186316e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6942
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.3125,
      "epoch": 0.3354754541940472,
      "grad_norm": 0.0,
      "learning_rate": 3.32286432160804e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6943
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.3125,
      "epoch": 0.33552377270970235,
      "grad_norm": 0.0,
      "learning_rate": 3.322622729029764e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6944
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.375,
      "epoch": 0.3355720912253576,
      "grad_norm": 0.0,
      "learning_rate": 3.322381136451488e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.2041158676147461,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 6945
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 88.25,
      "epoch": 0.33562040974101276,
      "grad_norm": 33.140646335790166,
      "learning_rate": 3.322139543873212e-07,
      "loss": -0.0,
      "reward": 1.421875,
      "reward_std": 0.20633062720298767,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 6946
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 85.28125,
      "epoch": 0.33566872825666794,
      "grad_norm": 16.18080997983362,
      "learning_rate": 3.321897951294936e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6947
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.0,
      "epoch": 0.33571704677232317,
      "grad_norm": 0.0,
      "learning_rate": 3.32165635871666e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6948
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.21875,
      "epoch": 0.33576536528797835,
      "grad_norm": 0.0,
      "learning_rate": 3.321414766138384e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6949
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.28125,
      "epoch": 0.3358136838036336,
      "grad_norm": 0.0,
      "learning_rate": 3.321173173560108e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6950
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.9375,
      "epoch": 0.33586200231928875,
      "grad_norm": 0.0,
      "learning_rate": 3.320931580981832e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 6951
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.59375,
      "epoch": 0.33591032083494393,
      "grad_norm": 19.93624496856053,
      "learning_rate": 3.3206899884035563e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.23047873377799988,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6952
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.71875,
      "epoch": 0.33595863935059916,
      "grad_norm": 0.0,
      "learning_rate": 3.32044839582528e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 6953
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.0625,
      "epoch": 0.33600695786625434,
      "grad_norm": 21.334488383036987,
      "learning_rate": 3.3202068032470045e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6954
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.84375,
      "epoch": 0.3360552763819096,
      "grad_norm": 0.0,
      "learning_rate": 3.319965210668728e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 6955
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.96875,
      "epoch": 0.33610359489756475,
      "grad_norm": 0.0,
      "learning_rate": 3.319723618090452e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6956
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.625,
      "epoch": 0.3361519134132199,
      "grad_norm": 0.0,
      "learning_rate": 3.3194820255121764e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6957
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.28125,
      "epoch": 0.33620023192887516,
      "grad_norm": 0.0,
      "learning_rate": 3.3192404329339e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6958
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.875,
      "epoch": 0.33624855044453034,
      "grad_norm": 0.0,
      "learning_rate": 3.318998840355624e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6959
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.65625,
      "epoch": 0.33629686896018557,
      "grad_norm": 0.0,
      "learning_rate": 3.318757247777348e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6960
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.03125,
      "epoch": 0.33634518747584075,
      "grad_norm": 0.0,
      "learning_rate": 3.3185156551990724e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6961
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.25,
      "epoch": 0.3363935059914959,
      "grad_norm": 0.0,
      "learning_rate": 3.318274062620796e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6962
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.625,
      "epoch": 0.33644182450715115,
      "grad_norm": 0.0,
      "learning_rate": 3.31803247004252e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6963
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.5,
      "epoch": 0.33649014302280633,
      "grad_norm": 0.0,
      "learning_rate": 3.3177908774642443e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 6964
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.1875,
      "epoch": 0.33653846153846156,
      "grad_norm": 0.0,
      "learning_rate": 3.317549284885968e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6965
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.84375,
      "epoch": 0.33658678005411674,
      "grad_norm": 0.0,
      "learning_rate": 3.317307692307692e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6966
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.125,
      "epoch": 0.3366350985697719,
      "grad_norm": 0.0,
      "learning_rate": 3.317066099729416e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6967
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.375,
      "epoch": 0.33668341708542715,
      "grad_norm": 0.0,
      "learning_rate": 3.3168245071511403e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6968
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.96875,
      "epoch": 0.3367317356010823,
      "grad_norm": 0.0,
      "learning_rate": 3.316582914572864e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6969
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.84375,
      "epoch": 0.33678005411673756,
      "grad_norm": 0.0,
      "learning_rate": 3.316341321994588e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6970
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.78125,
      "epoch": 0.33682837263239274,
      "grad_norm": 0.0,
      "learning_rate": 3.316099729416312e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6971
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.25,
      "epoch": 0.3368766911480479,
      "grad_norm": 0.0,
      "learning_rate": 3.3158581368380364e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 6972
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.25,
      "epoch": 0.33692500966370315,
      "grad_norm": 0.0,
      "learning_rate": 3.31561654425976e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 6973
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.84375,
      "epoch": 0.3369733281793583,
      "grad_norm": 0.0,
      "learning_rate": 3.3153749516814846e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 6974
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.78125,
      "epoch": 0.33702164669501355,
      "grad_norm": 0.0,
      "learning_rate": 3.315133359103208e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 6975
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.1875,
      "epoch": 0.33706996521066873,
      "grad_norm": 0.0,
      "learning_rate": 3.314891766524932e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 6976
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.15625,
      "epoch": 0.3371182837263239,
      "grad_norm": 0.0,
      "learning_rate": 3.3146501739466565e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6977
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.15625,
      "epoch": 0.33716660224197914,
      "grad_norm": 0.0,
      "learning_rate": 3.31440858136838e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 6978
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.375,
      "epoch": 0.3372149207576343,
      "grad_norm": 0.0,
      "learning_rate": 3.314166988790104e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 6979
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.25,
      "epoch": 0.33726323927328955,
      "grad_norm": 0.0,
      "learning_rate": 3.313925396211828e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 6980
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.0625,
      "epoch": 0.3373115577889447,
      "grad_norm": 0.0,
      "learning_rate": 3.3136838036335525e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 6981
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.3125,
      "epoch": 0.3373598763045999,
      "grad_norm": 0.0,
      "learning_rate": 3.313442211055276e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 6982
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.78125,
      "epoch": 0.33740819482025514,
      "grad_norm": 0.0,
      "learning_rate": 3.3132006184770003e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6983
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.65625,
      "epoch": 0.3374565133359103,
      "grad_norm": 0.0,
      "learning_rate": 3.3129590258987244e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 6984
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.9375,
      "epoch": 0.33750483185156555,
      "grad_norm": 11.18522729157728,
      "learning_rate": 3.3127174333204485e-07,
      "loss": -0.0,
      "reward": 2.421875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 6985
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.15625,
      "epoch": 0.3375531503672207,
      "grad_norm": 0.0,
      "learning_rate": 3.312475840742172e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6986
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.5,
      "epoch": 0.3376014688828759,
      "grad_norm": 0.0,
      "learning_rate": 3.3122342481638963e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 6987
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.09375,
      "epoch": 0.33764978739853113,
      "grad_norm": 0.0,
      "learning_rate": 3.3119926555856204e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6988
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.25,
      "epoch": 0.3376981059141863,
      "grad_norm": 0.0,
      "learning_rate": 3.311751063007344e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6989
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.59375,
      "epoch": 0.33774642442984154,
      "grad_norm": 0.0,
      "learning_rate": 3.311509470429068e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 6990
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.875,
      "epoch": 0.3377947429454967,
      "grad_norm": 0.0,
      "learning_rate": 3.3112678778507923e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6991
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.03125,
      "epoch": 0.3378430614611519,
      "grad_norm": 0.0,
      "learning_rate": 3.3110262852725164e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6992
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.4375,
      "epoch": 0.3378913799768071,
      "grad_norm": 0.0,
      "learning_rate": 3.31078469269424e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 6993
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.5625,
      "epoch": 0.3379396984924623,
      "grad_norm": 0.0,
      "learning_rate": 3.3105431001159647e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 6994
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.65625,
      "epoch": 0.33798801700811754,
      "grad_norm": 0.0,
      "learning_rate": 3.3103015075376883e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 6995
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.78125,
      "epoch": 0.3380363355237727,
      "grad_norm": 0.0,
      "learning_rate": 3.310059914959412e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6996
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.90625,
      "epoch": 0.3380846540394279,
      "grad_norm": 0.0,
      "learning_rate": 3.3098183223811366e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 6997
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.25,
      "epoch": 0.3381329725550831,
      "grad_norm": 0.0,
      "learning_rate": 3.30957672980286e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 6998
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.84375,
      "epoch": 0.3381812910707383,
      "grad_norm": 0.0,
      "learning_rate": 3.3093351372245843e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 6999
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.6875,
      "epoch": 0.33822960958639353,
      "grad_norm": 0.0,
      "learning_rate": 3.309093544646308e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7000
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.53125,
      "epoch": 0.3382779281020487,
      "grad_norm": 0.0,
      "learning_rate": 3.3088519520680326e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7001
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.03125,
      "epoch": 0.3383262466177039,
      "grad_norm": 0.0,
      "learning_rate": 3.308610359489756e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7002
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.78125,
      "epoch": 0.3383745651333591,
      "grad_norm": 0.0,
      "learning_rate": 3.3083687669114803e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7003
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.40625,
      "epoch": 0.3384228836490143,
      "grad_norm": 0.0,
      "learning_rate": 3.3081271743332045e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 7004
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.625,
      "epoch": 0.3384712021646695,
      "grad_norm": 0.0,
      "learning_rate": 3.3078855817549286e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7005
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.96875,
      "epoch": 0.3385195206803247,
      "grad_norm": 0.0,
      "learning_rate": 3.307643989176652e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7006
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.53125,
      "epoch": 0.3385678391959799,
      "grad_norm": 0.0,
      "learning_rate": 3.3074023965983764e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7007
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.65625,
      "epoch": 0.3386161577116351,
      "grad_norm": 0.0,
      "learning_rate": 3.3071608040201005e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7008
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.4375,
      "epoch": 0.3386644762272903,
      "grad_norm": 0.0,
      "learning_rate": 3.306919211441824e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7009
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.5,
      "epoch": 0.3387127947429455,
      "grad_norm": 6.397086581643783,
      "learning_rate": 3.306677618863548e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7010
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.21875,
      "epoch": 0.3387611132586007,
      "grad_norm": 0.0,
      "learning_rate": 3.3064360262852724e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7011
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.625,
      "epoch": 0.3388094317742559,
      "grad_norm": 0.0,
      "learning_rate": 3.3061944337069965e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7012
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.71875,
      "epoch": 0.3388577502899111,
      "grad_norm": 0.0,
      "learning_rate": 3.30595284112872e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7013
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.75,
      "epoch": 0.3389060688055663,
      "grad_norm": 0.0,
      "learning_rate": 3.305711248550445e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7014
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.90625,
      "epoch": 0.3389543873212215,
      "grad_norm": 0.0,
      "learning_rate": 3.3054696559721684e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7015
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.59375,
      "epoch": 0.3390027058368767,
      "grad_norm": 0.0,
      "learning_rate": 3.3052280633938925e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7016
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.0625,
      "epoch": 0.33905102435253187,
      "grad_norm": 29.22196889314011,
      "learning_rate": 3.3049864708156167e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.20633062720298767,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7017
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.75,
      "epoch": 0.3390993428681871,
      "grad_norm": 0.0,
      "learning_rate": 3.3047448782373403e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7018
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.0625,
      "epoch": 0.3391476613838423,
      "grad_norm": 0.0,
      "learning_rate": 3.3045032856590644e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7019
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.21875,
      "epoch": 0.3391959798994975,
      "grad_norm": 0.0,
      "learning_rate": 3.304261693080788e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7020
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.375,
      "epoch": 0.3392442984151527,
      "grad_norm": 0.0,
      "learning_rate": 3.3040201005025127e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7021
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.65625,
      "epoch": 0.33929261693080787,
      "grad_norm": 0.0,
      "learning_rate": 3.3037785079242363e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7022
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.40625,
      "epoch": 0.3393409354464631,
      "grad_norm": 0.0,
      "learning_rate": 3.3035369153459604e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7023
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.53125,
      "epoch": 0.3393892539621183,
      "grad_norm": 0.0,
      "learning_rate": 3.3032953227676846e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7024
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.96875,
      "epoch": 0.3394375724777735,
      "grad_norm": 0.0,
      "learning_rate": 3.3030537301894087e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7025
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.0,
      "epoch": 0.3394858909934287,
      "grad_norm": 0.0,
      "learning_rate": 3.3028121376111323e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 7026
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.78125,
      "epoch": 0.33953420950908386,
      "grad_norm": 0.0,
      "learning_rate": 3.302570545032857e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7027
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.34375,
      "epoch": 0.3395825280247391,
      "grad_norm": 0.0,
      "learning_rate": 3.3023289524545806e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7028
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 96.1875,
      "epoch": 0.33963084654039427,
      "grad_norm": 13.530960382708281,
      "learning_rate": 3.302087359876304e-07,
      "loss": -0.0,
      "reward": 1.46875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7029
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.125,
      "epoch": 0.3396791650560495,
      "grad_norm": 0.0,
      "learning_rate": 3.3018457672980283e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 7030
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.78125,
      "epoch": 0.3397274835717047,
      "grad_norm": 0.0,
      "learning_rate": 3.3016041747197525e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7031
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.9375,
      "epoch": 0.33977580208735986,
      "grad_norm": 0.0,
      "learning_rate": 3.3013625821414766e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7032
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.75,
      "epoch": 0.3398241206030151,
      "grad_norm": 0.0,
      "learning_rate": 3.3011209895632e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7033
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.71875,
      "epoch": 0.33987243911867027,
      "grad_norm": 0.0,
      "learning_rate": 3.300879396984925e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 7034
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.5,
      "epoch": 0.3399207576343255,
      "grad_norm": 0.0,
      "learning_rate": 3.3006378044066485e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 7035
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.59375,
      "epoch": 0.3399690761499807,
      "grad_norm": 0.0,
      "learning_rate": 3.3003962118283726e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7036
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.375,
      "epoch": 0.34001739466563585,
      "grad_norm": 0.0,
      "learning_rate": 3.300154619250097e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7037
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.15625,
      "epoch": 0.3400657131812911,
      "grad_norm": 0.0,
      "learning_rate": 3.2999130266718203e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 7038
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.40625,
      "epoch": 0.34011403169694626,
      "grad_norm": 0.0,
      "learning_rate": 3.2996714340935445e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7039
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.25,
      "epoch": 0.3401623502126015,
      "grad_norm": 0.0,
      "learning_rate": 3.299429841515268e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7040
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.46875,
      "epoch": 0.34021066872825667,
      "grad_norm": 0.0,
      "learning_rate": 3.299188248936993e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7041
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.625,
      "epoch": 0.34025898724391185,
      "grad_norm": 0.0,
      "learning_rate": 3.2989466563587164e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 7042
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.46875,
      "epoch": 0.3403073057595671,
      "grad_norm": 0.0,
      "learning_rate": 3.2987050637804405e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7043
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.65625,
      "epoch": 0.34035562427522226,
      "grad_norm": 0.0,
      "learning_rate": 3.2984634712021646e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7044
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.125,
      "epoch": 0.3404039427908775,
      "grad_norm": 0.0,
      "learning_rate": 3.298221878623889e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7045
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 93.21875,
      "epoch": 0.34045226130653267,
      "grad_norm": 12.308558808303507,
      "learning_rate": 3.2979802860456124e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7046
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.75,
      "epoch": 0.34050057982218784,
      "grad_norm": 0.0,
      "learning_rate": 3.297738693467337e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7047
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.8125,
      "epoch": 0.3405488983378431,
      "grad_norm": 44.509205992262146,
      "learning_rate": 3.2974971008890607e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.30509042739868164,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7048
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.03125,
      "epoch": 0.34059721685349825,
      "grad_norm": 0.0,
      "learning_rate": 3.297255508310784e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7049
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.25,
      "epoch": 0.3406455353691535,
      "grad_norm": 0.0,
      "learning_rate": 3.2970139157325084e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7050
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 106.21875,
      "epoch": 0.34069385388480866,
      "grad_norm": 0.0,
      "learning_rate": 3.2967723231542325e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 7051
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.46875,
      "epoch": 0.34074217240046384,
      "grad_norm": 0.0,
      "learning_rate": 3.2965307305759567e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7052
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 89.46875,
      "epoch": 0.34079049091611907,
      "grad_norm": 6.989511632595147,
      "learning_rate": 3.2962891379976803e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7053
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 95.96875,
      "epoch": 0.34083880943177425,
      "grad_norm": 3.475989310452592,
      "learning_rate": 3.296047545419405e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7054
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.09375,
      "epoch": 0.3408871279474295,
      "grad_norm": 0.0,
      "learning_rate": 3.2958059528411285e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7055
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.09375,
      "epoch": 0.34093544646308466,
      "grad_norm": 0.0,
      "learning_rate": 3.2955643602628527e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7056
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.3125,
      "epoch": 0.34098376497873983,
      "grad_norm": 0.0,
      "learning_rate": 3.295322767684577e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7057
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.5625,
      "epoch": 0.34103208349439507,
      "grad_norm": 0.0,
      "learning_rate": 3.2950811751063004e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7058
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.9375,
      "epoch": 0.34108040201005024,
      "grad_norm": 0.0,
      "learning_rate": 3.2948395825280246e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7059
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.5625,
      "epoch": 0.3411287205257055,
      "grad_norm": 0.0,
      "learning_rate": 3.294597989949748e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7060
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.4375,
      "epoch": 0.34117703904136065,
      "grad_norm": 0.0,
      "learning_rate": 3.294356397371473e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7061
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.375,
      "epoch": 0.34122535755701583,
      "grad_norm": 0.0,
      "learning_rate": 3.2941148047931964e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7062
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.03125,
      "epoch": 0.34127367607267106,
      "grad_norm": 0.0,
      "learning_rate": 3.2938732122149206e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7063
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 85.59375,
      "epoch": 0.34132199458832624,
      "grad_norm": 18.422333543441678,
      "learning_rate": 3.2936316196366447e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7064
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.65625,
      "epoch": 0.34137031310398147,
      "grad_norm": 0.0,
      "learning_rate": 3.293390027058369e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7065
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 88.65625,
      "epoch": 0.34141863161963665,
      "grad_norm": 8.504123607953193,
      "learning_rate": 3.2931484344800925e-07,
      "loss": 0.0,
      "reward": 1.5625,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7066
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.53125,
      "epoch": 0.3414669501352918,
      "grad_norm": 0.0,
      "learning_rate": 3.292906841901817e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7067
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.5,
      "epoch": 0.34151526865094706,
      "grad_norm": 0.0,
      "learning_rate": 3.2926652493235407e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7068
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.15625,
      "epoch": 0.34156358716660223,
      "grad_norm": 0.0,
      "learning_rate": 3.2924236567452643e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7069
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.125,
      "epoch": 0.34161190568225747,
      "grad_norm": 0.0,
      "learning_rate": 3.2921820641669885e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7070
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 83.8125,
      "epoch": 0.34166022419791264,
      "grad_norm": 9.398922543441154,
      "learning_rate": 3.2919404715887126e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7071
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.65625,
      "epoch": 0.3417085427135678,
      "grad_norm": 0.0,
      "learning_rate": 3.291698879010437e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7072
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.90625,
      "epoch": 0.34175686122922305,
      "grad_norm": 0.0,
      "learning_rate": 3.2914572864321604e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7073
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.09375,
      "epoch": 0.34180517974487823,
      "grad_norm": 0.0,
      "learning_rate": 3.291215693853885e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7074
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.375,
      "epoch": 0.34185349826053346,
      "grad_norm": 0.0,
      "learning_rate": 3.2909741012756086e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7075
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 89.75,
      "epoch": 0.34190181677618864,
      "grad_norm": 5.690799163483488,
      "learning_rate": 3.290732508697333e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7076
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.25,
      "epoch": 0.3419501352918438,
      "grad_norm": 0.0,
      "learning_rate": 3.290490916119057e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7077
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.875,
      "epoch": 0.34199845380749905,
      "grad_norm": 0.0,
      "learning_rate": 3.290249323540781e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7078
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.84375,
      "epoch": 0.3420467723231542,
      "grad_norm": 0.0,
      "learning_rate": 3.2900077309625046e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7079
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.25,
      "epoch": 0.34209509083880946,
      "grad_norm": 0.0,
      "learning_rate": 3.289766138384228e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7080
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 93.78125,
      "epoch": 0.34214340935446463,
      "grad_norm": 9.945341900362935,
      "learning_rate": 3.289524545805953e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7081
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.34375,
      "epoch": 0.3421917278701198,
      "grad_norm": 0.0,
      "learning_rate": 3.2892829532276765e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7082
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.65625,
      "epoch": 0.34224004638577504,
      "grad_norm": 0.0,
      "learning_rate": 3.2890413606494007e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7083
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.375,
      "epoch": 0.3422883649014302,
      "grad_norm": 0.0,
      "learning_rate": 3.288799768071125e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7084
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.875,
      "epoch": 0.34233668341708545,
      "grad_norm": 0.0,
      "learning_rate": 3.288558175492849e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7085
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.34375,
      "epoch": 0.34238500193274063,
      "grad_norm": 0.0,
      "learning_rate": 3.2883165829145725e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7086
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.25,
      "epoch": 0.3424333204483958,
      "grad_norm": 0.0,
      "learning_rate": 3.288074990336297e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7087
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.78125,
      "epoch": 0.34248163896405104,
      "grad_norm": 0.0,
      "learning_rate": 3.287833397758021e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7088
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.75,
      "epoch": 0.3425299574797062,
      "grad_norm": 0.0,
      "learning_rate": 3.2875918051797444e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7089
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.8125,
      "epoch": 0.34257827599536145,
      "grad_norm": 0.0,
      "learning_rate": 3.2873502126014686e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7090
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.71875,
      "epoch": 0.3426265945110166,
      "grad_norm": 0.0,
      "learning_rate": 3.2871086200231927e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7091
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 93.8125,
      "epoch": 0.3426749130266718,
      "grad_norm": 2.105112904983074,
      "learning_rate": 3.286867027444917e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.3319803476333618,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 0.96875,
      "step": 7092
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.84375,
      "epoch": 0.34272323154232703,
      "grad_norm": 0.0,
      "learning_rate": 3.2866254348666404e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 0.96875,
      "step": 7093
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.25,
      "epoch": 0.3427715500579822,
      "grad_norm": 0.0,
      "learning_rate": 3.286383842288365e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7094
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.25,
      "epoch": 0.34281986857363744,
      "grad_norm": 4.862312474989772,
      "learning_rate": 3.2861422497100887e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.34374770522117615,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 0.96875,
      "step": 7095
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.15625,
      "epoch": 0.3428681870892926,
      "grad_norm": 0.0,
      "learning_rate": 3.285900657131813e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7096
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.34375,
      "epoch": 0.3429165056049478,
      "grad_norm": 0.0,
      "learning_rate": 3.285659064553537e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7097
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.8125,
      "epoch": 0.34296482412060303,
      "grad_norm": 0.0,
      "learning_rate": 3.285417471975261e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7098
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.96875,
      "epoch": 0.3430131426362582,
      "grad_norm": 0.0,
      "learning_rate": 3.2851758793969847e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7099
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.6875,
      "epoch": 0.34306146115191344,
      "grad_norm": 0.0,
      "learning_rate": 3.2849342868187083e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7100
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.125,
      "epoch": 0.3431097796675686,
      "grad_norm": 0.0,
      "learning_rate": 3.284692694240433e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7101
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.0,
      "epoch": 0.3431580981832238,
      "grad_norm": 0.0,
      "learning_rate": 3.2844511016621566e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7102
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.65625,
      "epoch": 0.343206416698879,
      "grad_norm": 0.0,
      "learning_rate": 3.284209509083881e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7103
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.1875,
      "epoch": 0.3432547352145342,
      "grad_norm": 0.0,
      "learning_rate": 3.283967916505605e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7104
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.0,
      "epoch": 0.34330305373018943,
      "grad_norm": 0.0,
      "learning_rate": 3.283726323927329e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7105
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.4375,
      "epoch": 0.3433513722458446,
      "grad_norm": 0.0,
      "learning_rate": 3.2834847313490526e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7106
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.3125,
      "epoch": 0.3433996907614998,
      "grad_norm": 0.0,
      "learning_rate": 3.2832431387707773e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7107
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.09375,
      "epoch": 0.343448009277155,
      "grad_norm": 0.0,
      "learning_rate": 3.283001546192501e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7108
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.15625,
      "epoch": 0.3434963277928102,
      "grad_norm": 0.0,
      "learning_rate": 3.2827599536142245e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7109
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.03125,
      "epoch": 0.34354464630846543,
      "grad_norm": 0.0,
      "learning_rate": 3.2825183610359486e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7110
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.9375,
      "epoch": 0.3435929648241206,
      "grad_norm": 0.0,
      "learning_rate": 3.282276768457673e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7111
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.9375,
      "epoch": 0.3436412833397758,
      "grad_norm": 0.0,
      "learning_rate": 3.282035175879397e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7112
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 83.6875,
      "epoch": 0.343689601855431,
      "grad_norm": 0.0,
      "learning_rate": 3.2817935833011205e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 0.96875,
      "step": 7113
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.5625,
      "epoch": 0.3437379203710862,
      "grad_norm": 0.0,
      "learning_rate": 3.281551990722845e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7114
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.65625,
      "epoch": 0.3437862388867414,
      "grad_norm": 0.0,
      "learning_rate": 3.281310398144569e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7115
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.125,
      "epoch": 0.3438345574023966,
      "grad_norm": 0.0,
      "learning_rate": 3.281068805566293e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7116
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 75.78125,
      "epoch": 0.3438828759180518,
      "grad_norm": 42.3221185634119,
      "learning_rate": 3.280827212988017e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.31983357667922974,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7117
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.0,
      "epoch": 0.343931194433707,
      "grad_norm": 0.0,
      "learning_rate": 3.280585620409741e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7118
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.25,
      "epoch": 0.3439795129493622,
      "grad_norm": 0.0,
      "learning_rate": 3.280344027831465e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7119
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.4375,
      "epoch": 0.3440278314650174,
      "grad_norm": 0.0,
      "learning_rate": 3.2801024352531884e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7120
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.0,
      "epoch": 0.3440761499806726,
      "grad_norm": 30.55330149451753,
      "learning_rate": 3.279860842674913e-07,
      "loss": -0.0,
      "reward": 2.140625,
      "reward_std": 0.2804819941520691,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7121
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.375,
      "epoch": 0.3441244684963278,
      "grad_norm": 0.0,
      "learning_rate": 3.2796192500966367e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7122
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.90625,
      "epoch": 0.344172787011983,
      "grad_norm": 0.0,
      "learning_rate": 3.279377657518361e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7123
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 82.625,
      "epoch": 0.3442211055276382,
      "grad_norm": 0.0,
      "learning_rate": 3.279136064940085e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 7124
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.0625,
      "epoch": 0.3442694240432934,
      "grad_norm": 0.0,
      "learning_rate": 3.278894472361809e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7125
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.03125,
      "epoch": 0.3443177425589486,
      "grad_norm": 0.0,
      "learning_rate": 3.2786528797835327e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7126
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.90625,
      "epoch": 0.34436606107460377,
      "grad_norm": 0.0,
      "learning_rate": 3.2784112872052574e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7127
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.90625,
      "epoch": 0.344414379590259,
      "grad_norm": 0.0,
      "learning_rate": 3.278169694626981e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 7128
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.3125,
      "epoch": 0.3444626981059142,
      "grad_norm": 0.0,
      "learning_rate": 3.277928102048705e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7129
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.21875,
      "epoch": 0.3445110166215694,
      "grad_norm": 0.0,
      "learning_rate": 3.2776865094704287e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7130
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.21875,
      "epoch": 0.3445593351372246,
      "grad_norm": 0.0,
      "learning_rate": 3.277444916892153e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7131
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.6875,
      "epoch": 0.34460765365287976,
      "grad_norm": 0.0,
      "learning_rate": 3.277203324313877e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7132
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.4375,
      "epoch": 0.344655972168535,
      "grad_norm": 0.0,
      "learning_rate": 3.2769617317356006e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7133
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.40625,
      "epoch": 0.3447042906841902,
      "grad_norm": 0.0,
      "learning_rate": 3.276720139157325e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7134
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.28125,
      "epoch": 0.3447526091998454,
      "grad_norm": 0.0,
      "learning_rate": 3.276478546579049e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7135
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.84375,
      "epoch": 0.3448009277155006,
      "grad_norm": 0.0,
      "learning_rate": 3.276236954000773e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7136
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.28125,
      "epoch": 0.34484924623115576,
      "grad_norm": 0.0,
      "learning_rate": 3.275995361422497e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7137
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 79.03125,
      "epoch": 0.344897564746811,
      "grad_norm": 12.129110891847967,
      "learning_rate": 3.2757537688442213e-07,
      "loss": -0.0,
      "reward": 1.328125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 7138
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.875,
      "epoch": 0.34494588326246617,
      "grad_norm": 0.0,
      "learning_rate": 3.275512176265945e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7139
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.84375,
      "epoch": 0.3449942017781214,
      "grad_norm": 0.0,
      "learning_rate": 3.2752705836876685e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7140
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.65625,
      "epoch": 0.3450425202937766,
      "grad_norm": 0.0,
      "learning_rate": 3.275028991109393e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7141
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.59375,
      "epoch": 0.34509083880943175,
      "grad_norm": 0.0,
      "learning_rate": 3.274787398531117e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7142
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.375,
      "epoch": 0.345139157325087,
      "grad_norm": 0.0,
      "learning_rate": 3.274545805952841e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7143
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.15625,
      "epoch": 0.34518747584074216,
      "grad_norm": 0.0,
      "learning_rate": 3.274304213374565e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7144
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.5625,
      "epoch": 0.3452357943563974,
      "grad_norm": 0.0,
      "learning_rate": 3.274062620796289e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7145
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.9375,
      "epoch": 0.3452841128720526,
      "grad_norm": 0.0,
      "learning_rate": 3.273821028218013e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7146
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.9375,
      "epoch": 0.34533243138770775,
      "grad_norm": 0.0,
      "learning_rate": 3.2735794356397374e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7147
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.3125,
      "epoch": 0.345380749903363,
      "grad_norm": 0.0,
      "learning_rate": 3.273337843061461e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7148
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.0,
      "epoch": 0.34542906841901816,
      "grad_norm": 0.0,
      "learning_rate": 3.273096250483185e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7149
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.0,
      "epoch": 0.3454773869346734,
      "grad_norm": 0.0,
      "learning_rate": 3.272854657904909e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7150
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.1875,
      "epoch": 0.34552570545032857,
      "grad_norm": 0.0,
      "learning_rate": 3.272613065326633e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7151
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.65625,
      "epoch": 0.34557402396598375,
      "grad_norm": 0.0,
      "learning_rate": 3.272371472748357e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7152
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.0,
      "epoch": 0.345622342481639,
      "grad_norm": 0.0,
      "learning_rate": 3.2721298801700807e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7153
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.84375,
      "epoch": 0.34567066099729415,
      "grad_norm": 0.0,
      "learning_rate": 3.2718882875918053e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7154
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.5,
      "epoch": 0.3457189795129494,
      "grad_norm": 0.0,
      "learning_rate": 3.271646695013529e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7155
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.46875,
      "epoch": 0.34576729802860456,
      "grad_norm": 0.0,
      "learning_rate": 3.271405102435253e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7156
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.34375,
      "epoch": 0.34581561654425974,
      "grad_norm": 0.0,
      "learning_rate": 3.271163509856977e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7157
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.0,
      "epoch": 0.345863935059915,
      "grad_norm": 0.0,
      "learning_rate": 3.2709219172787014e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7158
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.5,
      "epoch": 0.34591225357557015,
      "grad_norm": 0.0,
      "learning_rate": 3.270680324700425e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7159
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.4375,
      "epoch": 0.3459605720912254,
      "grad_norm": 0.0,
      "learning_rate": 3.2704387321221486e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7160
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.5625,
      "epoch": 0.34600889060688056,
      "grad_norm": 0.0,
      "learning_rate": 3.270197139543873e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7161
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.8125,
      "epoch": 0.34605720912253574,
      "grad_norm": 0.0,
      "learning_rate": 3.269955546965597e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7162
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.0,
      "epoch": 0.34610552763819097,
      "grad_norm": 0.0,
      "learning_rate": 3.269713954387321e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7163
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.9375,
      "epoch": 0.34615384615384615,
      "grad_norm": 0.0,
      "learning_rate": 3.269472361809045e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7164
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.40625,
      "epoch": 0.3462021646695014,
      "grad_norm": 0.0,
      "learning_rate": 3.269230769230769e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7165
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.4375,
      "epoch": 0.34625048318515655,
      "grad_norm": 0.0,
      "learning_rate": 3.268989176652493e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7166
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.84375,
      "epoch": 0.34629880170081173,
      "grad_norm": 0.0,
      "learning_rate": 3.2687475840742175e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7167
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 70.75,
      "epoch": 0.34634712021646696,
      "grad_norm": 0.0,
      "learning_rate": 3.268505991495941e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 7168
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.8125,
      "epoch": 0.34639543873212214,
      "grad_norm": 0.0,
      "learning_rate": 3.268264398917665e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7169
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.90625,
      "epoch": 0.3464437572477774,
      "grad_norm": 0.0,
      "learning_rate": 3.268022806339389e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7170
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.84375,
      "epoch": 0.34649207576343255,
      "grad_norm": 0.0,
      "learning_rate": 3.267781213761113e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7171
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.5,
      "epoch": 0.3465403942790877,
      "grad_norm": 0.0,
      "learning_rate": 3.267539621182837e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7172
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.3125,
      "epoch": 0.34658871279474296,
      "grad_norm": 0.0,
      "learning_rate": 3.267298028604561e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7173
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.09375,
      "epoch": 0.34663703131039814,
      "grad_norm": 0.0,
      "learning_rate": 3.2670564360262854e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7174
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.84375,
      "epoch": 0.34668534982605337,
      "grad_norm": 0.0,
      "learning_rate": 3.266814843448009e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7175
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.40625,
      "epoch": 0.34673366834170855,
      "grad_norm": 0.0,
      "learning_rate": 3.266573250869733e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7176
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.375,
      "epoch": 0.3467819868573637,
      "grad_norm": 0.0,
      "learning_rate": 3.2663316582914573e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7177
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.96875,
      "epoch": 0.34683030537301895,
      "grad_norm": 0.0,
      "learning_rate": 3.2660900657131814e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7178
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.78125,
      "epoch": 0.34687862388867413,
      "grad_norm": 0.0,
      "learning_rate": 3.265848473134905e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7179
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.75,
      "epoch": 0.34692694240432936,
      "grad_norm": 0.0,
      "learning_rate": 3.265606880556629e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7180
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.71875,
      "epoch": 0.34697526091998454,
      "grad_norm": 0.0,
      "learning_rate": 3.2653652879783533e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7181
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.03125,
      "epoch": 0.3470235794356397,
      "grad_norm": 0.0,
      "learning_rate": 3.265123695400077e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7182
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.65625,
      "epoch": 0.34707189795129495,
      "grad_norm": 0.0,
      "learning_rate": 3.264882102821801e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7183
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.40625,
      "epoch": 0.3471202164669501,
      "grad_norm": 0.0,
      "learning_rate": 3.264640510243525e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7184
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.4375,
      "epoch": 0.34716853498260536,
      "grad_norm": 0.0,
      "learning_rate": 3.2643989176652493e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7185
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.40625,
      "epoch": 0.34721685349826054,
      "grad_norm": 0.0,
      "learning_rate": 3.264157325086973e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7186
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.46875,
      "epoch": 0.3472651720139157,
      "grad_norm": 0.0,
      "learning_rate": 3.2639157325086976e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7187
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.21875,
      "epoch": 0.34731349052957095,
      "grad_norm": 0.0,
      "learning_rate": 3.263674139930421e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7188
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.28125,
      "epoch": 0.3473618090452261,
      "grad_norm": 0.0,
      "learning_rate": 3.2634325473521453e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7189
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.71875,
      "epoch": 0.34741012756088135,
      "grad_norm": 0.0,
      "learning_rate": 3.263190954773869e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7190
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.0,
      "epoch": 0.34745844607653653,
      "grad_norm": 0.0,
      "learning_rate": 3.2629493621955936e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7191
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.53125,
      "epoch": 0.3475067645921917,
      "grad_norm": 0.0,
      "learning_rate": 3.262707769617317e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7192
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.90625,
      "epoch": 0.34755508310784694,
      "grad_norm": 0.0,
      "learning_rate": 3.262466177039041e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7193
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 72.25,
      "epoch": 0.3476034016235021,
      "grad_norm": 19.94779119914433,
      "learning_rate": 3.2622245844607655e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.20633062720298767,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7194
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.15625,
      "epoch": 0.34765172013915735,
      "grad_norm": 0.0,
      "learning_rate": 3.261982991882489e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 7195
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.625,
      "epoch": 0.3477000386548125,
      "grad_norm": 0.0,
      "learning_rate": 3.261741399304213e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7196
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.5625,
      "epoch": 0.3477483571704677,
      "grad_norm": 0.0,
      "learning_rate": 3.2614998067259374e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7197
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.6875,
      "epoch": 0.34779667568612294,
      "grad_norm": 0.0,
      "learning_rate": 3.2612582141476615e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7198
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.8125,
      "epoch": 0.3478449942017781,
      "grad_norm": 0.0,
      "learning_rate": 3.261016621569385e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7199
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.875,
      "epoch": 0.34789331271743335,
      "grad_norm": 0.0,
      "learning_rate": 3.260775028991109e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7200
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.96875,
      "epoch": 0.3479416312330885,
      "grad_norm": 0.0,
      "learning_rate": 3.2605334364128334e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7201
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.28125,
      "epoch": 0.3479899497487437,
      "grad_norm": 0.0,
      "learning_rate": 3.260291843834557e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7202
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 63.8125,
      "epoch": 0.34803826826439893,
      "grad_norm": 0.0,
      "learning_rate": 3.260050251256281e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 7203
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.25,
      "epoch": 0.3480865867800541,
      "grad_norm": 0.0,
      "learning_rate": 3.2598086586780053e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7204
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.84375,
      "epoch": 0.34813490529570934,
      "grad_norm": 0.0,
      "learning_rate": 3.2595670660997294e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7205
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.9375,
      "epoch": 0.3481832238113645,
      "grad_norm": 0.0,
      "learning_rate": 3.259325473521453e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7206
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 71.71875,
      "epoch": 0.3482315423270197,
      "grad_norm": 0.0,
      "learning_rate": 3.2590838809431777e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 7207
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.96875,
      "epoch": 0.3482798608426749,
      "grad_norm": 0.0,
      "learning_rate": 3.2588422883649013e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7208
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.84375,
      "epoch": 0.3483281793583301,
      "grad_norm": 0.0,
      "learning_rate": 3.2586006957866254e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7209
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.875,
      "epoch": 0.34837649787398534,
      "grad_norm": 0.0,
      "learning_rate": 3.258359103208349e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7210
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.53125,
      "epoch": 0.3484248163896405,
      "grad_norm": 0.0,
      "learning_rate": 3.2581175106300737e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7211
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.875,
      "epoch": 0.3484731349052957,
      "grad_norm": 0.0,
      "learning_rate": 3.2578759180517973e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7212
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.03125,
      "epoch": 0.3485214534209509,
      "grad_norm": 0.0,
      "learning_rate": 3.257634325473521e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7213
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.78125,
      "epoch": 0.3485697719366061,
      "grad_norm": 0.0,
      "learning_rate": 3.2573927328952456e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7214
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.46875,
      "epoch": 0.34861809045226133,
      "grad_norm": 0.0,
      "learning_rate": 3.257151140316969e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7215
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.1875,
      "epoch": 0.3486664089679165,
      "grad_norm": 0.0,
      "learning_rate": 3.2569095477386933e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7216
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.375,
      "epoch": 0.3487147274835717,
      "grad_norm": 0.0,
      "learning_rate": 3.2566679551604175e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7217
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.4375,
      "epoch": 0.3487630459992269,
      "grad_norm": 0.0,
      "learning_rate": 3.2564263625821416e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7218
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.375,
      "epoch": 0.3488113645148821,
      "grad_norm": 0.0,
      "learning_rate": 3.256184770003865e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7219
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.4375,
      "epoch": 0.3488596830305373,
      "grad_norm": 0.0,
      "learning_rate": 3.2559431774255893e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7220
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.1875,
      "epoch": 0.3489080015461925,
      "grad_norm": 0.0,
      "learning_rate": 3.2557015848473135e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7221
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.15625,
      "epoch": 0.3489563200618477,
      "grad_norm": 0.0,
      "learning_rate": 3.255459992269037e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7222
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.53125,
      "epoch": 0.3490046385775029,
      "grad_norm": 0.0,
      "learning_rate": 3.255218399690761e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7223
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.4375,
      "epoch": 0.3490529570931581,
      "grad_norm": 0.0,
      "learning_rate": 3.2549768071124853e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7224
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.0,
      "epoch": 0.3491012756088133,
      "grad_norm": 0.0,
      "learning_rate": 3.2547352145342095e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 7225
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.3125,
      "epoch": 0.3491495941244685,
      "grad_norm": 0.0,
      "learning_rate": 3.254493621955933e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7226
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.0,
      "epoch": 0.3491979126401237,
      "grad_norm": 0.0,
      "learning_rate": 3.254252029377658e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7227
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.78125,
      "epoch": 0.3492462311557789,
      "grad_norm": 0.0,
      "learning_rate": 3.2540104367993814e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7228
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.15625,
      "epoch": 0.3492945496714341,
      "grad_norm": 0.0,
      "learning_rate": 3.2537688442211055e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 7229
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.1875,
      "epoch": 0.3493428681870893,
      "grad_norm": 0.0,
      "learning_rate": 3.253527251642829e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7230
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.09375,
      "epoch": 0.3493911867027445,
      "grad_norm": 0.0,
      "learning_rate": 3.253285659064554e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7231
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.0,
      "epoch": 0.34943950521839967,
      "grad_norm": 0.0,
      "learning_rate": 3.2530440664862774e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 7232
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.59375,
      "epoch": 0.3494878237340549,
      "grad_norm": 0.0,
      "learning_rate": 3.252802473908001e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7233
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.59375,
      "epoch": 0.3495361422497101,
      "grad_norm": 0.0,
      "learning_rate": 3.2525608813297256e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7234
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.78125,
      "epoch": 0.3495844607653653,
      "grad_norm": 0.0,
      "learning_rate": 3.252319288751449e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7235
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.0625,
      "epoch": 0.3496327792810205,
      "grad_norm": 0.0,
      "learning_rate": 3.2520776961731734e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 7236
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.28125,
      "epoch": 0.34968109779667567,
      "grad_norm": 0.0,
      "learning_rate": 3.2518361035948975e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7237
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.125,
      "epoch": 0.3497294163123309,
      "grad_norm": 0.0,
      "learning_rate": 3.2515945110166217e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7238
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 67.03125,
      "epoch": 0.3497777348279861,
      "grad_norm": 0.0,
      "learning_rate": 3.2513529184383453e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7239
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.1875,
      "epoch": 0.3498260533436413,
      "grad_norm": 0.0,
      "learning_rate": 3.2511113258600694e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7240
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.1875,
      "epoch": 0.3498743718592965,
      "grad_norm": 0.0,
      "learning_rate": 3.2508697332817935e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7241
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.15625,
      "epoch": 0.34992269037495166,
      "grad_norm": 0.0,
      "learning_rate": 3.2506281407035177e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 7242
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.9375,
      "epoch": 0.3499710088906069,
      "grad_norm": 0.0,
      "learning_rate": 3.2503865481252413e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7243
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.8125,
      "epoch": 0.35001932740626207,
      "grad_norm": 0.0,
      "learning_rate": 3.2501449555469654e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7244
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.84375,
      "epoch": 0.3500676459219173,
      "grad_norm": 0.0,
      "learning_rate": 3.2499033629686896e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7245
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.78125,
      "epoch": 0.3501159644375725,
      "grad_norm": 0.0,
      "learning_rate": 3.249661770390413e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7246
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.5,
      "epoch": 0.35016428295322766,
      "grad_norm": 0.0,
      "learning_rate": 3.249420177812138e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7247
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.3125,
      "epoch": 0.3502126014688829,
      "grad_norm": 0.0,
      "learning_rate": 3.2491785852338614e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 7248
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.09375,
      "epoch": 0.35026091998453807,
      "grad_norm": 0.0,
      "learning_rate": 3.2489369926555856e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7249
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.21875,
      "epoch": 0.3503092385001933,
      "grad_norm": 0.0,
      "learning_rate": 3.248695400077309e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7250
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.15625,
      "epoch": 0.3503575570158485,
      "grad_norm": 0.0,
      "learning_rate": 3.248453807499034e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7251
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.78125,
      "epoch": 0.35040587553150365,
      "grad_norm": 0.0,
      "learning_rate": 3.2482122149207575e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7252
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.59375,
      "epoch": 0.3504541940471589,
      "grad_norm": 0.0,
      "learning_rate": 3.247970622342481e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7253
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.0,
      "epoch": 0.35050251256281406,
      "grad_norm": 0.0,
      "learning_rate": 3.2477290297642057e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7254
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.59375,
      "epoch": 0.3505508310784693,
      "grad_norm": 0.0,
      "learning_rate": 3.2474874371859293e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7255
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.15625,
      "epoch": 0.35059914959412447,
      "grad_norm": 0.0,
      "learning_rate": 3.2472458446076535e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7256
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.0625,
      "epoch": 0.35064746810977965,
      "grad_norm": 0.0,
      "learning_rate": 3.2470042520293776e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7257
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.75,
      "epoch": 0.3506957866254349,
      "grad_norm": 0.0,
      "learning_rate": 3.246762659451102e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7258
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.0625,
      "epoch": 0.35074410514109006,
      "grad_norm": 0.0,
      "learning_rate": 3.2465210668728254e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7259
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.59375,
      "epoch": 0.3507924236567453,
      "grad_norm": 0.0,
      "learning_rate": 3.2462794742945495e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7260
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.34375,
      "epoch": 0.35084074217240047,
      "grad_norm": 0.0,
      "learning_rate": 3.2460378817162736e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7261
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.03125,
      "epoch": 0.35088906068805564,
      "grad_norm": 0.0,
      "learning_rate": 3.245796289137998e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 7262
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.75,
      "epoch": 0.3509373792037109,
      "grad_norm": 0.0,
      "learning_rate": 3.2455546965597214e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7263
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.28125,
      "epoch": 0.35098569771936605,
      "grad_norm": 0.0,
      "learning_rate": 3.2453131039814455e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 7264
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.53125,
      "epoch": 0.3510340162350213,
      "grad_norm": 0.0,
      "learning_rate": 3.2450715114031696e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7265
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.875,
      "epoch": 0.35108233475067646,
      "grad_norm": 0.0,
      "learning_rate": 3.244829918824893e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7266
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.96875,
      "epoch": 0.35113065326633164,
      "grad_norm": 0.0,
      "learning_rate": 3.244588326246618e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7267
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.90625,
      "epoch": 0.35117897178198687,
      "grad_norm": 0.0,
      "learning_rate": 3.2443467336683415e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7268
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.90625,
      "epoch": 0.35122729029764205,
      "grad_norm": 0.0,
      "learning_rate": 3.2441051410900657e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 7269
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.96875,
      "epoch": 0.3512756088132973,
      "grad_norm": 0.0,
      "learning_rate": 3.243863548511789e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 7270
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.03125,
      "epoch": 0.35132392732895246,
      "grad_norm": 0.0,
      "learning_rate": 3.243621955933514e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7271
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.03125,
      "epoch": 0.35137224584460763,
      "grad_norm": 0.0,
      "learning_rate": 3.2433803633552375e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7272
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 73.6875,
      "epoch": 0.35142056436026287,
      "grad_norm": 55.48835453648831,
      "learning_rate": 3.2431387707769617e-07,
      "loss": -0.0,
      "reward": 1.484375,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7273
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 62.71875,
      "epoch": 0.35146888287591804,
      "grad_norm": 22.626181337399235,
      "learning_rate": 3.242897178198686e-07,
      "loss": -0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 7274
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.4375,
      "epoch": 0.3515172013915733,
      "grad_norm": 0.0,
      "learning_rate": 3.2426555856204094e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7275
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.5,
      "epoch": 0.35156551990722845,
      "grad_norm": 0.0,
      "learning_rate": 3.2424139930421335e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7276
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.71875,
      "epoch": 0.35161383842288363,
      "grad_norm": 0.0,
      "learning_rate": 3.2421724004638577e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7277
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.90625,
      "epoch": 0.35166215693853886,
      "grad_norm": 0.0,
      "learning_rate": 3.241930807885582e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7278
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.59375,
      "epoch": 0.35171047545419404,
      "grad_norm": 0.0,
      "learning_rate": 3.2416892153073054e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7279
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.28125,
      "epoch": 0.35175879396984927,
      "grad_norm": 0.0,
      "learning_rate": 3.2414476227290296e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7280
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.59375,
      "epoch": 0.35180711248550445,
      "grad_norm": 0.0,
      "learning_rate": 3.2412060301507537e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7281
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.34375,
      "epoch": 0.3518554310011596,
      "grad_norm": 0.0,
      "learning_rate": 3.240964437572478e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 7282
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.46875,
      "epoch": 0.35190374951681486,
      "grad_norm": 0.0,
      "learning_rate": 3.2407228449942014e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7283
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.5625,
      "epoch": 0.35195206803247003,
      "grad_norm": 0.0,
      "learning_rate": 3.240481252415926e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 7284
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.375,
      "epoch": 0.35200038654812527,
      "grad_norm": 0.0,
      "learning_rate": 3.2402396598376497e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 7285
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.03125,
      "epoch": 0.35204870506378044,
      "grad_norm": 0.0,
      "learning_rate": 3.2399980672593733e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7286
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.40625,
      "epoch": 0.3520970235794356,
      "grad_norm": 0.0,
      "learning_rate": 3.239756474681098e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7287
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.5625,
      "epoch": 0.35214534209509085,
      "grad_norm": 0.0,
      "learning_rate": 3.2395148821028216e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7288
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.84375,
      "epoch": 0.35219366061074603,
      "grad_norm": 0.0,
      "learning_rate": 3.2392732895245457e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7289
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.4375,
      "epoch": 0.35224197912640126,
      "grad_norm": 0.0,
      "learning_rate": 3.2390316969462693e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7290
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.0625,
      "epoch": 0.35229029764205644,
      "grad_norm": 0.0,
      "learning_rate": 3.238790104367994e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7291
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.34375,
      "epoch": 0.3523386161577116,
      "grad_norm": 0.0,
      "learning_rate": 3.2385485117897176e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7292
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.625,
      "epoch": 0.35238693467336685,
      "grad_norm": 0.0,
      "learning_rate": 3.238306919211442e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7293
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.34375,
      "epoch": 0.352435253189022,
      "grad_norm": 0.0,
      "learning_rate": 3.238065326633166e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 7294
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.8125,
      "epoch": 0.35248357170467726,
      "grad_norm": 0.0,
      "learning_rate": 3.2378237340548895e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7295
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.84375,
      "epoch": 0.35253189022033243,
      "grad_norm": 0.0,
      "learning_rate": 3.2375821414766136e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7296
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.15625,
      "epoch": 0.3525802087359876,
      "grad_norm": 7.163014124026222,
      "learning_rate": 3.237340548898338e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7297
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.21875,
      "epoch": 0.35262852725164284,
      "grad_norm": 0.0,
      "learning_rate": 3.237098956320062e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7298
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.25,
      "epoch": 0.352676845767298,
      "grad_norm": 0.0,
      "learning_rate": 3.2368573637417855e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7299
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.25,
      "epoch": 0.35272516428295325,
      "grad_norm": 0.0,
      "learning_rate": 3.2366157711635096e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7300
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.65625,
      "epoch": 0.35277348279860843,
      "grad_norm": 0.0,
      "learning_rate": 3.236374178585234e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7301
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.84375,
      "epoch": 0.3528218013142636,
      "grad_norm": 0.0,
      "learning_rate": 3.236132586006958e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 7302
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.28125,
      "epoch": 0.35287011982991884,
      "grad_norm": 0.0,
      "learning_rate": 3.2358909934286815e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7303
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.46875,
      "epoch": 0.352918438345574,
      "grad_norm": 0.0,
      "learning_rate": 3.235649400850406e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7304
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.34375,
      "epoch": 0.35296675686122925,
      "grad_norm": 0.0,
      "learning_rate": 3.23540780827213e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7305
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 58.46875,
      "epoch": 0.3530150753768844,
      "grad_norm": 44.01608693280639,
      "learning_rate": 3.2351662156938534e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.2529362142086029,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7306
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.03125,
      "epoch": 0.3530633938925396,
      "grad_norm": 0.0,
      "learning_rate": 3.234924623115578e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7307
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.15625,
      "epoch": 0.35311171240819483,
      "grad_norm": 0.0,
      "learning_rate": 3.2346830305373017e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7308
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.96875,
      "epoch": 0.35316003092385,
      "grad_norm": 0.0,
      "learning_rate": 3.234441437959026e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7309
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.84375,
      "epoch": 0.35320834943950524,
      "grad_norm": 0.0,
      "learning_rate": 3.2341998453807494e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7310
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.8125,
      "epoch": 0.3532566679551604,
      "grad_norm": 0.0,
      "learning_rate": 3.233958252802474e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7311
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.0625,
      "epoch": 0.3533049864708156,
      "grad_norm": 0.0,
      "learning_rate": 3.2337166602241977e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7312
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.1875,
      "epoch": 0.35335330498647083,
      "grad_norm": 0.0,
      "learning_rate": 3.233475067645922e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7313
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.125,
      "epoch": 0.353401623502126,
      "grad_norm": 29.680498817366395,
      "learning_rate": 3.233233475067646e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.31663835048675537,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7314
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 61.4375,
      "epoch": 0.35344994201778124,
      "grad_norm": 13.113958141618555,
      "learning_rate": 3.2329918824893696e-07,
      "loss": -0.0,
      "reward": 2.3125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7315
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.6875,
      "epoch": 0.3534982605334364,
      "grad_norm": 0.0,
      "learning_rate": 3.2327502899110937e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7316
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.78125,
      "epoch": 0.3535465790490916,
      "grad_norm": 0.0,
      "learning_rate": 3.232508697332818e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7317
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.34375,
      "epoch": 0.3535948975647468,
      "grad_norm": 0.0,
      "learning_rate": 3.232267104754542e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7318
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.4375,
      "epoch": 0.353643216080402,
      "grad_norm": 0.0,
      "learning_rate": 3.2320255121762656e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7319
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.90625,
      "epoch": 0.35369153459605723,
      "grad_norm": 0.0,
      "learning_rate": 3.2317839195979897e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7320
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.21875,
      "epoch": 0.3537398531117124,
      "grad_norm": 0.0,
      "learning_rate": 3.231542327019714e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7321
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.375,
      "epoch": 0.3537881716273676,
      "grad_norm": 0.0,
      "learning_rate": 3.231300734441438e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 7322
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.21875,
      "epoch": 0.3538364901430228,
      "grad_norm": 0.0,
      "learning_rate": 3.2310591418631616e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7323
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.9375,
      "epoch": 0.353884808658678,
      "grad_norm": 0.0,
      "learning_rate": 3.2308175492848863e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7324
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.4375,
      "epoch": 0.35393312717433323,
      "grad_norm": 0.0,
      "learning_rate": 3.23057595670661e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7325
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.71875,
      "epoch": 0.3539814456899884,
      "grad_norm": 0.0,
      "learning_rate": 3.2303343641283335e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7326
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.8125,
      "epoch": 0.3540297642056436,
      "grad_norm": 0.0,
      "learning_rate": 3.230092771550058e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7327
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.8125,
      "epoch": 0.3540780827212988,
      "grad_norm": 12.965795721318083,
      "learning_rate": 3.229851178971782e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.18174266815185547,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7328
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.75,
      "epoch": 0.354126401236954,
      "grad_norm": 0.0,
      "learning_rate": 3.229609586393506e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7329
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.21875,
      "epoch": 0.3541747197526092,
      "grad_norm": 0.0,
      "learning_rate": 3.2293679938152295e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7330
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.75,
      "epoch": 0.3542230382682644,
      "grad_norm": 0.0,
      "learning_rate": 3.229126401236954e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7331
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.65625,
      "epoch": 0.3542713567839196,
      "grad_norm": 0.0,
      "learning_rate": 3.228884808658678e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7332
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.34375,
      "epoch": 0.3543196752995748,
      "grad_norm": 0.0,
      "learning_rate": 3.228643216080402e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7333
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.59375,
      "epoch": 0.35436799381523,
      "grad_norm": 0.0,
      "learning_rate": 3.228401623502126e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7334
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.0,
      "epoch": 0.3544163123308852,
      "grad_norm": 0.0,
      "learning_rate": 3.22816003092385e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7335
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.25,
      "epoch": 0.3544646308465404,
      "grad_norm": 0.0,
      "learning_rate": 3.227918438345574e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7336
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.4375,
      "epoch": 0.3545129493621956,
      "grad_norm": 0.0,
      "learning_rate": 3.227676845767298e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7337
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.65625,
      "epoch": 0.3545612678778508,
      "grad_norm": 0.0,
      "learning_rate": 3.227435253189022e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7338
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.8125,
      "epoch": 0.354609586393506,
      "grad_norm": 0.0,
      "learning_rate": 3.2271936606107457e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7339
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.71875,
      "epoch": 0.3546579049091612,
      "grad_norm": 0.0,
      "learning_rate": 3.22695206803247e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7340
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.46875,
      "epoch": 0.3547062234248164,
      "grad_norm": 30.796774593239693,
      "learning_rate": 3.226710475454194e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7341
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.84375,
      "epoch": 0.35475454194047157,
      "grad_norm": 0.0,
      "learning_rate": 3.226468882875918e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7342
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.875,
      "epoch": 0.3548028604561268,
      "grad_norm": 0.0,
      "learning_rate": 3.2262272902976417e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7343
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.875,
      "epoch": 0.354851178971782,
      "grad_norm": 0.0,
      "learning_rate": 3.2259856977193663e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7344
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.375,
      "epoch": 0.3548994974874372,
      "grad_norm": 0.0,
      "learning_rate": 3.22574410514109e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7345
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.3125,
      "epoch": 0.3549478160030924,
      "grad_norm": 0.0,
      "learning_rate": 3.2255025125628136e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7346
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.5625,
      "epoch": 0.35499613451874756,
      "grad_norm": 0.0,
      "learning_rate": 3.225260919984538e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7347
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.5,
      "epoch": 0.3550444530344028,
      "grad_norm": 111.37345927620636,
      "learning_rate": 3.225019327406262e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.2961486577987671,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7348
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.15625,
      "epoch": 0.355092771550058,
      "grad_norm": 0.0,
      "learning_rate": 3.224777734827986e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7349
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.84375,
      "epoch": 0.3551410900657132,
      "grad_norm": 0.0,
      "learning_rate": 3.2245361422497096e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7350
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.625,
      "epoch": 0.3551894085813684,
      "grad_norm": 0.0,
      "learning_rate": 3.224294549671434e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7351
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.71875,
      "epoch": 0.35523772709702356,
      "grad_norm": 0.0,
      "learning_rate": 3.224052957093158e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7352
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.375,
      "epoch": 0.3552860456126788,
      "grad_norm": 0.0,
      "learning_rate": 3.223811364514882e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7353
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.53125,
      "epoch": 0.35533436412833397,
      "grad_norm": 0.0,
      "learning_rate": 3.223569771936606e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7354
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.6875,
      "epoch": 0.3553826826439892,
      "grad_norm": 0.0,
      "learning_rate": 3.22332817935833e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7355
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.46875,
      "epoch": 0.3554310011596444,
      "grad_norm": 0.0,
      "learning_rate": 3.223086586780054e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7356
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.125,
      "epoch": 0.35547931967529955,
      "grad_norm": 0.0,
      "learning_rate": 3.222844994201778e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7357
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.59375,
      "epoch": 0.3555276381909548,
      "grad_norm": 0.0,
      "learning_rate": 3.222603401623502e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7358
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.625,
      "epoch": 0.35557595670660996,
      "grad_norm": 0.0,
      "learning_rate": 3.222361809045226e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7359
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.875,
      "epoch": 0.3556242752222652,
      "grad_norm": 0.0,
      "learning_rate": 3.22212021646695e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7360
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.90625,
      "epoch": 0.3556725937379204,
      "grad_norm": 0.0,
      "learning_rate": 3.221878623888674e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7361
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.5625,
      "epoch": 0.35572091225357555,
      "grad_norm": 0.0,
      "learning_rate": 3.221637031310398e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7362
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.375,
      "epoch": 0.3557692307692308,
      "grad_norm": 0.0,
      "learning_rate": 3.221395438732122e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7363
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.96875,
      "epoch": 0.35581754928488596,
      "grad_norm": 0.0,
      "learning_rate": 3.2211538461538464e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7364
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.6875,
      "epoch": 0.3558658678005412,
      "grad_norm": 0.0,
      "learning_rate": 3.22091225357557e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7365
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 65.71875,
      "epoch": 0.35591418631619637,
      "grad_norm": 0.0,
      "learning_rate": 3.2206706609972936e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 7366
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.875,
      "epoch": 0.35596250483185155,
      "grad_norm": 0.0,
      "learning_rate": 3.2204290684190183e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7367
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.4375,
      "epoch": 0.3560108233475068,
      "grad_norm": 0.0,
      "learning_rate": 3.220187475840742e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7368
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 71.40625,
      "epoch": 0.35605914186316195,
      "grad_norm": 28.55510736335018,
      "learning_rate": 3.219945883262466e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7369
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.625,
      "epoch": 0.3561074603788172,
      "grad_norm": 0.0,
      "learning_rate": 3.2197042906841897e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7370
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.9375,
      "epoch": 0.35615577889447236,
      "grad_norm": 0.0,
      "learning_rate": 3.2194626981059143e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7371
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.3125,
      "epoch": 0.35620409741012754,
      "grad_norm": 0.0,
      "learning_rate": 3.219221105527638e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7372
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.90625,
      "epoch": 0.3562524159257828,
      "grad_norm": 0.0,
      "learning_rate": 3.218979512949362e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7373
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.59375,
      "epoch": 0.35630073444143795,
      "grad_norm": 0.0,
      "learning_rate": 3.218737920371086e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7374
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.875,
      "epoch": 0.3563490529570932,
      "grad_norm": 0.0,
      "learning_rate": 3.2184963277928103e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7375
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.90625,
      "epoch": 0.35639737147274836,
      "grad_norm": 0.0,
      "learning_rate": 3.218254735214534e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7376
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.53125,
      "epoch": 0.35644568998840354,
      "grad_norm": 0.0,
      "learning_rate": 3.218013142636258e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7377
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.0,
      "epoch": 0.35649400850405877,
      "grad_norm": 0.0,
      "learning_rate": 3.217771550057982e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7378
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.28125,
      "epoch": 0.35654232701971394,
      "grad_norm": 0.0,
      "learning_rate": 3.217529957479706e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7379
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.25,
      "epoch": 0.3565906455353692,
      "grad_norm": 0.0,
      "learning_rate": 3.21728836490143e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7380
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.3125,
      "epoch": 0.35663896405102435,
      "grad_norm": 0.0,
      "learning_rate": 3.217046772323154e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7381
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.53125,
      "epoch": 0.35668728256667953,
      "grad_norm": 0.0,
      "learning_rate": 3.216805179744878e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7382
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.03125,
      "epoch": 0.35673560108233476,
      "grad_norm": 0.0,
      "learning_rate": 3.216563587166602e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7383
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.3125,
      "epoch": 0.35678391959798994,
      "grad_norm": 0.0,
      "learning_rate": 3.2163219945883265e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7384
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.71875,
      "epoch": 0.3568322381136452,
      "grad_norm": 0.0,
      "learning_rate": 3.21608040201005e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7385
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.3125,
      "epoch": 0.35688055662930035,
      "grad_norm": 0.0,
      "learning_rate": 3.215838809431774e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7386
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.875,
      "epoch": 0.3569288751449555,
      "grad_norm": 0.0,
      "learning_rate": 3.2155972168534984e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7387
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.1875,
      "epoch": 0.35697719366061076,
      "grad_norm": 0.0,
      "learning_rate": 3.215355624275222e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7388
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.4375,
      "epoch": 0.35702551217626594,
      "grad_norm": 0.0,
      "learning_rate": 3.215114031696946e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7389
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.75,
      "epoch": 0.35707383069192117,
      "grad_norm": 0.0,
      "learning_rate": 3.2148724391186697e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7390
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.25,
      "epoch": 0.35712214920757634,
      "grad_norm": 0.0,
      "learning_rate": 3.2146308465403944e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7391
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.6875,
      "epoch": 0.3571704677232315,
      "grad_norm": 0.0,
      "learning_rate": 3.214389253962118e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7392
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.125,
      "epoch": 0.35721878623888675,
      "grad_norm": 0.0,
      "learning_rate": 3.214147661383842e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7393
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.28125,
      "epoch": 0.35726710475454193,
      "grad_norm": 0.0,
      "learning_rate": 3.2139060688055663e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7394
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.84375,
      "epoch": 0.35731542327019716,
      "grad_norm": 0.0,
      "learning_rate": 3.2136644762272904e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7395
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.4375,
      "epoch": 0.35736374178585234,
      "grad_norm": 0.0,
      "learning_rate": 3.213422883649014e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7396
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.4375,
      "epoch": 0.3574120603015075,
      "grad_norm": 0.0,
      "learning_rate": 3.2131812910707387e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7397
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.8125,
      "epoch": 0.35746037881716275,
      "grad_norm": 0.0,
      "learning_rate": 3.2129396984924623e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7398
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.9375,
      "epoch": 0.3575086973328179,
      "grad_norm": 0.0,
      "learning_rate": 3.212698105914186e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7399
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.5,
      "epoch": 0.35755701584847316,
      "grad_norm": 0.0,
      "learning_rate": 3.21245651333591e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7400
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.75,
      "epoch": 0.35760533436412834,
      "grad_norm": 48.999119730216115,
      "learning_rate": 3.212214920757634e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7401
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.15625,
      "epoch": 0.3576536528797835,
      "grad_norm": 0.0,
      "learning_rate": 3.2119733281793583e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7402
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.71875,
      "epoch": 0.35770197139543874,
      "grad_norm": 0.0,
      "learning_rate": 3.211731735601082e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7403
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.5625,
      "epoch": 0.3577502899110939,
      "grad_norm": 0.0,
      "learning_rate": 3.2114901430228066e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7404
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.15625,
      "epoch": 0.35779860842674915,
      "grad_norm": 0.0,
      "learning_rate": 3.21124855044453e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7405
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.875,
      "epoch": 0.35784692694240433,
      "grad_norm": 0.0,
      "learning_rate": 3.2110069578662543e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7406
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.5,
      "epoch": 0.3578952454580595,
      "grad_norm": 0.0,
      "learning_rate": 3.2107653652879785e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7407
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.84375,
      "epoch": 0.35794356397371474,
      "grad_norm": 0.0,
      "learning_rate": 3.210523772709702e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7408
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.5,
      "epoch": 0.3579918824893699,
      "grad_norm": 0.0,
      "learning_rate": 3.210282180131426e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7409
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 71.78125,
      "epoch": 0.35804020100502515,
      "grad_norm": 0.0,
      "learning_rate": 3.21004058755315e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 7410
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.625,
      "epoch": 0.3580885195206803,
      "grad_norm": 0.0,
      "learning_rate": 3.2097989949748745e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7411
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.71875,
      "epoch": 0.3581368380363355,
      "grad_norm": 0.0,
      "learning_rate": 3.209557402396598e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 7412
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.65625,
      "epoch": 0.35818515655199074,
      "grad_norm": 0.0,
      "learning_rate": 3.209315809818322e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7413
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.84375,
      "epoch": 0.3582334750676459,
      "grad_norm": 0.0,
      "learning_rate": 3.2090742172400464e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7414
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.53125,
      "epoch": 0.35828179358330114,
      "grad_norm": 0.0,
      "learning_rate": 3.2088326246617705e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7415
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.75,
      "epoch": 0.3583301120989563,
      "grad_norm": 0.0,
      "learning_rate": 3.208591032083494e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7416
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.75,
      "epoch": 0.3583784306146115,
      "grad_norm": 0.0,
      "learning_rate": 3.208349439505219e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7417
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.625,
      "epoch": 0.35842674913026673,
      "grad_norm": 0.0,
      "learning_rate": 3.2081078469269424e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7418
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 67.28125,
      "epoch": 0.3584750676459219,
      "grad_norm": 0.0,
      "learning_rate": 3.207866254348666e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7419
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.09375,
      "epoch": 0.35852338616157714,
      "grad_norm": 0.0,
      "learning_rate": 3.20762466177039e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7420
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.5625,
      "epoch": 0.3585717046772323,
      "grad_norm": 0.0,
      "learning_rate": 3.207383069192114e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7421
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.625,
      "epoch": 0.3586200231928875,
      "grad_norm": 0.0,
      "learning_rate": 3.2071414766138384e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7422
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.25,
      "epoch": 0.3586683417085427,
      "grad_norm": 0.0,
      "learning_rate": 3.206899884035562e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7423
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.09375,
      "epoch": 0.3587166602241979,
      "grad_norm": 0.0,
      "learning_rate": 3.2066582914572867e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7424
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.40625,
      "epoch": 0.35876497873985314,
      "grad_norm": 0.0,
      "learning_rate": 3.2064166988790103e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7425
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 67.09375,
      "epoch": 0.3588132972555083,
      "grad_norm": 0.0,
      "learning_rate": 3.2061751063007344e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 7426
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.03125,
      "epoch": 0.3588616157711635,
      "grad_norm": 0.0,
      "learning_rate": 3.2059335137224585e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7427
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.5,
      "epoch": 0.3589099342868187,
      "grad_norm": 0.0,
      "learning_rate": 3.205691921144182e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7428
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.34375,
      "epoch": 0.3589582528024739,
      "grad_norm": 0.0,
      "learning_rate": 3.2054503285659063e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7429
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.875,
      "epoch": 0.35900657131812913,
      "grad_norm": 0.0,
      "learning_rate": 3.20520873598763e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7430
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.8125,
      "epoch": 0.3590548898337843,
      "grad_norm": 0.0,
      "learning_rate": 3.2049671434093546e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7431
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.15625,
      "epoch": 0.3591032083494395,
      "grad_norm": 0.0,
      "learning_rate": 3.204725550831078e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7432
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.15625,
      "epoch": 0.3591515268650947,
      "grad_norm": 0.0,
      "learning_rate": 3.2044839582528023e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7433
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.3125,
      "epoch": 0.3591998453807499,
      "grad_norm": 0.0,
      "learning_rate": 3.2042423656745264e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7434
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.34375,
      "epoch": 0.3592481638964051,
      "grad_norm": 0.0,
      "learning_rate": 3.2040007730962506e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7435
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.71875,
      "epoch": 0.3592964824120603,
      "grad_norm": 0.0,
      "learning_rate": 3.203759180517974e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7436
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.84375,
      "epoch": 0.3593448009277155,
      "grad_norm": 0.0,
      "learning_rate": 3.203517587939699e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7437
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.9375,
      "epoch": 0.3593931194433707,
      "grad_norm": 0.0,
      "learning_rate": 3.2032759953614225e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7438
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.0,
      "epoch": 0.3594414379590259,
      "grad_norm": 0.0,
      "learning_rate": 3.203034402783146e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7439
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 68.78125,
      "epoch": 0.3594897564746811,
      "grad_norm": 0.0,
      "learning_rate": 3.20279281020487e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 7440
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.9375,
      "epoch": 0.3595380749903363,
      "grad_norm": 0.0,
      "learning_rate": 3.2025512176265943e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7441
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.53125,
      "epoch": 0.3595863935059915,
      "grad_norm": 0.0,
      "learning_rate": 3.2023096250483185e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 7442
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.53125,
      "epoch": 0.3596347120216467,
      "grad_norm": 0.0,
      "learning_rate": 3.202068032470042e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7443
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.25,
      "epoch": 0.3596830305373019,
      "grad_norm": 0.0,
      "learning_rate": 3.201826439891767e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7444
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.84375,
      "epoch": 0.3597313490529571,
      "grad_norm": 0.0,
      "learning_rate": 3.2015848473134903e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7445
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.9375,
      "epoch": 0.3597796675686123,
      "grad_norm": 0.0,
      "learning_rate": 3.2013432547352145e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7446
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.15625,
      "epoch": 0.35982798608426747,
      "grad_norm": 0.0,
      "learning_rate": 3.2011016621569386e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7447
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.21875,
      "epoch": 0.3598763045999227,
      "grad_norm": 0.0,
      "learning_rate": 3.200860069578663e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7448
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.28125,
      "epoch": 0.3599246231155779,
      "grad_norm": 0.0,
      "learning_rate": 3.2006184770003864e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7449
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.0,
      "epoch": 0.3599729416312331,
      "grad_norm": 0.0,
      "learning_rate": 3.20037688442211e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7450
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.9375,
      "epoch": 0.3600212601468883,
      "grad_norm": 0.0,
      "learning_rate": 3.2001352918438346e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7451
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 67.96875,
      "epoch": 0.36006957866254347,
      "grad_norm": 48.93406302361326,
      "learning_rate": 3.199893699265558e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7452
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.125,
      "epoch": 0.3601178971781987,
      "grad_norm": 0.0,
      "learning_rate": 3.1996521066872824e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7453
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.9375,
      "epoch": 0.3601662156938539,
      "grad_norm": 0.0,
      "learning_rate": 3.1994105141090065e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7454
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.40625,
      "epoch": 0.3602145342095091,
      "grad_norm": 0.0,
      "learning_rate": 3.1991689215307307e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 7455
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.1875,
      "epoch": 0.3602628527251643,
      "grad_norm": 0.0,
      "learning_rate": 3.198927328952454e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7456
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.25,
      "epoch": 0.36031117124081946,
      "grad_norm": 0.0,
      "learning_rate": 3.198685736374179e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7457
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.78125,
      "epoch": 0.3603594897564747,
      "grad_norm": 0.0,
      "learning_rate": 3.1984441437959025e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7458
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.65625,
      "epoch": 0.36040780827212987,
      "grad_norm": 0.0,
      "learning_rate": 3.198202551217626e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7459
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.28125,
      "epoch": 0.3604561267877851,
      "grad_norm": 0.0,
      "learning_rate": 3.1979609586393503e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7460
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.21875,
      "epoch": 0.3605044453034403,
      "grad_norm": 0.0,
      "learning_rate": 3.1977193660610744e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7461
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.25,
      "epoch": 0.36055276381909546,
      "grad_norm": 0.0,
      "learning_rate": 3.1974777734827985e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7462
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.0,
      "epoch": 0.3606010823347507,
      "grad_norm": 0.0,
      "learning_rate": 3.197236180904522e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7463
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.5625,
      "epoch": 0.36064940085040587,
      "grad_norm": 0.0,
      "learning_rate": 3.196994588326247e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7464
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.9375,
      "epoch": 0.3606977193660611,
      "grad_norm": 0.0,
      "learning_rate": 3.1967529957479704e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7465
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.09375,
      "epoch": 0.3607460378817163,
      "grad_norm": 0.0,
      "learning_rate": 3.1965114031696946e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 7466
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.59375,
      "epoch": 0.36079435639737145,
      "grad_norm": 0.0,
      "learning_rate": 3.1962698105914187e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7467
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.84375,
      "epoch": 0.3608426749130267,
      "grad_norm": 0.0,
      "learning_rate": 3.196028218013143e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 7468
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.9375,
      "epoch": 0.36089099342868186,
      "grad_norm": 0.0,
      "learning_rate": 3.1957866254348664e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7469
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.03125,
      "epoch": 0.3609393119443371,
      "grad_norm": 0.0,
      "learning_rate": 3.19554503285659e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7470
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.78125,
      "epoch": 0.36098763045999227,
      "grad_norm": 0.0,
      "learning_rate": 3.1953034402783147e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7471
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.125,
      "epoch": 0.36103594897564745,
      "grad_norm": 0.0,
      "learning_rate": 3.1950618477000383e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7472
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.03125,
      "epoch": 0.3610842674913027,
      "grad_norm": 0.0,
      "learning_rate": 3.1948202551217625e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7473
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.40625,
      "epoch": 0.36113258600695786,
      "grad_norm": 0.0,
      "learning_rate": 3.1945786625434866e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7474
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.375,
      "epoch": 0.3611809045226131,
      "grad_norm": 0.0,
      "learning_rate": 3.1943370699652107e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7475
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.96875,
      "epoch": 0.36122922303826827,
      "grad_norm": 0.0,
      "learning_rate": 3.1940954773869343e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7476
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.59375,
      "epoch": 0.36127754155392344,
      "grad_norm": 0.0,
      "learning_rate": 3.193853884808659e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7477
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.09375,
      "epoch": 0.3613258600695787,
      "grad_norm": 0.0,
      "learning_rate": 3.1936122922303826e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7478
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.46875,
      "epoch": 0.36137417858523385,
      "grad_norm": 0.0,
      "learning_rate": 3.193370699652107e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7479
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.90625,
      "epoch": 0.3614224971008891,
      "grad_norm": 0.0,
      "learning_rate": 3.1931291070738304e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7480
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.90625,
      "epoch": 0.36147081561654426,
      "grad_norm": 0.0,
      "learning_rate": 3.1928875144955545e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7481
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.09375,
      "epoch": 0.36151913413219944,
      "grad_norm": 0.0,
      "learning_rate": 3.1926459219172786e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 7482
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.4375,
      "epoch": 0.36156745264785467,
      "grad_norm": 0.0,
      "learning_rate": 3.192404329339002e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7483
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.4375,
      "epoch": 0.36161577116350985,
      "grad_norm": 0.0,
      "learning_rate": 3.192162736760727e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7484
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.59375,
      "epoch": 0.3616640896791651,
      "grad_norm": 0.0,
      "learning_rate": 3.1919211441824505e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7485
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 73.75,
      "epoch": 0.36171240819482026,
      "grad_norm": 169.2028239661782,
      "learning_rate": 3.1916795516041746e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7486
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.625,
      "epoch": 0.36176072671047543,
      "grad_norm": 0.0,
      "learning_rate": 3.191437959025899e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7487
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.3125,
      "epoch": 0.36180904522613067,
      "grad_norm": 0.0,
      "learning_rate": 3.191196366447623e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7488
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.15625,
      "epoch": 0.36185736374178584,
      "grad_norm": 0.0,
      "learning_rate": 3.1909547738693465e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7489
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.6875,
      "epoch": 0.3619056822574411,
      "grad_norm": 0.0,
      "learning_rate": 3.19071318129107e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7490
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.53125,
      "epoch": 0.36195400077309625,
      "grad_norm": 0.0,
      "learning_rate": 3.190471588712795e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7491
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.25,
      "epoch": 0.36200231928875143,
      "grad_norm": 0.0,
      "learning_rate": 3.1902299961345184e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7492
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.96875,
      "epoch": 0.36205063780440666,
      "grad_norm": 0.0,
      "learning_rate": 3.1899884035562425e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7493
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.0625,
      "epoch": 0.36209895632006184,
      "grad_norm": 0.0,
      "learning_rate": 3.1897468109779667e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 7494
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.5,
      "epoch": 0.36214727483571707,
      "grad_norm": 0.0,
      "learning_rate": 3.189505218399691e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7495
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.3125,
      "epoch": 0.36219559335137225,
      "grad_norm": 0.0,
      "learning_rate": 3.1892636258214144e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7496
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.8125,
      "epoch": 0.3622439118670274,
      "grad_norm": 0.0,
      "learning_rate": 3.189022033243139e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7497
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.71875,
      "epoch": 0.36229223038268266,
      "grad_norm": 0.0,
      "learning_rate": 3.1887804406648627e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7498
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.09375,
      "epoch": 0.36234054889833783,
      "grad_norm": 0.0,
      "learning_rate": 3.188538848086587e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7499
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.3125,
      "epoch": 0.36238886741399307,
      "grad_norm": 0.0,
      "learning_rate": 3.1882972555083104e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7500
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.84375,
      "epoch": 0.36243718592964824,
      "grad_norm": 0.0,
      "learning_rate": 3.1880556629300346e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7501
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.3125,
      "epoch": 0.3624855044453034,
      "grad_norm": 0.0,
      "learning_rate": 3.1878140703517587e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7502
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.8125,
      "epoch": 0.36253382296095865,
      "grad_norm": 0.0,
      "learning_rate": 3.1875724777734823e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7503
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.375,
      "epoch": 0.36258214147661383,
      "grad_norm": 0.0,
      "learning_rate": 3.187330885195207e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7504
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.375,
      "epoch": 0.36263045999226906,
      "grad_norm": 0.0,
      "learning_rate": 3.1870892926169306e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7505
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.5,
      "epoch": 0.36267877850792424,
      "grad_norm": 0.0,
      "learning_rate": 3.1868477000386547e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7506
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.71875,
      "epoch": 0.3627270970235794,
      "grad_norm": 0.0,
      "learning_rate": 3.186606107460379e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 7507
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 70.25,
      "epoch": 0.36277541553923465,
      "grad_norm": 13.614146847402878,
      "learning_rate": 3.186364514882103e-07,
      "loss": 0.0,
      "reward": 1.21875,
      "reward_std": 0.18600594997406006,
      "rewards/accuracy_reward": 0.15625,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 7508
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.65625,
      "epoch": 0.3628237340548898,
      "grad_norm": 0.0,
      "learning_rate": 3.1861229223038266e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7509
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.21875,
      "epoch": 0.36287205257054506,
      "grad_norm": 0.0,
      "learning_rate": 3.18588132972555e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7510
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.0,
      "epoch": 0.36292037108620023,
      "grad_norm": 0.0,
      "learning_rate": 3.185639737147275e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7511
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.625,
      "epoch": 0.3629686896018554,
      "grad_norm": 0.0,
      "learning_rate": 3.1853981445689985e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7512
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.71875,
      "epoch": 0.36301700811751064,
      "grad_norm": 0.0,
      "learning_rate": 3.1851565519907226e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 7513
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.28125,
      "epoch": 0.3630653266331658,
      "grad_norm": 0.0,
      "learning_rate": 3.184914959412447e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7514
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.375,
      "epoch": 0.36311364514882105,
      "grad_norm": 0.0,
      "learning_rate": 3.184673366834171e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7515
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.75,
      "epoch": 0.36316196366447623,
      "grad_norm": 0.0,
      "learning_rate": 3.1844317742558945e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7516
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.3125,
      "epoch": 0.3632102821801314,
      "grad_norm": 0.0,
      "learning_rate": 3.184190181677619e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7517
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.09375,
      "epoch": 0.36325860069578664,
      "grad_norm": 0.0,
      "learning_rate": 3.183948589099343e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7518
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.75,
      "epoch": 0.3633069192114418,
      "grad_norm": 0.0,
      "learning_rate": 3.183706996521067e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7519
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.78125,
      "epoch": 0.36335523772709705,
      "grad_norm": 0.0,
      "learning_rate": 3.1834654039427905e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7520
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.09375,
      "epoch": 0.3634035562427522,
      "grad_norm": 0.0,
      "learning_rate": 3.1832238113645146e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7521
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.0625,
      "epoch": 0.3634518747584074,
      "grad_norm": 0.0,
      "learning_rate": 3.182982218786239e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7522
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.84375,
      "epoch": 0.36350019327406263,
      "grad_norm": 0.0,
      "learning_rate": 3.1827406262079624e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7523
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.21875,
      "epoch": 0.3635485117897178,
      "grad_norm": 0.0,
      "learning_rate": 3.182499033629687e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7524
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.625,
      "epoch": 0.36359683030537304,
      "grad_norm": 0.0,
      "learning_rate": 3.1822574410514107e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7525
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.875,
      "epoch": 0.3636451488210282,
      "grad_norm": 0.0,
      "learning_rate": 3.182015848473135e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7526
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.625,
      "epoch": 0.3636934673366834,
      "grad_norm": 0.0,
      "learning_rate": 3.181774255894859e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 7527
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.28125,
      "epoch": 0.36374178585233863,
      "grad_norm": 0.0,
      "learning_rate": 3.181532663316583e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7528
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.875,
      "epoch": 0.3637901043679938,
      "grad_norm": 0.0,
      "learning_rate": 3.1812910707383067e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7529
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.5,
      "epoch": 0.36383842288364904,
      "grad_norm": 0.0,
      "learning_rate": 3.181049478160031e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7530
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.625,
      "epoch": 0.3638867413993042,
      "grad_norm": 0.0,
      "learning_rate": 3.180807885581755e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7531
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.84375,
      "epoch": 0.3639350599149594,
      "grad_norm": 0.0,
      "learning_rate": 3.1805662930034786e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7532
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.6875,
      "epoch": 0.3639833784306146,
      "grad_norm": 0.0,
      "learning_rate": 3.1803247004252027e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7533
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.59375,
      "epoch": 0.3640316969462698,
      "grad_norm": 0.0,
      "learning_rate": 3.180083107846927e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7534
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.96875,
      "epoch": 0.36408001546192503,
      "grad_norm": 0.0,
      "learning_rate": 3.179841515268651e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 7535
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.84375,
      "epoch": 0.3641283339775802,
      "grad_norm": 0.0,
      "learning_rate": 3.1795999226903746e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7536
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.625,
      "epoch": 0.3641766524932354,
      "grad_norm": 0.0,
      "learning_rate": 3.179358330112099e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7537
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.8125,
      "epoch": 0.3642249710088906,
      "grad_norm": 0.0,
      "learning_rate": 3.179116737533823e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7538
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.375,
      "epoch": 0.3642732895245458,
      "grad_norm": 0.0,
      "learning_rate": 3.178875144955547e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7539
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.09375,
      "epoch": 0.36432160804020103,
      "grad_norm": 0.0,
      "learning_rate": 3.1786335523772706e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7540
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.1875,
      "epoch": 0.3643699265558562,
      "grad_norm": 0.0,
      "learning_rate": 3.178391959798995e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7541
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.625,
      "epoch": 0.3644182450715114,
      "grad_norm": 0.0,
      "learning_rate": 3.178150367220719e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7542
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.0,
      "epoch": 0.3644665635871666,
      "grad_norm": 0.0,
      "learning_rate": 3.1779087746424425e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7543
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 64.90625,
      "epoch": 0.3645148821028218,
      "grad_norm": 0.0,
      "learning_rate": 3.177667182064167e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 7544
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.53125,
      "epoch": 0.364563200618477,
      "grad_norm": 0.0,
      "learning_rate": 3.177425589485891e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7545
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 69.5625,
      "epoch": 0.3646115191341322,
      "grad_norm": 0.0,
      "learning_rate": 3.177183996907615e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7546
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.9375,
      "epoch": 0.3646598376497874,
      "grad_norm": 0.0,
      "learning_rate": 3.176942404329339e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7547
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.65625,
      "epoch": 0.3647081561654426,
      "grad_norm": 0.0,
      "learning_rate": 3.176700811751063e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7548
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.78125,
      "epoch": 0.3647564746810978,
      "grad_norm": 0.0,
      "learning_rate": 3.176459219172787e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7549
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 62.0625,
      "epoch": 0.364804793196753,
      "grad_norm": 0.0,
      "learning_rate": 3.176217626594511e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 7550
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.71875,
      "epoch": 0.3648531117124082,
      "grad_norm": 0.0,
      "learning_rate": 3.175976034016235e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7551
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.75,
      "epoch": 0.3649014302280634,
      "grad_norm": 0.0,
      "learning_rate": 3.1757344414379586e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7552
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.0,
      "epoch": 0.3649497487437186,
      "grad_norm": 0.0,
      "learning_rate": 3.175492848859683e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7553
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 58.65625,
      "epoch": 0.3649980672593738,
      "grad_norm": 0.0,
      "learning_rate": 3.175251256281407e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 7554
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.4375,
      "epoch": 0.365046385775029,
      "grad_norm": 0.0,
      "learning_rate": 3.175009663703131e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7555
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.90625,
      "epoch": 0.3650947042906842,
      "grad_norm": 0.0,
      "learning_rate": 3.1747680711248547e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7556
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.34375,
      "epoch": 0.36514302280633937,
      "grad_norm": 0.0,
      "learning_rate": 3.1745264785465793e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7557
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.90625,
      "epoch": 0.3651913413219946,
      "grad_norm": 0.0,
      "learning_rate": 3.174284885968303e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7558
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.375,
      "epoch": 0.3652396598376498,
      "grad_norm": 0.0,
      "learning_rate": 3.174043293390027e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7559
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.25,
      "epoch": 0.365287978353305,
      "grad_norm": 0.0,
      "learning_rate": 3.1738017008117507e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7560
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.375,
      "epoch": 0.3653362968689602,
      "grad_norm": 0.0,
      "learning_rate": 3.1735601082334753e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7561
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.375,
      "epoch": 0.36538461538461536,
      "grad_norm": 0.0,
      "learning_rate": 3.173318515655199e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7562
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.5,
      "epoch": 0.3654329339002706,
      "grad_norm": 0.0,
      "learning_rate": 3.1730769230769225e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7563
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.15625,
      "epoch": 0.3654812524159258,
      "grad_norm": 0.0,
      "learning_rate": 3.172835330498647e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7564
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.21875,
      "epoch": 0.365529570931581,
      "grad_norm": 0.0,
      "learning_rate": 3.172593737920371e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7565
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.78125,
      "epoch": 0.3655778894472362,
      "grad_norm": 0.0,
      "learning_rate": 3.172352145342095e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7566
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.1875,
      "epoch": 0.36562620796289136,
      "grad_norm": 0.0,
      "learning_rate": 3.172110552763819e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7567
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.0625,
      "epoch": 0.3656745264785466,
      "grad_norm": 0.0,
      "learning_rate": 3.171868960185543e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7568
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.78125,
      "epoch": 0.36572284499420177,
      "grad_norm": 0.0,
      "learning_rate": 3.171627367607267e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7569
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.75,
      "epoch": 0.365771163509857,
      "grad_norm": 0.0,
      "learning_rate": 3.171385775028991e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7570
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.5,
      "epoch": 0.3658194820255122,
      "grad_norm": 0.0,
      "learning_rate": 3.171144182450715e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7571
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.28125,
      "epoch": 0.36586780054116735,
      "grad_norm": 0.0,
      "learning_rate": 3.1709025898724387e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7572
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.90625,
      "epoch": 0.3659161190568226,
      "grad_norm": 0.0,
      "learning_rate": 3.170660997294163e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7573
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.34375,
      "epoch": 0.36596443757247776,
      "grad_norm": 0.0,
      "learning_rate": 3.170419404715887e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7574
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.21875,
      "epoch": 0.366012756088133,
      "grad_norm": 0.0,
      "learning_rate": 3.170177812137611e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7575
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.0,
      "epoch": 0.3660610746037882,
      "grad_norm": 0.0,
      "learning_rate": 3.1699362195593347e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7576
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.3125,
      "epoch": 0.36610939311944335,
      "grad_norm": 0.0,
      "learning_rate": 3.1696946269810594e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7577
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.5625,
      "epoch": 0.3661577116350986,
      "grad_norm": 0.0,
      "learning_rate": 3.169453034402783e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7578
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.84375,
      "epoch": 0.36620603015075376,
      "grad_norm": 0.0,
      "learning_rate": 3.169211441824507e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7579
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.9375,
      "epoch": 0.366254348666409,
      "grad_norm": 0.0,
      "learning_rate": 3.168969849246231e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7580
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.65625,
      "epoch": 0.36630266718206417,
      "grad_norm": 0.0,
      "learning_rate": 3.1687282566679554e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7581
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.21875,
      "epoch": 0.36635098569771934,
      "grad_norm": 0.0,
      "learning_rate": 3.168486664089679e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7582
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.75,
      "epoch": 0.3663993042133746,
      "grad_norm": 0.0,
      "learning_rate": 3.1682450715114026e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7583
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.6875,
      "epoch": 0.36644762272902975,
      "grad_norm": 0.0,
      "learning_rate": 3.1680034789331273e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7584
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.90625,
      "epoch": 0.366495941244685,
      "grad_norm": 0.0,
      "learning_rate": 3.167761886354851e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7585
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.65625,
      "epoch": 0.36654425976034016,
      "grad_norm": 0.0,
      "learning_rate": 3.167520293776575e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7586
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.875,
      "epoch": 0.36659257827599534,
      "grad_norm": 0.0,
      "learning_rate": 3.167278701198299e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7587
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.65625,
      "epoch": 0.3666408967916506,
      "grad_norm": 0.0,
      "learning_rate": 3.1670371086200233e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7588
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.0,
      "epoch": 0.36668921530730575,
      "grad_norm": 0.0,
      "learning_rate": 3.166795516041747e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7589
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.96875,
      "epoch": 0.366737533822961,
      "grad_norm": 0.0,
      "learning_rate": 3.166553923463471e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7590
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.125,
      "epoch": 0.36678585233861616,
      "grad_norm": 0.0,
      "learning_rate": 3.166312330885195e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7591
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.1875,
      "epoch": 0.36683417085427134,
      "grad_norm": 0.0,
      "learning_rate": 3.1660707383069193e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7592
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.46875,
      "epoch": 0.36688248936992657,
      "grad_norm": 0.0,
      "learning_rate": 3.165829145728643e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7593
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.5,
      "epoch": 0.36693080788558174,
      "grad_norm": 0.0,
      "learning_rate": 3.165587553150367e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7594
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.15625,
      "epoch": 0.366979126401237,
      "grad_norm": 0.0,
      "learning_rate": 3.165345960572091e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7595
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.15625,
      "epoch": 0.36702744491689215,
      "grad_norm": 0.0,
      "learning_rate": 3.165104367993815e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7596
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.25,
      "epoch": 0.36707576343254733,
      "grad_norm": 0.0,
      "learning_rate": 3.1648627754155395e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7597
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 66.40625,
      "epoch": 0.36712408194820256,
      "grad_norm": 26.080778094321936,
      "learning_rate": 3.164621182837263e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7598
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.65625,
      "epoch": 0.36717240046385774,
      "grad_norm": 0.0,
      "learning_rate": 3.164379590258987e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7599
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.25,
      "epoch": 0.367220718979513,
      "grad_norm": 0.0,
      "learning_rate": 3.164137997680711e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7600
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.8125,
      "epoch": 0.36726903749516815,
      "grad_norm": 0.0,
      "learning_rate": 3.1638964051024355e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 7601
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.59375,
      "epoch": 0.3673173560108233,
      "grad_norm": 0.0,
      "learning_rate": 3.163654812524159e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7602
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.9375,
      "epoch": 0.36736567452647856,
      "grad_norm": 0.0,
      "learning_rate": 3.1634132199458827e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7603
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 67.65625,
      "epoch": 0.36741399304213374,
      "grad_norm": 0.0,
      "learning_rate": 3.1631716273676074e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 7604
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.9375,
      "epoch": 0.36746231155778897,
      "grad_norm": 0.0,
      "learning_rate": 3.162930034789331e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7605
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.53125,
      "epoch": 0.36751063007344414,
      "grad_norm": 0.0,
      "learning_rate": 3.162688442211055e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7606
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.53125,
      "epoch": 0.3675589485890993,
      "grad_norm": 0.0,
      "learning_rate": 3.162446849632779e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7607
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.53125,
      "epoch": 0.36760726710475455,
      "grad_norm": 0.0,
      "learning_rate": 3.1622052570545034e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7608
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.53125,
      "epoch": 0.36765558562040973,
      "grad_norm": 0.0,
      "learning_rate": 3.161963664476227e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7609
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.09375,
      "epoch": 0.36770390413606496,
      "grad_norm": 0.0,
      "learning_rate": 3.161722071897951e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7610
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.53125,
      "epoch": 0.36775222265172014,
      "grad_norm": 0.0,
      "learning_rate": 3.161480479319675e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7611
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.4375,
      "epoch": 0.3678005411673753,
      "grad_norm": 0.0,
      "learning_rate": 3.1612388867413994e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7612
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.4375,
      "epoch": 0.36784885968303055,
      "grad_norm": 0.0,
      "learning_rate": 3.160997294163123e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7613
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.90625,
      "epoch": 0.3678971781986857,
      "grad_norm": 0.0,
      "learning_rate": 3.160755701584847e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7614
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.375,
      "epoch": 0.36794549671434096,
      "grad_norm": 0.0,
      "learning_rate": 3.1605141090065713e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7615
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 65.0625,
      "epoch": 0.36799381522999614,
      "grad_norm": 0.0,
      "learning_rate": 3.160272516428295e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7616
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.1875,
      "epoch": 0.3680421337456513,
      "grad_norm": 0.0,
      "learning_rate": 3.1600309238500196e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7617
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.71875,
      "epoch": 0.36809045226130654,
      "grad_norm": 0.0,
      "learning_rate": 3.159789331271743e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7618
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.875,
      "epoch": 0.3681387707769617,
      "grad_norm": 0.0,
      "learning_rate": 3.1595477386934673e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7619
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 67.40625,
      "epoch": 0.36818708929261695,
      "grad_norm": 0.0,
      "learning_rate": 3.159306146115191e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 7620
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.0625,
      "epoch": 0.36823540780827213,
      "grad_norm": 0.0,
      "learning_rate": 3.1590645535369156e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7621
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.46875,
      "epoch": 0.3682837263239273,
      "grad_norm": 0.0,
      "learning_rate": 3.158822960958639e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7622
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.125,
      "epoch": 0.36833204483958254,
      "grad_norm": 0.0,
      "learning_rate": 3.158581368380363e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7623
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.1875,
      "epoch": 0.3683803633552377,
      "grad_norm": 0.0,
      "learning_rate": 3.1583397758020874e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7624
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.46875,
      "epoch": 0.36842868187089295,
      "grad_norm": 0.0,
      "learning_rate": 3.158098183223811e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7625
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.90625,
      "epoch": 0.3684770003865481,
      "grad_norm": 0.0,
      "learning_rate": 3.157856590645535e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7626
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.34375,
      "epoch": 0.3685253189022033,
      "grad_norm": 0.0,
      "learning_rate": 3.1576149980672593e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7627
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.6875,
      "epoch": 0.36857363741785854,
      "grad_norm": 0.0,
      "learning_rate": 3.1573734054889835e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7628
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.15625,
      "epoch": 0.3686219559335137,
      "grad_norm": 0.0,
      "learning_rate": 3.157131812910707e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7629
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.09375,
      "epoch": 0.36867027444916894,
      "grad_norm": 0.0,
      "learning_rate": 3.156890220332431e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7630
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.28125,
      "epoch": 0.3687185929648241,
      "grad_norm": 0.0,
      "learning_rate": 3.1566486277541553e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7631
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.21875,
      "epoch": 0.3687669114804793,
      "grad_norm": 0.0,
      "learning_rate": 3.1564070351758795e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7632
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.5625,
      "epoch": 0.36881522999613453,
      "grad_norm": 0.0,
      "learning_rate": 3.156165442597603e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7633
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.1875,
      "epoch": 0.3688635485117897,
      "grad_norm": 0.0,
      "learning_rate": 3.155923850019327e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7634
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.3125,
      "epoch": 0.36891186702744494,
      "grad_norm": 0.0,
      "learning_rate": 3.1556822574410514e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7635
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.8125,
      "epoch": 0.3689601855431001,
      "grad_norm": 0.0,
      "learning_rate": 3.155440664862775e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7636
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.75,
      "epoch": 0.3690085040587553,
      "grad_norm": 0.0,
      "learning_rate": 3.1551990722844996e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7637
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.1875,
      "epoch": 0.3690568225744105,
      "grad_norm": 0.0,
      "learning_rate": 3.154957479706223e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7638
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.4375,
      "epoch": 0.3691051410900657,
      "grad_norm": 0.0,
      "learning_rate": 3.1547158871279474e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7639
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.0625,
      "epoch": 0.36915345960572093,
      "grad_norm": 0.0,
      "learning_rate": 3.154474294549671e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7640
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.1875,
      "epoch": 0.3692017781213761,
      "grad_norm": 0.0,
      "learning_rate": 3.1542327019713956e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7641
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.53125,
      "epoch": 0.3692500966370313,
      "grad_norm": 0.0,
      "learning_rate": 3.153991109393119e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7642
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.25,
      "epoch": 0.3692984151526865,
      "grad_norm": 0.0,
      "learning_rate": 3.1537495168148434e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7643
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.21875,
      "epoch": 0.3693467336683417,
      "grad_norm": 0.0,
      "learning_rate": 3.1535079242365675e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7644
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.09375,
      "epoch": 0.36939505218399693,
      "grad_norm": 0.0,
      "learning_rate": 3.153266331658291e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7645
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.78125,
      "epoch": 0.3694433706996521,
      "grad_norm": 0.0,
      "learning_rate": 3.1530247390800153e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7646
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.34375,
      "epoch": 0.3694916892153073,
      "grad_norm": 0.0,
      "learning_rate": 3.1527831465017394e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7647
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 71.875,
      "epoch": 0.3695400077309625,
      "grad_norm": 0.0,
      "learning_rate": 3.1525415539234635e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7648
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.25,
      "epoch": 0.3695883262466177,
      "grad_norm": 0.0,
      "learning_rate": 3.152299961345187e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 7649
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.84375,
      "epoch": 0.3696366447622729,
      "grad_norm": 0.0,
      "learning_rate": 3.1520583687669113e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7650
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.5625,
      "epoch": 0.3696849632779281,
      "grad_norm": 0.0,
      "learning_rate": 3.1518167761886354e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7651
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.8125,
      "epoch": 0.3697332817935833,
      "grad_norm": 0.0,
      "learning_rate": 3.1515751836103596e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7652
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.6875,
      "epoch": 0.3697816003092385,
      "grad_norm": 0.0,
      "learning_rate": 3.151333591032083e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7653
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 67.8125,
      "epoch": 0.3698299188248937,
      "grad_norm": 0.0,
      "learning_rate": 3.151091998453808e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7654
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.9375,
      "epoch": 0.3698782373405489,
      "grad_norm": 0.0,
      "learning_rate": 3.1508504058755314e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7655
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.09375,
      "epoch": 0.3699265558562041,
      "grad_norm": 0.0,
      "learning_rate": 3.150608813297255e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7656
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.8125,
      "epoch": 0.3699748743718593,
      "grad_norm": 0.0,
      "learning_rate": 3.1503672207189797e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7657
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.65625,
      "epoch": 0.3700231928875145,
      "grad_norm": 0.0,
      "learning_rate": 3.1501256281407033e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7658
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.25,
      "epoch": 0.3700715114031697,
      "grad_norm": 0.0,
      "learning_rate": 3.1498840355624275e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7659
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.8125,
      "epoch": 0.3701198299188249,
      "grad_norm": 0.0,
      "learning_rate": 3.149642442984151e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7660
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.03125,
      "epoch": 0.3701681484344801,
      "grad_norm": 0.0,
      "learning_rate": 3.1494008504058757e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7661
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.96875,
      "epoch": 0.37021646695013527,
      "grad_norm": 0.0,
      "learning_rate": 3.1491592578275993e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7662
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.71875,
      "epoch": 0.3702647854657905,
      "grad_norm": 0.0,
      "learning_rate": 3.1489176652493235e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 7663
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.96875,
      "epoch": 0.3703131039814457,
      "grad_norm": 0.0,
      "learning_rate": 3.1486760726710476e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7664
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.21875,
      "epoch": 0.3703614224971009,
      "grad_norm": 0.0,
      "learning_rate": 3.148434480092771e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7665
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.5,
      "epoch": 0.3704097410127561,
      "grad_norm": 0.0,
      "learning_rate": 3.1481928875144953e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7666
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.0625,
      "epoch": 0.37045805952841127,
      "grad_norm": 0.0,
      "learning_rate": 3.1479512949362195e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7667
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.5625,
      "epoch": 0.3705063780440665,
      "grad_norm": 0.0,
      "learning_rate": 3.1477097023579436e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7668
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.96875,
      "epoch": 0.3705546965597217,
      "grad_norm": 0.0,
      "learning_rate": 3.147468109779667e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7669
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.0,
      "epoch": 0.3706030150753769,
      "grad_norm": 0.0,
      "learning_rate": 3.1472265172013914e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7670
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.09375,
      "epoch": 0.3706513335910321,
      "grad_norm": 0.0,
      "learning_rate": 3.1469849246231155e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7671
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.125,
      "epoch": 0.37069965210668726,
      "grad_norm": 0.0,
      "learning_rate": 3.1467433320448396e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7672
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.28125,
      "epoch": 0.3707479706223425,
      "grad_norm": 0.0,
      "learning_rate": 3.146501739466563e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7673
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.1875,
      "epoch": 0.37079628913799767,
      "grad_norm": 0.0,
      "learning_rate": 3.146260146888288e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7674
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.8125,
      "epoch": 0.3708446076536529,
      "grad_norm": 0.0,
      "learning_rate": 3.1460185543100115e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7675
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.09375,
      "epoch": 0.3708929261693081,
      "grad_norm": 0.0,
      "learning_rate": 3.145776961731735e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7676
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.125,
      "epoch": 0.37094124468496326,
      "grad_norm": 0.0,
      "learning_rate": 3.14553536915346e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7677
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.46875,
      "epoch": 0.3709895632006185,
      "grad_norm": 0.0,
      "learning_rate": 3.1452937765751834e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7678
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.65625,
      "epoch": 0.37103788171627367,
      "grad_norm": 0.0,
      "learning_rate": 3.1450521839969075e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7679
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.875,
      "epoch": 0.3710862002319289,
      "grad_norm": 0.0,
      "learning_rate": 3.144810591418631e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7680
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.5,
      "epoch": 0.3711345187475841,
      "grad_norm": 0.0,
      "learning_rate": 3.144568998840356e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7681
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.84375,
      "epoch": 0.37118283726323925,
      "grad_norm": 0.0,
      "learning_rate": 3.1443274062620794e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 7682
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.71875,
      "epoch": 0.3712311557788945,
      "grad_norm": 0.0,
      "learning_rate": 3.1440858136838035e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7683
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.375,
      "epoch": 0.37127947429454966,
      "grad_norm": 0.0,
      "learning_rate": 3.1438442211055277e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7684
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.75,
      "epoch": 0.3713277928102049,
      "grad_norm": 0.0,
      "learning_rate": 3.1436026285272513e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7685
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.4375,
      "epoch": 0.37137611132586007,
      "grad_norm": 0.0,
      "learning_rate": 3.1433610359489754e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7686
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.875,
      "epoch": 0.37142442984151525,
      "grad_norm": 0.0,
      "learning_rate": 3.1431194433706996e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7687
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.875,
      "epoch": 0.3714727483571705,
      "grad_norm": 0.0,
      "learning_rate": 3.1428778507924237e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7688
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 67.46875,
      "epoch": 0.37152106687282566,
      "grad_norm": 0.0,
      "learning_rate": 3.1426362582141473e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 7689
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.9375,
      "epoch": 0.3715693853884809,
      "grad_norm": 0.0,
      "learning_rate": 3.1423946656358714e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7690
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.5625,
      "epoch": 0.37161770390413607,
      "grad_norm": 0.0,
      "learning_rate": 3.1421530730575956e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 7691
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.96875,
      "epoch": 0.37166602241979124,
      "grad_norm": 0.0,
      "learning_rate": 3.1419114804793197e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7692
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.5625,
      "epoch": 0.3717143409354465,
      "grad_norm": 0.0,
      "learning_rate": 3.1416698879010433e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7693
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.4375,
      "epoch": 0.37176265945110165,
      "grad_norm": 0.0,
      "learning_rate": 3.141428295322768e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7694
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.9375,
      "epoch": 0.3718109779667569,
      "grad_norm": 0.0,
      "learning_rate": 3.1411867027444916e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7695
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.59375,
      "epoch": 0.37185929648241206,
      "grad_norm": 0.0,
      "learning_rate": 3.140945110166215e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7696
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.53125,
      "epoch": 0.37190761499806724,
      "grad_norm": 0.0,
      "learning_rate": 3.14070351758794e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7697
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.5625,
      "epoch": 0.37195593351372247,
      "grad_norm": 0.0,
      "learning_rate": 3.1404619250096635e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7698
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.75,
      "epoch": 0.37200425202937765,
      "grad_norm": 0.0,
      "learning_rate": 3.1402203324313876e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7699
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.1875,
      "epoch": 0.3720525705450329,
      "grad_norm": 0.0,
      "learning_rate": 3.139978739853111e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7700
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.9375,
      "epoch": 0.37210088906068806,
      "grad_norm": 0.0,
      "learning_rate": 3.139737147274836e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7701
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.53125,
      "epoch": 0.37214920757634323,
      "grad_norm": 0.0,
      "learning_rate": 3.1394955546965595e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7702
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.875,
      "epoch": 0.37219752609199847,
      "grad_norm": 0.0,
      "learning_rate": 3.1392539621182836e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7703
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.09375,
      "epoch": 0.37224584460765364,
      "grad_norm": 0.0,
      "learning_rate": 3.139012369540008e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7704
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.71875,
      "epoch": 0.3722941631233089,
      "grad_norm": 0.0,
      "learning_rate": 3.138770776961732e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7705
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.59375,
      "epoch": 0.37234248163896405,
      "grad_norm": 0.0,
      "learning_rate": 3.1385291843834555e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7706
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.125,
      "epoch": 0.37239080015461923,
      "grad_norm": 0.0,
      "learning_rate": 3.1382875918051796e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7707
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.90625,
      "epoch": 0.37243911867027446,
      "grad_norm": 0.0,
      "learning_rate": 3.138045999226904e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7708
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.8125,
      "epoch": 0.37248743718592964,
      "grad_norm": 0.0,
      "learning_rate": 3.1378044066486274e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7709
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.3125,
      "epoch": 0.37253575570158487,
      "grad_norm": 0.0,
      "learning_rate": 3.1375628140703515e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7710
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 68.5625,
      "epoch": 0.37258407421724005,
      "grad_norm": 0.0,
      "learning_rate": 3.1373212214920757e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7711
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.3125,
      "epoch": 0.3726323927328952,
      "grad_norm": 0.0,
      "learning_rate": 3.1370796289138e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7712
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.09375,
      "epoch": 0.37268071124855046,
      "grad_norm": 0.0,
      "learning_rate": 3.1368380363355234e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7713
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.4375,
      "epoch": 0.37272902976420563,
      "grad_norm": 0.0,
      "learning_rate": 3.136596443757248e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7714
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.9375,
      "epoch": 0.37277734827986087,
      "grad_norm": 0.0,
      "learning_rate": 3.1363548511789717e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7715
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.875,
      "epoch": 0.37282566679551604,
      "grad_norm": 0.0,
      "learning_rate": 3.1361132586006953e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7716
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 63.34375,
      "epoch": 0.3728739853111712,
      "grad_norm": 0.0,
      "learning_rate": 3.13587166602242e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 7717
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.21875,
      "epoch": 0.37292230382682645,
      "grad_norm": 0.0,
      "learning_rate": 3.1356300734441436e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7718
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.53125,
      "epoch": 0.37297062234248163,
      "grad_norm": 0.0,
      "learning_rate": 3.1353884808658677e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7719
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.71875,
      "epoch": 0.37301894085813686,
      "grad_norm": 0.0,
      "learning_rate": 3.1351468882875913e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7720
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.0625,
      "epoch": 0.37306725937379204,
      "grad_norm": 0.0,
      "learning_rate": 3.134905295709316e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7721
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.875,
      "epoch": 0.3731155778894472,
      "grad_norm": 0.0,
      "learning_rate": 3.1346637031310396e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7722
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 69.4375,
      "epoch": 0.37316389640510245,
      "grad_norm": 0.0,
      "learning_rate": 3.1344221105527637e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 7723
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.4375,
      "epoch": 0.3732122149207576,
      "grad_norm": 0.0,
      "learning_rate": 3.134180517974488e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7724
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.34375,
      "epoch": 0.37326053343641286,
      "grad_norm": 0.0,
      "learning_rate": 3.133938925396212e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7725
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.5,
      "epoch": 0.37330885195206803,
      "grad_norm": 0.0,
      "learning_rate": 3.1336973328179356e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7726
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.84375,
      "epoch": 0.3733571704677232,
      "grad_norm": 0.0,
      "learning_rate": 3.1334557402396597e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7727
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.375,
      "epoch": 0.37340548898337844,
      "grad_norm": 0.0,
      "learning_rate": 3.133214147661384e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7728
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.5625,
      "epoch": 0.3734538074990336,
      "grad_norm": 0.0,
      "learning_rate": 3.1329725550831075e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7729
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.3125,
      "epoch": 0.37350212601468885,
      "grad_norm": 0.0,
      "learning_rate": 3.1327309625048316e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7730
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.90625,
      "epoch": 0.37355044453034403,
      "grad_norm": 0.0,
      "learning_rate": 3.132489369926556e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7731
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.875,
      "epoch": 0.3735987630459992,
      "grad_norm": 0.0,
      "learning_rate": 3.13224777734828e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7732
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.75,
      "epoch": 0.37364708156165444,
      "grad_norm": 0.0,
      "learning_rate": 3.1320061847700035e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7733
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.1875,
      "epoch": 0.3736954000773096,
      "grad_norm": 0.0,
      "learning_rate": 3.131764592191728e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7734
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.0,
      "epoch": 0.37374371859296485,
      "grad_norm": 0.0,
      "learning_rate": 3.131522999613452e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7735
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.15625,
      "epoch": 0.37379203710862,
      "grad_norm": 0.0,
      "learning_rate": 3.131281407035176e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7736
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.40625,
      "epoch": 0.3738403556242752,
      "grad_norm": 0.0,
      "learning_rate": 3.1310398144569e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7737
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.65625,
      "epoch": 0.37388867413993043,
      "grad_norm": 0.0,
      "learning_rate": 3.1307982218786236e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7738
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 67.9375,
      "epoch": 0.3739369926555856,
      "grad_norm": 0.0,
      "learning_rate": 3.130556629300348e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 7739
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.9375,
      "epoch": 0.37398531117124084,
      "grad_norm": 0.0,
      "learning_rate": 3.1303150367220714e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7740
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.71875,
      "epoch": 0.374033629686896,
      "grad_norm": 0.0,
      "learning_rate": 3.130073444143796e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7741
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.5,
      "epoch": 0.3740819482025512,
      "grad_norm": 0.0,
      "learning_rate": 3.1298318515655196e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7742
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 71.90625,
      "epoch": 0.37413026671820643,
      "grad_norm": 54.89470111536654,
      "learning_rate": 3.129590258987244e-07,
      "loss": -0.0,
      "reward": 1.1875,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.125,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 7743
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.4375,
      "epoch": 0.3741785852338616,
      "grad_norm": 0.0,
      "learning_rate": 3.129348666408968e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7744
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.0625,
      "epoch": 0.37422690374951684,
      "grad_norm": 0.0,
      "learning_rate": 3.129107073830692e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7745
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.46875,
      "epoch": 0.374275222265172,
      "grad_norm": 0.0,
      "learning_rate": 3.1288654812524157e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7746
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.4375,
      "epoch": 0.3743235407808272,
      "grad_norm": 0.0,
      "learning_rate": 3.1286238886741403e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7747
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.71875,
      "epoch": 0.3743718592964824,
      "grad_norm": 0.0,
      "learning_rate": 3.128382296095864e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 7748
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.34375,
      "epoch": 0.3744201778121376,
      "grad_norm": 0.0,
      "learning_rate": 3.1281407035175875e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7749
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.25,
      "epoch": 0.37446849632779283,
      "grad_norm": 0.0,
      "learning_rate": 3.1278991109393117e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7750
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 67.4375,
      "epoch": 0.374516814843448,
      "grad_norm": 0.0,
      "learning_rate": 3.127657518361036e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 7751
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.96875,
      "epoch": 0.3745651333591032,
      "grad_norm": 0.0,
      "learning_rate": 3.12741592578276e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7752
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.15625,
      "epoch": 0.3746134518747584,
      "grad_norm": 0.0,
      "learning_rate": 3.1271743332044836e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7753
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.0625,
      "epoch": 0.3746617703904136,
      "grad_norm": 0.0,
      "learning_rate": 3.126932740626208e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7754
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.28125,
      "epoch": 0.37471008890606883,
      "grad_norm": 0.0,
      "learning_rate": 3.126691148047932e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7755
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.8125,
      "epoch": 0.374758407421724,
      "grad_norm": 0.0,
      "learning_rate": 3.126449555469656e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7756
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.75,
      "epoch": 0.3748067259373792,
      "grad_norm": 0.0,
      "learning_rate": 3.12620796289138e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7757
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.5,
      "epoch": 0.3748550444530344,
      "grad_norm": 0.0,
      "learning_rate": 3.1259663703131037e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7758
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.0625,
      "epoch": 0.3749033629686896,
      "grad_norm": 0.0,
      "learning_rate": 3.125724777734828e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 7759
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.3125,
      "epoch": 0.3749516814843448,
      "grad_norm": 0.0,
      "learning_rate": 3.1254831851565515e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7760
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.96875,
      "epoch": 0.375,
      "grad_norm": 0.0,
      "learning_rate": 3.125241592578276e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7761
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.6875,
      "epoch": 0.3750483185156552,
      "grad_norm": 0.0,
      "learning_rate": 3.1249999999999997e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7762
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.34375,
      "epoch": 0.3750966370313104,
      "grad_norm": 0.0,
      "learning_rate": 3.124758407421724e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7763
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.6875,
      "epoch": 0.3751449555469656,
      "grad_norm": 0.0,
      "learning_rate": 3.124516814843448e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7764
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.125,
      "epoch": 0.3751932740626208,
      "grad_norm": 0.0,
      "learning_rate": 3.124275222265172e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7765
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.75,
      "epoch": 0.375241592578276,
      "grad_norm": 0.0,
      "learning_rate": 3.124033629686896e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7766
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.90625,
      "epoch": 0.37528991109393117,
      "grad_norm": 0.0,
      "learning_rate": 3.1237920371086193e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 7767
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.625,
      "epoch": 0.3753382296095864,
      "grad_norm": 0.0,
      "learning_rate": 3.123550444530344e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7768
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 64.28125,
      "epoch": 0.3753865481252416,
      "grad_norm": 0.0,
      "learning_rate": 3.1233088519520676e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7769
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.21875,
      "epoch": 0.3754348666408968,
      "grad_norm": 0.0,
      "learning_rate": 3.123067259373792e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7770
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.21875,
      "epoch": 0.375483185156552,
      "grad_norm": 0.0,
      "learning_rate": 3.122825666795516e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7771
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.6875,
      "epoch": 0.37553150367220717,
      "grad_norm": 0.0,
      "learning_rate": 3.12258407421724e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7772
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.40625,
      "epoch": 0.3755798221878624,
      "grad_norm": 0.0,
      "learning_rate": 3.1223424816389636e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7773
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.5625,
      "epoch": 0.3756281407035176,
      "grad_norm": 0.0,
      "learning_rate": 3.1221008890606883e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7774
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.53125,
      "epoch": 0.3756764592191728,
      "grad_norm": 0.0,
      "learning_rate": 3.121859296482412e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7775
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.40625,
      "epoch": 0.375724777734828,
      "grad_norm": 0.0,
      "learning_rate": 3.121617703904136e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7776
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 72.0,
      "epoch": 0.37577309625048316,
      "grad_norm": 0.0,
      "learning_rate": 3.1213761113258597e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7777
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.90625,
      "epoch": 0.3758214147661384,
      "grad_norm": 0.0,
      "learning_rate": 3.121134518747584e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7778
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.96875,
      "epoch": 0.37586973328179357,
      "grad_norm": 0.0,
      "learning_rate": 3.120892926169308e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7779
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.65625,
      "epoch": 0.3759180517974488,
      "grad_norm": 0.0,
      "learning_rate": 3.1206513335910315e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7780
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.53125,
      "epoch": 0.375966370313104,
      "grad_norm": 0.0,
      "learning_rate": 3.120409741012756e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7781
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.9375,
      "epoch": 0.37601468882875916,
      "grad_norm": 0.0,
      "learning_rate": 3.12016814843448e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7782
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.625,
      "epoch": 0.3760630073444144,
      "grad_norm": 0.0,
      "learning_rate": 3.119926555856204e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7783
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.90625,
      "epoch": 0.37611132586006957,
      "grad_norm": 0.0,
      "learning_rate": 3.119684963277928e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7784
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.25,
      "epoch": 0.3761596443757248,
      "grad_norm": 0.0,
      "learning_rate": 3.119443370699652e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7785
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.53125,
      "epoch": 0.37620796289138,
      "grad_norm": 0.0,
      "learning_rate": 3.119201778121376e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7786
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.78125,
      "epoch": 0.37625628140703515,
      "grad_norm": 0.0,
      "learning_rate": 3.1189601855431e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7787
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.9375,
      "epoch": 0.3763045999226904,
      "grad_norm": 0.0,
      "learning_rate": 3.118718592964824e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7788
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.8125,
      "epoch": 0.37635291843834556,
      "grad_norm": 0.0,
      "learning_rate": 3.1184770003865477e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7789
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.96875,
      "epoch": 0.3764012369540008,
      "grad_norm": 0.0,
      "learning_rate": 3.118235407808272e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7790
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.96875,
      "epoch": 0.37644955546965597,
      "grad_norm": 0.0,
      "learning_rate": 3.117993815229996e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7791
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.78125,
      "epoch": 0.37649787398531115,
      "grad_norm": 0.0,
      "learning_rate": 3.11775222265172e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7792
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.03125,
      "epoch": 0.3765461925009664,
      "grad_norm": 0.0,
      "learning_rate": 3.1175106300734437e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7793
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.5,
      "epoch": 0.37659451101662156,
      "grad_norm": 0.0,
      "learning_rate": 3.1172690374951684e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7794
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.8125,
      "epoch": 0.3766428295322768,
      "grad_norm": 0.0,
      "learning_rate": 3.117027444916892e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7795
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.25,
      "epoch": 0.37669114804793197,
      "grad_norm": 0.0,
      "learning_rate": 3.116785852338616e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7796
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 73.78125,
      "epoch": 0.37673946656358714,
      "grad_norm": 22.962554986157162,
      "learning_rate": 3.1165442597603397e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7797
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.125,
      "epoch": 0.3767877850792424,
      "grad_norm": 0.0,
      "learning_rate": 3.1163026671820644e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7798
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.09375,
      "epoch": 0.37683610359489755,
      "grad_norm": 0.0,
      "learning_rate": 3.116061074603788e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7799
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.71875,
      "epoch": 0.3768844221105528,
      "grad_norm": 0.0,
      "learning_rate": 3.1158194820255116e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7800
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.8125,
      "epoch": 0.37693274062620796,
      "grad_norm": 0.0,
      "learning_rate": 3.1155778894472363e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7801
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.65625,
      "epoch": 0.37698105914186314,
      "grad_norm": 0.0,
      "learning_rate": 3.11533629686896e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7802
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.40625,
      "epoch": 0.37702937765751837,
      "grad_norm": 0.0,
      "learning_rate": 3.115094704290684e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7803
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.125,
      "epoch": 0.37707769617317355,
      "grad_norm": 0.0,
      "learning_rate": 3.114853111712408e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7804
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.59375,
      "epoch": 0.3771260146888288,
      "grad_norm": 0.0,
      "learning_rate": 3.1146115191341323e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7805
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.21875,
      "epoch": 0.37717433320448396,
      "grad_norm": 0.0,
      "learning_rate": 3.114369926555856e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7806
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.5,
      "epoch": 0.37722265172013913,
      "grad_norm": 0.0,
      "learning_rate": 3.11412833397758e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7807
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.40625,
      "epoch": 0.37727097023579437,
      "grad_norm": 0.0,
      "learning_rate": 3.113886741399304e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7808
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.8125,
      "epoch": 0.37731928875144954,
      "grad_norm": 0.0,
      "learning_rate": 3.113645148821028e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7809
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.25,
      "epoch": 0.3773676072671048,
      "grad_norm": 0.0,
      "learning_rate": 3.113403556242752e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 7810
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.84375,
      "epoch": 0.37741592578275995,
      "grad_norm": 0.0,
      "learning_rate": 3.113161963664476e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7811
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.03125,
      "epoch": 0.37746424429841513,
      "grad_norm": 0.0,
      "learning_rate": 3.1129203710862e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7812
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.8125,
      "epoch": 0.37751256281407036,
      "grad_norm": 0.0,
      "learning_rate": 3.112678778507924e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7813
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.78125,
      "epoch": 0.37756088132972554,
      "grad_norm": 0.0,
      "learning_rate": 3.1124371859296485e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7814
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.1875,
      "epoch": 0.37760919984538077,
      "grad_norm": 0.0,
      "learning_rate": 3.112195593351372e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7815
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.5625,
      "epoch": 0.37765751836103595,
      "grad_norm": 0.0,
      "learning_rate": 3.111954000773096e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7816
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.6875,
      "epoch": 0.3777058368766911,
      "grad_norm": 0.0,
      "learning_rate": 3.11171240819482e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7817
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.34375,
      "epoch": 0.37775415539234636,
      "grad_norm": 0.0,
      "learning_rate": 3.1114708156165445e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7818
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.09375,
      "epoch": 0.37780247390800153,
      "grad_norm": 0.0,
      "learning_rate": 3.111229223038268e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7819
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.09375,
      "epoch": 0.37785079242365677,
      "grad_norm": 0.0,
      "learning_rate": 3.1109876304599917e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7820
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.9375,
      "epoch": 0.37789911093931194,
      "grad_norm": 0.0,
      "learning_rate": 3.1107460378817164e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7821
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.0625,
      "epoch": 0.3779474294549671,
      "grad_norm": 0.0,
      "learning_rate": 3.11050444530344e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7822
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.21875,
      "epoch": 0.37799574797062235,
      "grad_norm": 0.0,
      "learning_rate": 3.110262852725164e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 7823
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.96875,
      "epoch": 0.37804406648627753,
      "grad_norm": 0.0,
      "learning_rate": 3.110021260146888e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7824
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.1875,
      "epoch": 0.37809238500193276,
      "grad_norm": 0.0,
      "learning_rate": 3.1097796675686124e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 7825
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.40625,
      "epoch": 0.37814070351758794,
      "grad_norm": 0.0,
      "learning_rate": 3.109538074990336e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7826
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.65625,
      "epoch": 0.3781890220332431,
      "grad_norm": 0.0,
      "learning_rate": 3.10929648241206e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7827
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.03125,
      "epoch": 0.37823734054889835,
      "grad_norm": 0.0,
      "learning_rate": 3.109054889833784e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7828
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.875,
      "epoch": 0.3782856590645535,
      "grad_norm": 0.0,
      "learning_rate": 3.108813297255508e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7829
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.34375,
      "epoch": 0.37833397758020876,
      "grad_norm": 0.0,
      "learning_rate": 3.108571704677232e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 7830
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.21875,
      "epoch": 0.37838229609586393,
      "grad_norm": 0.0,
      "learning_rate": 3.108330112098956e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7831
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.5,
      "epoch": 0.3784306146115191,
      "grad_norm": 0.0,
      "learning_rate": 3.1080885195206803e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7832
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.25,
      "epoch": 0.37847893312717434,
      "grad_norm": 0.0,
      "learning_rate": 3.107846926942404e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7833
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.15625,
      "epoch": 0.3785272516428295,
      "grad_norm": 0.0,
      "learning_rate": 3.1076053343641285e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7834
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.4375,
      "epoch": 0.37857557015848475,
      "grad_norm": 0.0,
      "learning_rate": 3.107363741785852e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7835
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.3125,
      "epoch": 0.37862388867413993,
      "grad_norm": 0.0,
      "learning_rate": 3.1071221492075763e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7836
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.0625,
      "epoch": 0.3786722071897951,
      "grad_norm": 0.0,
      "learning_rate": 3.1068805566293e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7837
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.09375,
      "epoch": 0.37872052570545034,
      "grad_norm": 0.0,
      "learning_rate": 3.1066389640510246e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7838
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.21875,
      "epoch": 0.3787688442211055,
      "grad_norm": 0.0,
      "learning_rate": 3.106397371472748e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7839
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.5625,
      "epoch": 0.37881716273676075,
      "grad_norm": 0.0,
      "learning_rate": 3.106155778894472e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7840
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.125,
      "epoch": 0.3788654812524159,
      "grad_norm": 0.0,
      "learning_rate": 3.1059141863161964e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7841
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.125,
      "epoch": 0.3789137997680711,
      "grad_norm": 0.0,
      "learning_rate": 3.10567259373792e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 0.96875,
      "step": 7842
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.21875,
      "epoch": 0.37896211828372633,
      "grad_norm": 0.0,
      "learning_rate": 3.105431001159644e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7843
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.125,
      "epoch": 0.3790104367993815,
      "grad_norm": 0.0,
      "learning_rate": 3.1051894085813683e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7844
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.6875,
      "epoch": 0.37905875531503674,
      "grad_norm": 0.0,
      "learning_rate": 3.1049478160030924e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2630178928375244,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7845
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.21875,
      "epoch": 0.3791070738306919,
      "grad_norm": 0.0,
      "learning_rate": 3.104706223424816e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7846
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.25,
      "epoch": 0.3791553923463471,
      "grad_norm": 0.0,
      "learning_rate": 3.10446463084654e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7847
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.6875,
      "epoch": 0.37920371086200233,
      "grad_norm": 0.0,
      "learning_rate": 3.1042230382682643e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7848
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.5625,
      "epoch": 0.3792520293776575,
      "grad_norm": 0.0,
      "learning_rate": 3.1039814456899885e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7849
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.03125,
      "epoch": 0.37930034789331274,
      "grad_norm": 0.0,
      "learning_rate": 3.103739853111712e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7850
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.375,
      "epoch": 0.3793486664089679,
      "grad_norm": 53.65999528159984,
      "learning_rate": 3.103498260533436e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.3676722049713135,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7851
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.3125,
      "epoch": 0.3793969849246231,
      "grad_norm": 0.0,
      "learning_rate": 3.1032566679551603e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7852
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.53125,
      "epoch": 0.3794453034402783,
      "grad_norm": 0.0,
      "learning_rate": 3.103015075376884e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 7853
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.34375,
      "epoch": 0.3794936219559335,
      "grad_norm": 0.0,
      "learning_rate": 3.1027734827986086e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7854
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 72.0,
      "epoch": 0.37954194047158873,
      "grad_norm": 26.14810045272694,
      "learning_rate": 3.102531890220332e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7855
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 71.375,
      "epoch": 0.3795902589872439,
      "grad_norm": 140.17328789344612,
      "learning_rate": 3.1022902976420564e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.3234682083129883,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7856
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.5,
      "epoch": 0.3796385775028991,
      "grad_norm": 0.0,
      "learning_rate": 3.10204870506378e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7857
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.375,
      "epoch": 0.3796868960185543,
      "grad_norm": 0.0,
      "learning_rate": 3.1018071124855046e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7858
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.625,
      "epoch": 0.3797352145342095,
      "grad_norm": 0.0,
      "learning_rate": 3.101565519907228e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7859
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.5,
      "epoch": 0.37978353304986473,
      "grad_norm": 0.0,
      "learning_rate": 3.101323927328952e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7860
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.75,
      "epoch": 0.3798318515655199,
      "grad_norm": 0.0,
      "learning_rate": 3.1010823347506765e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7861
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.34375,
      "epoch": 0.3798801700811751,
      "grad_norm": 0.0,
      "learning_rate": 3.1008407421724e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7862
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.09375,
      "epoch": 0.3799284885968303,
      "grad_norm": 0.0,
      "learning_rate": 3.100599149594124e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7863
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.34375,
      "epoch": 0.3799768071124855,
      "grad_norm": 0.0,
      "learning_rate": 3.1003575570158484e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7864
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.71875,
      "epoch": 0.3800251256281407,
      "grad_norm": 0.0,
      "learning_rate": 3.1001159644375725e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7865
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.1875,
      "epoch": 0.3800734441437959,
      "grad_norm": 0.0,
      "learning_rate": 3.099874371859296e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7866
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.0625,
      "epoch": 0.3801217626594511,
      "grad_norm": 0.0,
      "learning_rate": 3.0996327792810203e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7867
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.625,
      "epoch": 0.3801700811751063,
      "grad_norm": 0.0,
      "learning_rate": 3.0993911867027444e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7868
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.84375,
      "epoch": 0.3802183996907615,
      "grad_norm": 0.0,
      "learning_rate": 3.0991495941244685e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7869
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.0625,
      "epoch": 0.3802667182064167,
      "grad_norm": 0.0,
      "learning_rate": 3.098908001546192e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7870
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 74.40625,
      "epoch": 0.3803150367220719,
      "grad_norm": 22.783135373838103,
      "learning_rate": 3.0986664089679163e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7871
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.71875,
      "epoch": 0.3803633552377271,
      "grad_norm": 0.0,
      "learning_rate": 3.0984248163896404e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7872
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 81.65625,
      "epoch": 0.3804116737533823,
      "grad_norm": 33.12134427571396,
      "learning_rate": 3.098183223811364e-07,
      "loss": -0.0,
      "reward": 1.484375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 7873
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.90625,
      "epoch": 0.3804599922690375,
      "grad_norm": 0.0,
      "learning_rate": 3.0979416312330887e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7874
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.84375,
      "epoch": 0.3805083107846927,
      "grad_norm": 0.0,
      "learning_rate": 3.0977000386548123e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7875
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.15625,
      "epoch": 0.3805566293003479,
      "grad_norm": 0.0,
      "learning_rate": 3.0974584460765364e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7876
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.15625,
      "epoch": 0.38060494781600307,
      "grad_norm": 0.0,
      "learning_rate": 3.09721685349826e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7877
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.75,
      "epoch": 0.3806532663316583,
      "grad_norm": 0.0,
      "learning_rate": 3.0969752609199847e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7878
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.25,
      "epoch": 0.3807015848473135,
      "grad_norm": 0.0,
      "learning_rate": 3.0967336683417083e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7879
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.46875,
      "epoch": 0.3807499033629687,
      "grad_norm": 0.0,
      "learning_rate": 3.096492075763432e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7880
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.1875,
      "epoch": 0.3807982218786239,
      "grad_norm": 0.0,
      "learning_rate": 3.0962504831851566e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 7881
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.875,
      "epoch": 0.38084654039427907,
      "grad_norm": 18.70817429151359,
      "learning_rate": 3.09600889060688e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7882
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.46875,
      "epoch": 0.3808948589099343,
      "grad_norm": 0.0,
      "learning_rate": 3.0957672980286043e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7883
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.59375,
      "epoch": 0.3809431774255895,
      "grad_norm": 0.0,
      "learning_rate": 3.0955257054503285e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7884
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.0,
      "epoch": 0.3809914959412447,
      "grad_norm": 0.0,
      "learning_rate": 3.0952841128720526e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 7885
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.09375,
      "epoch": 0.3810398144568999,
      "grad_norm": 0.0,
      "learning_rate": 3.095042520293776e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7886
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.65625,
      "epoch": 0.38108813297255506,
      "grad_norm": 0.0,
      "learning_rate": 3.0948009277155004e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7887
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.25,
      "epoch": 0.3811364514882103,
      "grad_norm": 0.0,
      "learning_rate": 3.0945593351372245e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7888
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.8125,
      "epoch": 0.38118477000386547,
      "grad_norm": 0.0,
      "learning_rate": 3.0943177425589486e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 7889
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.78125,
      "epoch": 0.3812330885195207,
      "grad_norm": 0.0,
      "learning_rate": 3.094076149980672e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7890
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.96875,
      "epoch": 0.3812814070351759,
      "grad_norm": 32.139030144652345,
      "learning_rate": 3.0938345574023964e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7891
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.6875,
      "epoch": 0.38132972555083106,
      "grad_norm": 0.0,
      "learning_rate": 3.0935929648241205e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7892
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.1875,
      "epoch": 0.3813780440664863,
      "grad_norm": 0.0,
      "learning_rate": 3.093351372245844e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7893
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.4375,
      "epoch": 0.38142636258214146,
      "grad_norm": 0.0,
      "learning_rate": 3.093109779667569e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7894
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.375,
      "epoch": 0.3814746810977967,
      "grad_norm": 0.0,
      "learning_rate": 3.0928681870892924e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7895
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.59375,
      "epoch": 0.3815229996134519,
      "grad_norm": 0.0,
      "learning_rate": 3.0926265945110165e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7896
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.5625,
      "epoch": 0.38157131812910705,
      "grad_norm": 0.0,
      "learning_rate": 3.09238500193274e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7897
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.15625,
      "epoch": 0.3816196366447623,
      "grad_norm": 0.0,
      "learning_rate": 3.092143409354465e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7898
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.15625,
      "epoch": 0.38166795516041746,
      "grad_norm": 0.0,
      "learning_rate": 3.0919018167761884e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7899
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.25,
      "epoch": 0.3817162736760727,
      "grad_norm": 0.0,
      "learning_rate": 3.0916602241979125e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7900
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.9375,
      "epoch": 0.38176459219172787,
      "grad_norm": 0.0,
      "learning_rate": 3.0914186316196367e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7901
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.90625,
      "epoch": 0.38181291070738305,
      "grad_norm": 0.0,
      "learning_rate": 3.0911770390413603e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7902
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.96875,
      "epoch": 0.3818612292230383,
      "grad_norm": 0.0,
      "learning_rate": 3.0909354464630844e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7903
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.25,
      "epoch": 0.38190954773869346,
      "grad_norm": 0.0,
      "learning_rate": 3.0906938538848085e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7904
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.4375,
      "epoch": 0.3819578662543487,
      "grad_norm": 0.0,
      "learning_rate": 3.0904522613065327e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7905
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.75,
      "epoch": 0.38200618477000386,
      "grad_norm": 0.0,
      "learning_rate": 3.0902106687282563e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7906
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.84375,
      "epoch": 0.38205450328565904,
      "grad_norm": 0.0,
      "learning_rate": 3.0899690761499804e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7907
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.15625,
      "epoch": 0.3821028218013143,
      "grad_norm": 0.0,
      "learning_rate": 3.0897274835717046e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7908
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.25,
      "epoch": 0.38215114031696945,
      "grad_norm": 13.891970336223613,
      "learning_rate": 3.0894858909934287e-07,
      "loss": -0.0,
      "reward": 2.140625,
      "reward_std": 0.3629639744758606,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7909
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.15625,
      "epoch": 0.3821994588326247,
      "grad_norm": 0.0,
      "learning_rate": 3.0892442984151523e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7910
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.0625,
      "epoch": 0.38224777734827986,
      "grad_norm": 0.0,
      "learning_rate": 3.089002705836877e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7911
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.78125,
      "epoch": 0.38229609586393504,
      "grad_norm": 0.0,
      "learning_rate": 3.0887611132586006e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7912
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.875,
      "epoch": 0.38234441437959027,
      "grad_norm": 0.0,
      "learning_rate": 3.088519520680324e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 7913
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.5625,
      "epoch": 0.38239273289524545,
      "grad_norm": 0.0,
      "learning_rate": 3.088277928102049e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7914
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.03125,
      "epoch": 0.3824410514109007,
      "grad_norm": 0.0,
      "learning_rate": 3.0880363355237725e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7915
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.53125,
      "epoch": 0.38248936992655586,
      "grad_norm": 0.0,
      "learning_rate": 3.0877947429454966e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7916
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.75,
      "epoch": 0.38253768844221103,
      "grad_norm": 0.0,
      "learning_rate": 3.08755315036722e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7917
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.15625,
      "epoch": 0.38258600695786626,
      "grad_norm": 0.0,
      "learning_rate": 3.087311557788945e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7918
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.0,
      "epoch": 0.38263432547352144,
      "grad_norm": 0.0,
      "learning_rate": 3.0870699652106685e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 7919
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.78125,
      "epoch": 0.3826826439891767,
      "grad_norm": 0.0,
      "learning_rate": 3.0868283726323926e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7920
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 97.5625,
      "epoch": 0.38273096250483185,
      "grad_norm": 0.0,
      "learning_rate": 3.086586780054117e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 7921
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.5,
      "epoch": 0.382779281020487,
      "grad_norm": 0.0,
      "learning_rate": 3.0863451874758404e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2130674123764038,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7922
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.5,
      "epoch": 0.38282759953614226,
      "grad_norm": 0.0,
      "learning_rate": 3.0861035948975645e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 7923
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.8125,
      "epoch": 0.38287591805179744,
      "grad_norm": 0.0,
      "learning_rate": 3.0858620023192886e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7924
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.84375,
      "epoch": 0.38292423656745267,
      "grad_norm": 0.0,
      "learning_rate": 3.085620409741013e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7925
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.3125,
      "epoch": 0.38297255508310785,
      "grad_norm": 0.0,
      "learning_rate": 3.0853788171627364e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7926
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.90625,
      "epoch": 0.383020873598763,
      "grad_norm": 0.0,
      "learning_rate": 3.0851372245844605e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7927
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.96875,
      "epoch": 0.38306919211441826,
      "grad_norm": 0.0,
      "learning_rate": 3.0848956320061846e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7928
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.21875,
      "epoch": 0.38311751063007343,
      "grad_norm": 0.0,
      "learning_rate": 3.084654039427909e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 7929
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.65625,
      "epoch": 0.38316582914572866,
      "grad_norm": 0.0,
      "learning_rate": 3.0844124468496324e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7930
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.75,
      "epoch": 0.38321414766138384,
      "grad_norm": 0.0,
      "learning_rate": 3.084170854271357e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7931
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.75,
      "epoch": 0.383262466177039,
      "grad_norm": 0.0,
      "learning_rate": 3.0839292616930807e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7932
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.625,
      "epoch": 0.38331078469269425,
      "grad_norm": 0.0,
      "learning_rate": 3.0836876691148043e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7933
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.21875,
      "epoch": 0.3833591032083494,
      "grad_norm": 0.0,
      "learning_rate": 3.083446076536529e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7934
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.96875,
      "epoch": 0.38340742172400466,
      "grad_norm": 0.0,
      "learning_rate": 3.0832044839582525e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7935
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.28125,
      "epoch": 0.38345574023965984,
      "grad_norm": 0.0,
      "learning_rate": 3.0829628913799767e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7936
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.5625,
      "epoch": 0.383504058755315,
      "grad_norm": 0.0,
      "learning_rate": 3.0827212988017003e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7937
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.40625,
      "epoch": 0.38355237727097025,
      "grad_norm": 0.0,
      "learning_rate": 3.082479706223425e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7938
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.5625,
      "epoch": 0.3836006957866254,
      "grad_norm": 0.0,
      "learning_rate": 3.0822381136451486e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7939
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.9375,
      "epoch": 0.38364901430228066,
      "grad_norm": 0.0,
      "learning_rate": 3.0819965210668727e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7940
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.03125,
      "epoch": 0.38369733281793583,
      "grad_norm": 0.0,
      "learning_rate": 3.081754928488597e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7941
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.21875,
      "epoch": 0.383745651333591,
      "grad_norm": 0.0,
      "learning_rate": 3.081513335910321e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 7942
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.125,
      "epoch": 0.38379396984924624,
      "grad_norm": 0.0,
      "learning_rate": 3.0812717433320446e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 7943
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.90625,
      "epoch": 0.3838422883649014,
      "grad_norm": 0.0,
      "learning_rate": 3.0810301507537687e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7944
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.15625,
      "epoch": 0.38389060688055665,
      "grad_norm": 0.0,
      "learning_rate": 3.080788558175493e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7945
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.9375,
      "epoch": 0.3839389253962118,
      "grad_norm": 0.0,
      "learning_rate": 3.0805469655972164e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7946
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.84375,
      "epoch": 0.383987243911867,
      "grad_norm": 0.0,
      "learning_rate": 3.0803053730189406e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7947
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.78125,
      "epoch": 0.38403556242752224,
      "grad_norm": 0.0,
      "learning_rate": 3.0800637804406647e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7948
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.28125,
      "epoch": 0.3840838809431774,
      "grad_norm": 0.0,
      "learning_rate": 3.079822187862389e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7949
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.875,
      "epoch": 0.38413219945883265,
      "grad_norm": 0.0,
      "learning_rate": 3.0795805952841125e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7950
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.28125,
      "epoch": 0.3841805179744878,
      "grad_norm": 0.0,
      "learning_rate": 3.079339002705837e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7951
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.3125,
      "epoch": 0.384228836490143,
      "grad_norm": 0.0,
      "learning_rate": 3.079097410127561e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7952
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.15625,
      "epoch": 0.38427715500579823,
      "grad_norm": 0.0,
      "learning_rate": 3.0788558175492843e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7953
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.6875,
      "epoch": 0.3843254735214534,
      "grad_norm": 0.0,
      "learning_rate": 3.078614224971009e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7954
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.96875,
      "epoch": 0.38437379203710864,
      "grad_norm": 0.0,
      "learning_rate": 3.0783726323927326e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7955
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.125,
      "epoch": 0.3844221105527638,
      "grad_norm": 0.0,
      "learning_rate": 3.078131039814457e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7956
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.3125,
      "epoch": 0.384470429068419,
      "grad_norm": 0.0,
      "learning_rate": 3.0778894472361804e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7957
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.0625,
      "epoch": 0.3845187475840742,
      "grad_norm": 0.0,
      "learning_rate": 3.077647854657905e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 7958
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.40625,
      "epoch": 0.3845670660997294,
      "grad_norm": 0.0,
      "learning_rate": 3.0774062620796286e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 7959
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 85.625,
      "epoch": 0.38461538461538464,
      "grad_norm": 0.0,
      "learning_rate": 3.077164669501353e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 7960
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.375,
      "epoch": 0.3846637031310398,
      "grad_norm": 0.0,
      "learning_rate": 3.076923076923077e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7961
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.625,
      "epoch": 0.384712021646695,
      "grad_norm": 0.0,
      "learning_rate": 3.076681484344801e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7962
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.21875,
      "epoch": 0.3847603401623502,
      "grad_norm": 0.0,
      "learning_rate": 3.0764398917665246e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7963
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.875,
      "epoch": 0.3848086586780054,
      "grad_norm": 0.0,
      "learning_rate": 3.076198299188249e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7964
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.5,
      "epoch": 0.38485697719366063,
      "grad_norm": 0.0,
      "learning_rate": 3.075956706609973e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7965
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.875,
      "epoch": 0.3849052957093158,
      "grad_norm": 0.0,
      "learning_rate": 3.0757151140316965e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7966
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.78125,
      "epoch": 0.384953614224971,
      "grad_norm": 0.0,
      "learning_rate": 3.0754735214534207e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7967
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.65625,
      "epoch": 0.3850019327406262,
      "grad_norm": 0.0,
      "learning_rate": 3.075231928875145e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 7968
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.875,
      "epoch": 0.3850502512562814,
      "grad_norm": 0.0,
      "learning_rate": 3.074990336296869e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 7969
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.15625,
      "epoch": 0.3850985697719366,
      "grad_norm": 0.0,
      "learning_rate": 3.0747487437185925e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7970
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.90625,
      "epoch": 0.3851468882875918,
      "grad_norm": 0.0,
      "learning_rate": 3.074507151140317e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7971
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.875,
      "epoch": 0.385195206803247,
      "grad_norm": 0.0,
      "learning_rate": 3.074265558562041e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7972
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.96875,
      "epoch": 0.3852435253189022,
      "grad_norm": 0.0,
      "learning_rate": 3.0740239659837644e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7973
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.59375,
      "epoch": 0.3852918438345574,
      "grad_norm": 0.0,
      "learning_rate": 3.073782373405489e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7974
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.78125,
      "epoch": 0.3853401623502126,
      "grad_norm": 0.0,
      "learning_rate": 3.0735407808272127e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7975
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.90625,
      "epoch": 0.3853884808658678,
      "grad_norm": 0.0,
      "learning_rate": 3.073299188248937e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7976
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.34375,
      "epoch": 0.385436799381523,
      "grad_norm": 0.0,
      "learning_rate": 3.0730575956706604e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 7977
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 76.84375,
      "epoch": 0.3854851178971782,
      "grad_norm": 13.741299698430463,
      "learning_rate": 3.072816003092385e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 7978
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.3125,
      "epoch": 0.3855334364128334,
      "grad_norm": 0.0,
      "learning_rate": 3.0725744105141087e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7979
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.3125,
      "epoch": 0.3855817549284886,
      "grad_norm": 0.0,
      "learning_rate": 3.072332817935833e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7980
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.78125,
      "epoch": 0.3856300734441438,
      "grad_norm": 0.0,
      "learning_rate": 3.072091225357557e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7981
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.5,
      "epoch": 0.38567839195979897,
      "grad_norm": 0.0,
      "learning_rate": 3.071849632779281e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 7982
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 81.28125,
      "epoch": 0.3857267104754542,
      "grad_norm": 0.0,
      "learning_rate": 3.0716080402010047e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 7983
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.53125,
      "epoch": 0.3857750289911094,
      "grad_norm": 0.0,
      "learning_rate": 3.071366447622729e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 7984
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.21875,
      "epoch": 0.3858233475067646,
      "grad_norm": 0.0,
      "learning_rate": 3.071124855044453e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 7985
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.09375,
      "epoch": 0.3858716660224198,
      "grad_norm": 0.0,
      "learning_rate": 3.0708832624661766e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 7986
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.46875,
      "epoch": 0.38591998453807497,
      "grad_norm": 0.0,
      "learning_rate": 3.070641669887901e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 7987
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.25,
      "epoch": 0.3859683030537302,
      "grad_norm": 0.0,
      "learning_rate": 3.070400077309625e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 7988
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.96875,
      "epoch": 0.3860166215693854,
      "grad_norm": 0.0,
      "learning_rate": 3.070158484731349e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7989
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.46875,
      "epoch": 0.3860649400850406,
      "grad_norm": 0.0,
      "learning_rate": 3.0699168921530726e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 7990
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.75,
      "epoch": 0.3861132586006958,
      "grad_norm": 0.0,
      "learning_rate": 3.0696752995747973e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 7991
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.46875,
      "epoch": 0.38616157711635096,
      "grad_norm": 0.0,
      "learning_rate": 3.069433706996521e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7992
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.71875,
      "epoch": 0.3862098956320062,
      "grad_norm": 0.0,
      "learning_rate": 3.069192114418245e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 7993
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.375,
      "epoch": 0.38625821414766137,
      "grad_norm": 0.0,
      "learning_rate": 3.068950521839969e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 7994
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.15625,
      "epoch": 0.3863065326633166,
      "grad_norm": 0.0,
      "learning_rate": 3.068708929261693e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 7995
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.09375,
      "epoch": 0.3863548511789718,
      "grad_norm": 0.0,
      "learning_rate": 3.068467336683417e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7996
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.21875,
      "epoch": 0.38640316969462696,
      "grad_norm": 0.0,
      "learning_rate": 3.0682257441051405e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 7997
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.53125,
      "epoch": 0.3864514882102822,
      "grad_norm": 0.0,
      "learning_rate": 3.067984151526865e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 7998
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.9375,
      "epoch": 0.38649980672593737,
      "grad_norm": 0.0,
      "learning_rate": 3.067742558948589e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 7999
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.40625,
      "epoch": 0.3865481252415926,
      "grad_norm": 0.0,
      "learning_rate": 3.067500966370313e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8000
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.625,
      "epoch": 0.3865964437572478,
      "grad_norm": 0.0,
      "learning_rate": 3.067259373792037e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8001
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.40625,
      "epoch": 0.38664476227290295,
      "grad_norm": 0.0,
      "learning_rate": 3.067017781213761e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8002
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.8125,
      "epoch": 0.3866930807885582,
      "grad_norm": 0.0,
      "learning_rate": 3.066776188635485e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 8003
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.59375,
      "epoch": 0.38674139930421336,
      "grad_norm": 0.0,
      "learning_rate": 3.0665345960572095e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8004
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.03125,
      "epoch": 0.3867897178198686,
      "grad_norm": 0.0,
      "learning_rate": 3.066293003478933e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8005
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.3125,
      "epoch": 0.38683803633552377,
      "grad_norm": 0.0,
      "learning_rate": 3.0660514109006567e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8006
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.3125,
      "epoch": 0.38688635485117895,
      "grad_norm": 0.0,
      "learning_rate": 3.065809818322381e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8007
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.8125,
      "epoch": 0.3869346733668342,
      "grad_norm": 0.0,
      "learning_rate": 3.065568225744105e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8008
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.25,
      "epoch": 0.38698299188248936,
      "grad_norm": 0.0,
      "learning_rate": 3.065326633165829e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8009
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.03125,
      "epoch": 0.3870313103981446,
      "grad_norm": 0.0,
      "learning_rate": 3.0650850405875527e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 8010
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.5625,
      "epoch": 0.38707962891379977,
      "grad_norm": 0.0,
      "learning_rate": 3.0648434480092774e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8011
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.28125,
      "epoch": 0.38712794742945494,
      "grad_norm": 11.0823996360274,
      "learning_rate": 3.064601855431001e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8012
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.09375,
      "epoch": 0.3871762659451102,
      "grad_norm": 0.0,
      "learning_rate": 3.064360262852725e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8013
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.0,
      "epoch": 0.38722458446076535,
      "grad_norm": 0.0,
      "learning_rate": 3.064118670274449e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8014
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.0,
      "epoch": 0.3872729029764206,
      "grad_norm": 0.0,
      "learning_rate": 3.063877077696173e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8015
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.0625,
      "epoch": 0.38732122149207576,
      "grad_norm": 0.0,
      "learning_rate": 3.063635485117897e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8016
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.40625,
      "epoch": 0.38736954000773094,
      "grad_norm": 0.0,
      "learning_rate": 3.0633938925396206e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8017
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.78125,
      "epoch": 0.38741785852338617,
      "grad_norm": 0.0,
      "learning_rate": 3.063152299961345e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8018
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.28125,
      "epoch": 0.38746617703904135,
      "grad_norm": 0.0,
      "learning_rate": 3.062910707383069e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8019
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.375,
      "epoch": 0.3875144955546966,
      "grad_norm": 0.0,
      "learning_rate": 3.062669114804793e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8020
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.8125,
      "epoch": 0.38756281407035176,
      "grad_norm": 0.0,
      "learning_rate": 3.062427522226517e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8021
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.375,
      "epoch": 0.38761113258600693,
      "grad_norm": 0.0,
      "learning_rate": 3.0621859296482413e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8022
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.46875,
      "epoch": 0.38765945110166217,
      "grad_norm": 0.0,
      "learning_rate": 3.061944337069965e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8023
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.71875,
      "epoch": 0.38770776961731734,
      "grad_norm": 0.0,
      "learning_rate": 3.0617027444916896e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8024
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.625,
      "epoch": 0.3877560881329726,
      "grad_norm": 0.0,
      "learning_rate": 3.061461151913413e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8025
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.5,
      "epoch": 0.38780440664862775,
      "grad_norm": 0.0,
      "learning_rate": 3.061219559335137e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8026
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.46875,
      "epoch": 0.38785272516428293,
      "grad_norm": 0.0,
      "learning_rate": 3.060977966756861e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 8027
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.8125,
      "epoch": 0.38790104367993816,
      "grad_norm": 0.0,
      "learning_rate": 3.060736374178585e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8028
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.59375,
      "epoch": 0.38794936219559334,
      "grad_norm": 0.0,
      "learning_rate": 3.060494781600309e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8029
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.6875,
      "epoch": 0.38799768071124857,
      "grad_norm": 0.0,
      "learning_rate": 3.060253189022033e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8030
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.53125,
      "epoch": 0.38804599922690375,
      "grad_norm": 0.0,
      "learning_rate": 3.0600115964437574e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8031
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.125,
      "epoch": 0.3880943177425589,
      "grad_norm": 0.0,
      "learning_rate": 3.059770003865481e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8032
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.78125,
      "epoch": 0.38814263625821416,
      "grad_norm": 0.0,
      "learning_rate": 3.059528411287205e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8033
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.09375,
      "epoch": 0.38819095477386933,
      "grad_norm": 0.0,
      "learning_rate": 3.0592868187089293e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8034
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.71875,
      "epoch": 0.38823927328952457,
      "grad_norm": 0.0,
      "learning_rate": 3.059045226130653e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8035
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.53125,
      "epoch": 0.38828759180517974,
      "grad_norm": 0.0,
      "learning_rate": 3.058803633552377e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8036
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.875,
      "epoch": 0.3883359103208349,
      "grad_norm": 0.0,
      "learning_rate": 3.0585620409741007e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8037
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.125,
      "epoch": 0.38838422883649015,
      "grad_norm": 0.0,
      "learning_rate": 3.0583204483958253e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8038
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.21875,
      "epoch": 0.38843254735214533,
      "grad_norm": 0.0,
      "learning_rate": 3.058078855817549e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8039
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.3125,
      "epoch": 0.38848086586780056,
      "grad_norm": 0.0,
      "learning_rate": 3.057837263239273e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8040
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.8125,
      "epoch": 0.38852918438345574,
      "grad_norm": 0.0,
      "learning_rate": 3.057595670660997e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 8041
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.90625,
      "epoch": 0.3885775028991109,
      "grad_norm": 0.0,
      "learning_rate": 3.0573540780827214e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8042
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.09375,
      "epoch": 0.38862582141476615,
      "grad_norm": 0.0,
      "learning_rate": 3.057112485504445e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8043
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.5,
      "epoch": 0.3886741399304213,
      "grad_norm": 0.0,
      "learning_rate": 3.0568708929261696e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 8044
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.96875,
      "epoch": 0.38872245844607656,
      "grad_norm": 0.0,
      "learning_rate": 3.056629300347893e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8045
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.03125,
      "epoch": 0.38877077696173173,
      "grad_norm": 0.0,
      "learning_rate": 3.056387707769617e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8046
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.90625,
      "epoch": 0.3888190954773869,
      "grad_norm": 0.0,
      "learning_rate": 3.056146115191341e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8047
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.09375,
      "epoch": 0.38886741399304214,
      "grad_norm": 0.0,
      "learning_rate": 3.055904522613065e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8048
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.25,
      "epoch": 0.3889157325086973,
      "grad_norm": 0.0,
      "learning_rate": 3.055662930034789e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8049
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.875,
      "epoch": 0.38896405102435255,
      "grad_norm": 0.0,
      "learning_rate": 3.055421337456513e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8050
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.34375,
      "epoch": 0.38901236954000773,
      "grad_norm": 0.0,
      "learning_rate": 3.0551797448782375e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8051
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.3125,
      "epoch": 0.3890606880556629,
      "grad_norm": 0.0,
      "learning_rate": 3.054938152299961e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8052
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.5625,
      "epoch": 0.38910900657131814,
      "grad_norm": 0.0,
      "learning_rate": 3.0546965597216853e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8053
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.03125,
      "epoch": 0.3891573250869733,
      "grad_norm": 0.0,
      "learning_rate": 3.0544549671434094e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8054
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.4375,
      "epoch": 0.38920564360262855,
      "grad_norm": 0.0,
      "learning_rate": 3.0542133745651335e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8055
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.0,
      "epoch": 0.3892539621182837,
      "grad_norm": 0.0,
      "learning_rate": 3.053971781986857e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8056
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.8125,
      "epoch": 0.3893022806339389,
      "grad_norm": 0.0,
      "learning_rate": 3.053730189408581e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8057
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.625,
      "epoch": 0.38935059914959413,
      "grad_norm": 0.0,
      "learning_rate": 3.0534885968303054e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8058
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.625,
      "epoch": 0.3893989176652493,
      "grad_norm": 0.0,
      "learning_rate": 3.053247004252029e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8059
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.34375,
      "epoch": 0.38944723618090454,
      "grad_norm": 0.0,
      "learning_rate": 3.053005411673753e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8060
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.59375,
      "epoch": 0.3894955546965597,
      "grad_norm": 0.0,
      "learning_rate": 3.0527638190954773e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8061
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.15625,
      "epoch": 0.3895438732122149,
      "grad_norm": 0.0,
      "learning_rate": 3.0525222265172014e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8062
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.4375,
      "epoch": 0.38959219172787013,
      "grad_norm": 0.0,
      "learning_rate": 3.052280633938925e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8063
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.34375,
      "epoch": 0.3896405102435253,
      "grad_norm": 0.0,
      "learning_rate": 3.0520390413606497e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8064
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.59375,
      "epoch": 0.38968882875918054,
      "grad_norm": 0.0,
      "learning_rate": 3.0517974487823733e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8065
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.625,
      "epoch": 0.3897371472748357,
      "grad_norm": 0.0,
      "learning_rate": 3.051555856204097e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8066
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.1875,
      "epoch": 0.3897854657904909,
      "grad_norm": 0.0,
      "learning_rate": 3.051314263625821e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8067
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.78125,
      "epoch": 0.3898337843061461,
      "grad_norm": 0.0,
      "learning_rate": 3.051072671047545e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 8068
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.90625,
      "epoch": 0.3898821028218013,
      "grad_norm": 0.0,
      "learning_rate": 3.0508310784692693e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8069
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.28125,
      "epoch": 0.38993042133745653,
      "grad_norm": 0.0,
      "learning_rate": 3.050589485890993e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8070
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.21875,
      "epoch": 0.3899787398531117,
      "grad_norm": 0.0,
      "learning_rate": 3.0503478933127176e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8071
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.0625,
      "epoch": 0.3900270583687669,
      "grad_norm": 0.0,
      "learning_rate": 3.050106300734441e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8072
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.96875,
      "epoch": 0.3900753768844221,
      "grad_norm": 0.0,
      "learning_rate": 3.0498647081561653e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8073
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.90625,
      "epoch": 0.3901236954000773,
      "grad_norm": 0.0,
      "learning_rate": 3.0496231155778895e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 8074
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 79.59375,
      "epoch": 0.39017201391573253,
      "grad_norm": 7.80940620795363,
      "learning_rate": 3.0493815229996136e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8075
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.03125,
      "epoch": 0.3902203324313877,
      "grad_norm": 0.0,
      "learning_rate": 3.049139930421337e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8076
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.84375,
      "epoch": 0.3902686509470429,
      "grad_norm": 0.0,
      "learning_rate": 3.048898337843061e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8077
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.3125,
      "epoch": 0.3903169694626981,
      "grad_norm": 0.0,
      "learning_rate": 3.0486567452647855e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8078
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.875,
      "epoch": 0.3903652879783533,
      "grad_norm": 0.0,
      "learning_rate": 3.048415152686509e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8079
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 90.40625,
      "epoch": 0.3904136064940085,
      "grad_norm": 0.0,
      "learning_rate": 3.048173560108233e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 8080
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.46875,
      "epoch": 0.3904619250096637,
      "grad_norm": 0.0,
      "learning_rate": 3.0479319675299574e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8081
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.96875,
      "epoch": 0.3905102435253189,
      "grad_norm": 0.0,
      "learning_rate": 3.0476903749516815e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8082
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.0625,
      "epoch": 0.3905585620409741,
      "grad_norm": 0.0,
      "learning_rate": 3.047448782373405e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8083
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.34375,
      "epoch": 0.3906068805566293,
      "grad_norm": 31.139091282969524,
      "learning_rate": 3.04720718979513e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2783603370189667,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8084
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.59375,
      "epoch": 0.3906551990722845,
      "grad_norm": 0.0,
      "learning_rate": 3.0469655972168534e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8085
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.46875,
      "epoch": 0.3907035175879397,
      "grad_norm": 0.0,
      "learning_rate": 3.046724004638577e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8086
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.875,
      "epoch": 0.3907518361035949,
      "grad_norm": 36.670542746860924,
      "learning_rate": 3.046482412060301e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.27564918994903564,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8087
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.53125,
      "epoch": 0.3908001546192501,
      "grad_norm": 0.0,
      "learning_rate": 3.0462408194820253e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8088
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.09375,
      "epoch": 0.3908484731349053,
      "grad_norm": 0.0,
      "learning_rate": 3.0459992269037494e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8089
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.78125,
      "epoch": 0.3908967916505605,
      "grad_norm": 0.0,
      "learning_rate": 3.045757634325473e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8090
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.3125,
      "epoch": 0.3909451101662157,
      "grad_norm": 0.0,
      "learning_rate": 3.0455160417471977e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8091
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.40625,
      "epoch": 0.39099342868187087,
      "grad_norm": 0.0,
      "learning_rate": 3.0452744491689213e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8092
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.25,
      "epoch": 0.3910417471975261,
      "grad_norm": 0.0,
      "learning_rate": 3.0450328565906454e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8093
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.0,
      "epoch": 0.3910900657131813,
      "grad_norm": 0.0,
      "learning_rate": 3.0447912640123696e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8094
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.28125,
      "epoch": 0.3911383842288365,
      "grad_norm": 0.0,
      "learning_rate": 3.0445496714340937e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8095
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.75,
      "epoch": 0.3911867027444917,
      "grad_norm": 0.0,
      "learning_rate": 3.0443080788558173e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 8096
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.125,
      "epoch": 0.39123502126014686,
      "grad_norm": 0.0,
      "learning_rate": 3.044066486277541e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8097
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.625,
      "epoch": 0.3912833397758021,
      "grad_norm": 0.0,
      "learning_rate": 3.0438248936992656e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8098
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.40625,
      "epoch": 0.3913316582914573,
      "grad_norm": 0.0,
      "learning_rate": 3.043583301120989e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8099
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.71875,
      "epoch": 0.3913799768071125,
      "grad_norm": 0.0,
      "learning_rate": 3.0433417085427133e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8100
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.9375,
      "epoch": 0.3914282953227677,
      "grad_norm": 0.0,
      "learning_rate": 3.0431001159644375e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8101
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.9375,
      "epoch": 0.39147661383842286,
      "grad_norm": 0.0,
      "learning_rate": 3.0428585233861616e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8102
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.46875,
      "epoch": 0.3915249323540781,
      "grad_norm": 0.0,
      "learning_rate": 3.042616930807885e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8103
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.0625,
      "epoch": 0.39157325086973327,
      "grad_norm": 0.0,
      "learning_rate": 3.04237533822961e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8104
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.90625,
      "epoch": 0.3916215693853885,
      "grad_norm": 0.0,
      "learning_rate": 3.0421337456513335e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8105
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.59375,
      "epoch": 0.3916698879010437,
      "grad_norm": 0.0,
      "learning_rate": 3.0418921530730576e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8106
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.71875,
      "epoch": 0.39171820641669886,
      "grad_norm": 0.0,
      "learning_rate": 3.041650560494781e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8107
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.53125,
      "epoch": 0.3917665249323541,
      "grad_norm": 0.0,
      "learning_rate": 3.0414089679165054e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8108
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.84375,
      "epoch": 0.39181484344800926,
      "grad_norm": 0.0,
      "learning_rate": 3.0411673753382295e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8109
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.4375,
      "epoch": 0.3918631619636645,
      "grad_norm": 0.0,
      "learning_rate": 3.040925782759953e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8110
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.96875,
      "epoch": 0.3919114804793197,
      "grad_norm": 0.0,
      "learning_rate": 3.040684190181678e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8111
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.28125,
      "epoch": 0.39195979899497485,
      "grad_norm": 0.0,
      "learning_rate": 3.0404425976034014e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8112
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.71875,
      "epoch": 0.3920081175106301,
      "grad_norm": 0.0,
      "learning_rate": 3.0402010050251255e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8113
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.375,
      "epoch": 0.39205643602628526,
      "grad_norm": 0.0,
      "learning_rate": 3.0399594124468496e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8114
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.1875,
      "epoch": 0.3921047545419405,
      "grad_norm": 0.0,
      "learning_rate": 3.039717819868574e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8115
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.84375,
      "epoch": 0.39215307305759567,
      "grad_norm": 0.0,
      "learning_rate": 3.0394762272902974e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8116
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.96875,
      "epoch": 0.39220139157325085,
      "grad_norm": 0.0,
      "learning_rate": 3.039234634712021e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8117
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.34375,
      "epoch": 0.3922497100889061,
      "grad_norm": 0.0,
      "learning_rate": 3.0389930421337457e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8118
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.625,
      "epoch": 0.39229802860456126,
      "grad_norm": 0.0,
      "learning_rate": 3.038751449555469e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8119
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.84375,
      "epoch": 0.3923463471202165,
      "grad_norm": 0.0,
      "learning_rate": 3.0385098569771934e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8120
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.0625,
      "epoch": 0.39239466563587166,
      "grad_norm": 0.0,
      "learning_rate": 3.0382682643989175e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8121
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.875,
      "epoch": 0.39244298415152684,
      "grad_norm": 0.0,
      "learning_rate": 3.0380266718206417e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8122
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.5625,
      "epoch": 0.3924913026671821,
      "grad_norm": 0.0,
      "learning_rate": 3.0377850792423653e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8123
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.5,
      "epoch": 0.39253962118283725,
      "grad_norm": 0.0,
      "learning_rate": 3.03754348666409e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8124
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.40625,
      "epoch": 0.3925879396984925,
      "grad_norm": 0.0,
      "learning_rate": 3.0373018940858136e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8125
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.90625,
      "epoch": 0.39263625821414766,
      "grad_norm": 0.0,
      "learning_rate": 3.0370603015075377e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8126
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.53125,
      "epoch": 0.39268457672980284,
      "grad_norm": 0.0,
      "learning_rate": 3.0368187089292613e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8127
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.3125,
      "epoch": 0.39273289524545807,
      "grad_norm": 0.0,
      "learning_rate": 3.0365771163509854e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8128
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.625,
      "epoch": 0.39278121376111325,
      "grad_norm": 14.784645635923054,
      "learning_rate": 3.0363355237727096e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8129
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.21875,
      "epoch": 0.3928295322767685,
      "grad_norm": 0.0,
      "learning_rate": 3.036093931194433e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8130
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.9375,
      "epoch": 0.39287785079242366,
      "grad_norm": 0.0,
      "learning_rate": 3.035852338616158e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8131
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.875,
      "epoch": 0.39292616930807883,
      "grad_norm": 0.0,
      "learning_rate": 3.0356107460378814e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8132
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 97.25,
      "epoch": 0.39297448782373406,
      "grad_norm": 0.0,
      "learning_rate": 3.0353691534596056e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8133
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.15625,
      "epoch": 0.39302280633938924,
      "grad_norm": 0.0,
      "learning_rate": 3.0351275608813297e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8134
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.5625,
      "epoch": 0.3930711248550445,
      "grad_norm": 0.0,
      "learning_rate": 3.034885968303054e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8135
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.125,
      "epoch": 0.39311944337069965,
      "grad_norm": 0.0,
      "learning_rate": 3.0346443757247775e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8136
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.9375,
      "epoch": 0.3931677618863548,
      "grad_norm": 0.0,
      "learning_rate": 3.034402783146501e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8137
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.1875,
      "epoch": 0.39321608040201006,
      "grad_norm": 0.0,
      "learning_rate": 3.0341611905682257e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8138
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.3125,
      "epoch": 0.39326439891766524,
      "grad_norm": 0.0,
      "learning_rate": 3.0339195979899493e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8139
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.84375,
      "epoch": 0.39331271743332047,
      "grad_norm": 0.0,
      "learning_rate": 3.0336780054116735e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8140
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.875,
      "epoch": 0.39336103594897565,
      "grad_norm": 0.0,
      "learning_rate": 3.0334364128333976e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8141
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.28125,
      "epoch": 0.3934093544646308,
      "grad_norm": 0.0,
      "learning_rate": 3.033194820255122e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8142
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.5625,
      "epoch": 0.39345767298028606,
      "grad_norm": 0.0,
      "learning_rate": 3.0329532276768454e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8143
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.625,
      "epoch": 0.39350599149594123,
      "grad_norm": 0.0,
      "learning_rate": 3.03271163509857e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8144
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.875,
      "epoch": 0.39355431001159646,
      "grad_norm": 0.0,
      "learning_rate": 3.0324700425202936e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 8145
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.84375,
      "epoch": 0.39360262852725164,
      "grad_norm": 0.0,
      "learning_rate": 3.032228449942018e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8146
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.0625,
      "epoch": 0.3936509470429068,
      "grad_norm": 0.0,
      "learning_rate": 3.0319868573637414e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8147
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.65625,
      "epoch": 0.39369926555856205,
      "grad_norm": 0.0,
      "learning_rate": 3.0317452647854655e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8148
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.5,
      "epoch": 0.3937475840742172,
      "grad_norm": 0.0,
      "learning_rate": 3.0315036722071896e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8149
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.125,
      "epoch": 0.39379590258987246,
      "grad_norm": 0.0,
      "learning_rate": 3.031262079628913e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8150
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.78125,
      "epoch": 0.39384422110552764,
      "grad_norm": 0.0,
      "learning_rate": 3.031020487050638e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8151
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.46875,
      "epoch": 0.3938925396211828,
      "grad_norm": 0.0,
      "learning_rate": 3.0307788944723615e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8152
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.625,
      "epoch": 0.39394085813683805,
      "grad_norm": 0.0,
      "learning_rate": 3.0305373018940857e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8153
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.4375,
      "epoch": 0.3939891766524932,
      "grad_norm": 0.0,
      "learning_rate": 3.03029570931581e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8154
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.84375,
      "epoch": 0.39403749516814845,
      "grad_norm": 0.0,
      "learning_rate": 3.030054116737534e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8155
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.3125,
      "epoch": 0.39408581368380363,
      "grad_norm": 0.0,
      "learning_rate": 3.0298125241592575e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8156
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.125,
      "epoch": 0.3941341321994588,
      "grad_norm": 0.0,
      "learning_rate": 3.0295709315809817e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8157
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.90625,
      "epoch": 0.39418245071511404,
      "grad_norm": 0.0,
      "learning_rate": 3.029329339002706e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8158
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.8125,
      "epoch": 0.3942307692307692,
      "grad_norm": 0.0,
      "learning_rate": 3.0290877464244294e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8159
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.15625,
      "epoch": 0.39427908774642445,
      "grad_norm": 0.0,
      "learning_rate": 3.0288461538461536e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8160
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.125,
      "epoch": 0.3943274062620796,
      "grad_norm": 0.0,
      "learning_rate": 3.0286045612678777e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8161
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.59375,
      "epoch": 0.3943757247777348,
      "grad_norm": 0.0,
      "learning_rate": 3.028362968689602e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8162
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.0625,
      "epoch": 0.39442404329339004,
      "grad_norm": 0.0,
      "learning_rate": 3.0281213761113254e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8163
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.75,
      "epoch": 0.3944723618090452,
      "grad_norm": 0.0,
      "learning_rate": 3.02787978353305e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8164
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.84375,
      "epoch": 0.39452068032470045,
      "grad_norm": 0.0,
      "learning_rate": 3.0276381909547737e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8165
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.40625,
      "epoch": 0.3945689988403556,
      "grad_norm": 0.0,
      "learning_rate": 3.027396598376498e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8166
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.875,
      "epoch": 0.3946173173560108,
      "grad_norm": 0.0,
      "learning_rate": 3.0271550057982215e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8167
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.375,
      "epoch": 0.39466563587166603,
      "grad_norm": 0.0,
      "learning_rate": 3.026913413219946e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8168
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.34375,
      "epoch": 0.3947139543873212,
      "grad_norm": 14.941724457343287,
      "learning_rate": 3.0266718206416697e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8169
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.0625,
      "epoch": 0.39476227290297644,
      "grad_norm": 0.0,
      "learning_rate": 3.0264302280633933e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8170
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.71875,
      "epoch": 0.3948105914186316,
      "grad_norm": 0.0,
      "learning_rate": 3.026188635485118e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8171
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.625,
      "epoch": 0.3948589099342868,
      "grad_norm": 0.0,
      "learning_rate": 3.0259470429068416e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8172
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.21875,
      "epoch": 0.394907228449942,
      "grad_norm": 0.0,
      "learning_rate": 3.025705450328566e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8173
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.34375,
      "epoch": 0.3949555469655972,
      "grad_norm": 0.0,
      "learning_rate": 3.02546385775029e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8174
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.9375,
      "epoch": 0.39500386548125244,
      "grad_norm": 0.0,
      "learning_rate": 3.025222265172014e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8175
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.28125,
      "epoch": 0.3950521839969076,
      "grad_norm": 0.0,
      "learning_rate": 3.0249806725937376e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8176
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 101.78125,
      "epoch": 0.3951005025125628,
      "grad_norm": 22.12680432865024,
      "learning_rate": 3.024739080015462e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8177
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.40625,
      "epoch": 0.395148821028218,
      "grad_norm": 0.0,
      "learning_rate": 3.024497487437186e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8178
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.46875,
      "epoch": 0.3951971395438732,
      "grad_norm": 0.0,
      "learning_rate": 3.0242558948589095e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8179
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.90625,
      "epoch": 0.39524545805952843,
      "grad_norm": 0.0,
      "learning_rate": 3.0240143022806336e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8180
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.75,
      "epoch": 0.3952937765751836,
      "grad_norm": 0.0,
      "learning_rate": 3.023772709702358e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8181
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.09375,
      "epoch": 0.3953420950908388,
      "grad_norm": 0.0,
      "learning_rate": 3.023531117124082e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8182
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 96.625,
      "epoch": 0.395390413606494,
      "grad_norm": 46.80118279211116,
      "learning_rate": 3.0232895245458055e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 8183
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.53125,
      "epoch": 0.3954387321221492,
      "grad_norm": 0.0,
      "learning_rate": 3.02304793196753e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8184
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.8125,
      "epoch": 0.3954870506378044,
      "grad_norm": 0.0,
      "learning_rate": 3.022806339389254e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8185
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.4375,
      "epoch": 0.3955353691534596,
      "grad_norm": 0.0,
      "learning_rate": 3.022564746810978e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8186
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.9375,
      "epoch": 0.3955836876691148,
      "grad_norm": 0.0,
      "learning_rate": 3.0223231542327015e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8187
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.375,
      "epoch": 0.39563200618477,
      "grad_norm": 0.0,
      "learning_rate": 3.022081561654426e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8188
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.625,
      "epoch": 0.3956803247004252,
      "grad_norm": 0.0,
      "learning_rate": 3.02183996907615e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8189
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.40625,
      "epoch": 0.3957286432160804,
      "grad_norm": 0.0,
      "learning_rate": 3.0215983764978734e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8190
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.65625,
      "epoch": 0.3957769617317356,
      "grad_norm": 0.0,
      "learning_rate": 3.021356783919598e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8191
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.0,
      "epoch": 0.3958252802473908,
      "grad_norm": 0.0,
      "learning_rate": 3.0211151913413217e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8192
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 93.0,
      "epoch": 0.395873598763046,
      "grad_norm": 5.460516444281666,
      "learning_rate": 3.020873598763046e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.3014557659626007,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8193
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 97.9375,
      "epoch": 0.3959219172787012,
      "grad_norm": 45.71288379251731,
      "learning_rate": 3.02063200618477e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8194
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.75,
      "epoch": 0.3959702357943564,
      "grad_norm": 0.0,
      "learning_rate": 3.020390413606494e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8195
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.0,
      "epoch": 0.3960185543100116,
      "grad_norm": 0.0,
      "learning_rate": 3.0201488210282177e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8196
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.59375,
      "epoch": 0.39606687282566677,
      "grad_norm": 0.0,
      "learning_rate": 3.019907228449942e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8197
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.6875,
      "epoch": 0.396115191341322,
      "grad_norm": 0.0,
      "learning_rate": 3.019665635871666e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8198
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.375,
      "epoch": 0.3961635098569772,
      "grad_norm": 0.0,
      "learning_rate": 3.01942404329339e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8199
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.59375,
      "epoch": 0.3962118283726324,
      "grad_norm": 0.0,
      "learning_rate": 3.0191824507151137e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8200
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.15625,
      "epoch": 0.3962601468882876,
      "grad_norm": 0.0,
      "learning_rate": 3.018940858136838e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8201
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.40625,
      "epoch": 0.39630846540394277,
      "grad_norm": 0.0,
      "learning_rate": 3.018699265558562e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8202
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.28125,
      "epoch": 0.396356783919598,
      "grad_norm": 0.0,
      "learning_rate": 3.0184576729802856e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8203
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.78125,
      "epoch": 0.3964051024352532,
      "grad_norm": 0.0,
      "learning_rate": 3.01821608040201e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8204
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.9375,
      "epoch": 0.3964534209509084,
      "grad_norm": 0.0,
      "learning_rate": 3.017974487823734e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8205
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.3125,
      "epoch": 0.3965017394665636,
      "grad_norm": 0.0,
      "learning_rate": 3.017732895245458e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8206
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.65625,
      "epoch": 0.39655005798221876,
      "grad_norm": 0.0,
      "learning_rate": 3.0174913026671816e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8207
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.84375,
      "epoch": 0.396598376497874,
      "grad_norm": 10.278926944334987,
      "learning_rate": 3.0172497100889063e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.27883464097976685,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8208
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.40625,
      "epoch": 0.39664669501352917,
      "grad_norm": 0.0,
      "learning_rate": 3.01700811751063e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 8209
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.0625,
      "epoch": 0.3966950135291844,
      "grad_norm": 0.0,
      "learning_rate": 3.0167665249323535e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8210
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.1875,
      "epoch": 0.3967433320448396,
      "grad_norm": 0.0,
      "learning_rate": 3.016524932354078e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8211
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.34375,
      "epoch": 0.39679165056049476,
      "grad_norm": 0.0,
      "learning_rate": 3.016283339775802e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8212
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.125,
      "epoch": 0.39683996907615,
      "grad_norm": 0.0,
      "learning_rate": 3.016041747197526e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8213
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.53125,
      "epoch": 0.39688828759180517,
      "grad_norm": 0.0,
      "learning_rate": 3.01580015461925e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8214
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 93.90625,
      "epoch": 0.3969366061074604,
      "grad_norm": 7.203206588295655,
      "learning_rate": 3.015558562040974e-07,
      "loss": -0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 8215
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.125,
      "epoch": 0.3969849246231156,
      "grad_norm": 0.0,
      "learning_rate": 3.015316969462698e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8216
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.46875,
      "epoch": 0.39703324313877075,
      "grad_norm": 0.0,
      "learning_rate": 3.015075376884422e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8217
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.5,
      "epoch": 0.397081561654426,
      "grad_norm": 0.0,
      "learning_rate": 3.014833784306146e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8218
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.03125,
      "epoch": 0.39712988017008116,
      "grad_norm": 0.0,
      "learning_rate": 3.01459219172787e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8219
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.84375,
      "epoch": 0.3971781986857364,
      "grad_norm": 0.0,
      "learning_rate": 3.014350599149594e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8220
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.625,
      "epoch": 0.39722651720139157,
      "grad_norm": 0.0,
      "learning_rate": 3.014109006571318e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8221
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.8125,
      "epoch": 0.39727483571704675,
      "grad_norm": 0.0,
      "learning_rate": 3.013867413993042e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8222
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.3125,
      "epoch": 0.397323154232702,
      "grad_norm": 0.0,
      "learning_rate": 3.0136258214147657e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8223
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.1875,
      "epoch": 0.39737147274835716,
      "grad_norm": 0.0,
      "learning_rate": 3.0133842288364903e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8224
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.65625,
      "epoch": 0.3974197912640124,
      "grad_norm": 0.0,
      "learning_rate": 3.013142636258214e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8225
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.625,
      "epoch": 0.39746810977966757,
      "grad_norm": 0.0,
      "learning_rate": 3.012901043679938e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8226
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.4375,
      "epoch": 0.39751642829532274,
      "grad_norm": 0.0,
      "learning_rate": 3.0126594511016617e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8227
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 93.125,
      "epoch": 0.397564746810978,
      "grad_norm": 77.53630526510182,
      "learning_rate": 3.0124178585233864e-07,
      "loss": -0.0,
      "reward": 1.5625,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8228
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.9375,
      "epoch": 0.39761306532663315,
      "grad_norm": 0.0,
      "learning_rate": 3.01217626594511e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.19727617502212524,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8229
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.40625,
      "epoch": 0.3976613838422884,
      "grad_norm": 0.0,
      "learning_rate": 3.0119346733668336e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8230
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.78125,
      "epoch": 0.39770970235794356,
      "grad_norm": 0.0,
      "learning_rate": 3.011693080788558e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8231
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.375,
      "epoch": 0.39775802087359874,
      "grad_norm": 0.0,
      "learning_rate": 3.011451488210282e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8232
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.40625,
      "epoch": 0.39780633938925397,
      "grad_norm": 0.0,
      "learning_rate": 3.011209895632006e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8233
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.59375,
      "epoch": 0.39785465790490915,
      "grad_norm": 0.0,
      "learning_rate": 3.01096830305373e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8234
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.21875,
      "epoch": 0.3979029764205644,
      "grad_norm": 0.0,
      "learning_rate": 3.010726710475454e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8235
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.21875,
      "epoch": 0.39795129493621956,
      "grad_norm": 0.0,
      "learning_rate": 3.010485117897178e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8236
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.84375,
      "epoch": 0.39799961345187473,
      "grad_norm": 0.0,
      "learning_rate": 3.010243525318902e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8237
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.09375,
      "epoch": 0.39804793196752997,
      "grad_norm": 0.0,
      "learning_rate": 3.010001932740626e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8238
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.46875,
      "epoch": 0.39809625048318514,
      "grad_norm": 0.0,
      "learning_rate": 3.00976034016235e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8239
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.875,
      "epoch": 0.3981445689988404,
      "grad_norm": 0.0,
      "learning_rate": 3.009518747584074e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8240
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.4375,
      "epoch": 0.39819288751449555,
      "grad_norm": 0.0,
      "learning_rate": 3.009277155005798e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8241
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.53125,
      "epoch": 0.39824120603015073,
      "grad_norm": 0.0,
      "learning_rate": 3.009035562427522e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8242
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.25,
      "epoch": 0.39828952454580596,
      "grad_norm": 0.0,
      "learning_rate": 3.008793969849246e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8243
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.34375,
      "epoch": 0.39833784306146114,
      "grad_norm": 0.0,
      "learning_rate": 3.0085523772709704e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8244
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.375,
      "epoch": 0.39838616157711637,
      "grad_norm": 0.0,
      "learning_rate": 3.008310784692694e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8245
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.96875,
      "epoch": 0.39843448009277155,
      "grad_norm": 0.0,
      "learning_rate": 3.008069192114418e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8246
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.5,
      "epoch": 0.3984827986084267,
      "grad_norm": 0.0,
      "learning_rate": 3.007827599536142e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8247
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.09375,
      "epoch": 0.39853111712408196,
      "grad_norm": 0.0,
      "learning_rate": 3.0075860069578664e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8248
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.21875,
      "epoch": 0.39857943563973713,
      "grad_norm": 0.0,
      "learning_rate": 3.00734441437959e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 8249
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.0625,
      "epoch": 0.39862775415539237,
      "grad_norm": 0.0,
      "learning_rate": 3.007102821801314e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8250
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.125,
      "epoch": 0.39867607267104754,
      "grad_norm": 0.0,
      "learning_rate": 3.0068612292230383e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8251
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.5,
      "epoch": 0.3987243911867027,
      "grad_norm": 0.0,
      "learning_rate": 3.006619636644762e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8252
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.1875,
      "epoch": 0.39877270970235795,
      "grad_norm": 0.0,
      "learning_rate": 3.006378044066486e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8253
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.59375,
      "epoch": 0.39882102821801313,
      "grad_norm": 0.0,
      "learning_rate": 3.00613645148821e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8254
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.1875,
      "epoch": 0.39886934673366836,
      "grad_norm": 0.0,
      "learning_rate": 3.0058948589099343e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8255
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.53125,
      "epoch": 0.39891766524932354,
      "grad_norm": 0.0,
      "learning_rate": 3.005653266331658e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8256
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.53125,
      "epoch": 0.3989659837649787,
      "grad_norm": 0.0,
      "learning_rate": 3.005411673753382e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8257
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 101.5,
      "epoch": 0.39901430228063395,
      "grad_norm": 0.0,
      "learning_rate": 3.005170081175106e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 8258
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.25,
      "epoch": 0.3990626207962891,
      "grad_norm": 0.0,
      "learning_rate": 3.0049284885968303e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8259
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.375,
      "epoch": 0.39911093931194436,
      "grad_norm": 16.746195136007888,
      "learning_rate": 3.004686896018554e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.2961486577987671,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8260
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.8125,
      "epoch": 0.39915925782759953,
      "grad_norm": 0.0,
      "learning_rate": 3.0044453034402786e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8261
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.5625,
      "epoch": 0.3992075763432547,
      "grad_norm": 0.0,
      "learning_rate": 3.004203710862002e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8262
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.75,
      "epoch": 0.39925589485890994,
      "grad_norm": 0.0,
      "learning_rate": 3.003962118283726e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8263
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.6875,
      "epoch": 0.3993042133745651,
      "grad_norm": 0.0,
      "learning_rate": 3.0037205257054505e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8264
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.40625,
      "epoch": 0.39935253189022035,
      "grad_norm": 0.0,
      "learning_rate": 3.003478933127174e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8265
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.5625,
      "epoch": 0.39940085040587553,
      "grad_norm": 0.0,
      "learning_rate": 3.003237340548898e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8266
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.84375,
      "epoch": 0.3994491689215307,
      "grad_norm": 0.0,
      "learning_rate": 3.002995747970622e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8267
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.5625,
      "epoch": 0.39949748743718594,
      "grad_norm": 0.0,
      "learning_rate": 3.0027541553923465e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8268
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.6875,
      "epoch": 0.3995458059528411,
      "grad_norm": 0.0,
      "learning_rate": 3.00251256281407e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8269
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.65625,
      "epoch": 0.39959412446849635,
      "grad_norm": 0.0,
      "learning_rate": 3.002270970235794e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8270
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.78125,
      "epoch": 0.3996424429841515,
      "grad_norm": 0.0,
      "learning_rate": 3.0020293776575184e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8271
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 89.6875,
      "epoch": 0.3996907614998067,
      "grad_norm": 15.670225093833546,
      "learning_rate": 3.001787785079242e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.20348434150218964,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8272
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.96875,
      "epoch": 0.39973908001546193,
      "grad_norm": 0.0,
      "learning_rate": 3.001546192500966e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8273
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.125,
      "epoch": 0.3997873985311171,
      "grad_norm": 0.0,
      "learning_rate": 3.0013045999226903e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8274
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.21875,
      "epoch": 0.39983571704677234,
      "grad_norm": 0.0,
      "learning_rate": 3.0010630073444144e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8275
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.84375,
      "epoch": 0.3998840355624275,
      "grad_norm": 0.0,
      "learning_rate": 3.000821414766138e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8276
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.125,
      "epoch": 0.3999323540780827,
      "grad_norm": 0.0,
      "learning_rate": 3.000579822187862e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8277
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.875,
      "epoch": 0.39998067259373793,
      "grad_norm": 0.0,
      "learning_rate": 3.0003382296095863e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8278
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.90625,
      "epoch": 0.4000289911093931,
      "grad_norm": 23.572535736515615,
      "learning_rate": 3.0000966370313104e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.3766237795352936,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8279
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.21875,
      "epoch": 0.40007730962504834,
      "grad_norm": 0.0,
      "learning_rate": 2.999855044453034e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8280
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.0,
      "epoch": 0.4001256281407035,
      "grad_norm": 0.0,
      "learning_rate": 2.9996134518747587e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8281
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.53125,
      "epoch": 0.4001739466563587,
      "grad_norm": 0.0,
      "learning_rate": 2.9993718592964823e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 8282
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.3125,
      "epoch": 0.4002222651720139,
      "grad_norm": 0.0,
      "learning_rate": 2.999130266718206e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8283
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.65625,
      "epoch": 0.4002705836876691,
      "grad_norm": 0.0,
      "learning_rate": 2.9988886741399306e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8284
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.34375,
      "epoch": 0.40031890220332433,
      "grad_norm": 0.0,
      "learning_rate": 2.998647081561654e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8285
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.09375,
      "epoch": 0.4003672207189795,
      "grad_norm": 0.0,
      "learning_rate": 2.9984054889833783e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8286
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 77.71875,
      "epoch": 0.4004155392346347,
      "grad_norm": 31.24787404025116,
      "learning_rate": 2.998163896405102e-07,
      "loss": -0.0,
      "reward": 2.28125,
      "reward_std": 0.29305070638656616,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8287
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.75,
      "epoch": 0.4004638577502899,
      "grad_norm": 0.0,
      "learning_rate": 2.9979223038268266e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 8288
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.5625,
      "epoch": 0.4005121762659451,
      "grad_norm": 0.0,
      "learning_rate": 2.99768071124855e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8289
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.0625,
      "epoch": 0.40056049478160033,
      "grad_norm": 0.0,
      "learning_rate": 2.9974391186702743e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8290
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.21875,
      "epoch": 0.4006088132972555,
      "grad_norm": 0.0,
      "learning_rate": 2.9971975260919985e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8291
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.09375,
      "epoch": 0.4006571318129107,
      "grad_norm": 0.0,
      "learning_rate": 2.996955933513722e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8292
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 76.71875,
      "epoch": 0.4007054503285659,
      "grad_norm": 42.409286026449124,
      "learning_rate": 2.996714340935446e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.2893187999725342,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8293
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.4375,
      "epoch": 0.4007537688442211,
      "grad_norm": 0.0,
      "learning_rate": 2.9964727483571703e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8294
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.15625,
      "epoch": 0.4008020873598763,
      "grad_norm": 0.0,
      "learning_rate": 2.9962311557788945e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8295
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.28125,
      "epoch": 0.4008504058755315,
      "grad_norm": 0.0,
      "learning_rate": 2.995989563200618e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8296
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.1875,
      "epoch": 0.4008987243911867,
      "grad_norm": 0.0,
      "learning_rate": 2.995747970622342e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8297
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.6875,
      "epoch": 0.4009470429068419,
      "grad_norm": 0.0,
      "learning_rate": 2.9955063780440664e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8298
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.8125,
      "epoch": 0.4009953614224971,
      "grad_norm": 0.0,
      "learning_rate": 2.9952647854657905e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8299
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.96875,
      "epoch": 0.4010436799381523,
      "grad_norm": 0.0,
      "learning_rate": 2.995023192887514e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8300
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.84375,
      "epoch": 0.4010919984538075,
      "grad_norm": 0.0,
      "learning_rate": 2.994781600309239e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8301
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.59375,
      "epoch": 0.4011403169694627,
      "grad_norm": 0.0,
      "learning_rate": 2.9945400077309624e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 8302
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.5625,
      "epoch": 0.4011886354851179,
      "grad_norm": 0.0,
      "learning_rate": 2.994298415152686e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 8303
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.125,
      "epoch": 0.4012369540007731,
      "grad_norm": 0.0,
      "learning_rate": 2.9940568225744107e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 8304
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.21875,
      "epoch": 0.4012852725164283,
      "grad_norm": 0.0,
      "learning_rate": 2.993815229996134e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8305
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.75,
      "epoch": 0.4013335910320835,
      "grad_norm": 0.0,
      "learning_rate": 2.9935736374178584e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8306
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.0,
      "epoch": 0.40138190954773867,
      "grad_norm": 0.0,
      "learning_rate": 2.993332044839582e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8307
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.8125,
      "epoch": 0.4014302280633939,
      "grad_norm": 0.0,
      "learning_rate": 2.9930904522613067e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8308
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.15625,
      "epoch": 0.4014785465790491,
      "grad_norm": 0.0,
      "learning_rate": 2.9928488596830303e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8309
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.09375,
      "epoch": 0.4015268650947043,
      "grad_norm": 0.0,
      "learning_rate": 2.9926072671047544e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8310
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.09375,
      "epoch": 0.4015751836103595,
      "grad_norm": 0.0,
      "learning_rate": 2.9923656745264785e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8311
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.15625,
      "epoch": 0.40162350212601466,
      "grad_norm": 0.0,
      "learning_rate": 2.9921240819482027e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8312
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.0625,
      "epoch": 0.4016718206416699,
      "grad_norm": 0.0,
      "learning_rate": 2.9918824893699263e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8313
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.03125,
      "epoch": 0.4017201391573251,
      "grad_norm": 0.0,
      "learning_rate": 2.9916408967916504e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8314
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.65625,
      "epoch": 0.4017684576729803,
      "grad_norm": 0.0,
      "learning_rate": 2.9913993042133746e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8315
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.75,
      "epoch": 0.4018167761886355,
      "grad_norm": 0.0,
      "learning_rate": 2.991157711635098e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8316
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.3125,
      "epoch": 0.40186509470429066,
      "grad_norm": 0.0,
      "learning_rate": 2.9909161190568223e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8317
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.90625,
      "epoch": 0.4019134132199459,
      "grad_norm": 0.0,
      "learning_rate": 2.9906745264785464e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8318
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.96875,
      "epoch": 0.40196173173560107,
      "grad_norm": 0.0,
      "learning_rate": 2.9904329339002706e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8319
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.8125,
      "epoch": 0.4020100502512563,
      "grad_norm": 0.0,
      "learning_rate": 2.990191341321994e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8320
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.5625,
      "epoch": 0.4020583687669115,
      "grad_norm": 0.0,
      "learning_rate": 2.989949748743719e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8321
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.96875,
      "epoch": 0.40210668728256665,
      "grad_norm": 0.0,
      "learning_rate": 2.9897081561654425e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8322
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.78125,
      "epoch": 0.4021550057982219,
      "grad_norm": 0.0,
      "learning_rate": 2.989466563587166e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8323
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.90625,
      "epoch": 0.40220332431387706,
      "grad_norm": 0.0,
      "learning_rate": 2.9892249710088907e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8324
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.84375,
      "epoch": 0.4022516428295323,
      "grad_norm": 0.0,
      "learning_rate": 2.9889833784306143e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8325
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.625,
      "epoch": 0.4022999613451875,
      "grad_norm": 0.0,
      "learning_rate": 2.9887417858523385e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8326
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 81.21875,
      "epoch": 0.40234827986084265,
      "grad_norm": 0.0,
      "learning_rate": 2.988500193274062e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 8327
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.84375,
      "epoch": 0.4023965983764979,
      "grad_norm": 0.0,
      "learning_rate": 2.988258600695787e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8328
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.25,
      "epoch": 0.40244491689215306,
      "grad_norm": 0.0,
      "learning_rate": 2.9880170081175104e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8329
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.65625,
      "epoch": 0.4024932354078083,
      "grad_norm": 0.0,
      "learning_rate": 2.9877754155392345e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8330
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.75,
      "epoch": 0.40254155392346347,
      "grad_norm": 0.0,
      "learning_rate": 2.9875338229609586e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 8331
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.5,
      "epoch": 0.40258987243911865,
      "grad_norm": 0.0,
      "learning_rate": 2.987292230382683e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8332
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.21875,
      "epoch": 0.4026381909547739,
      "grad_norm": 0.0,
      "learning_rate": 2.9870506378044064e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8333
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 77.4375,
      "epoch": 0.40268650947042905,
      "grad_norm": 29.141867170582746,
      "learning_rate": 2.9868090452261305e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8334
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.40625,
      "epoch": 0.4027348279860843,
      "grad_norm": 0.0,
      "learning_rate": 2.9865674526478546e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8335
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.59375,
      "epoch": 0.40278314650173946,
      "grad_norm": 0.0,
      "learning_rate": 2.986325860069578e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8336
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.75,
      "epoch": 0.40283146501739464,
      "grad_norm": 0.0,
      "learning_rate": 2.9860842674913024e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8337
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.625,
      "epoch": 0.4028797835330499,
      "grad_norm": 0.0,
      "learning_rate": 2.9858426749130265e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8338
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.78125,
      "epoch": 0.40292810204870505,
      "grad_norm": 0.0,
      "learning_rate": 2.9856010823347507e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8339
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.84375,
      "epoch": 0.4029764205643603,
      "grad_norm": 0.0,
      "learning_rate": 2.985359489756474e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8340
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.40625,
      "epoch": 0.40302473908001546,
      "grad_norm": 0.0,
      "learning_rate": 2.985117897178199e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8341
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.625,
      "epoch": 0.40307305759567064,
      "grad_norm": 0.0,
      "learning_rate": 2.9848763045999225e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8342
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.4375,
      "epoch": 0.40312137611132587,
      "grad_norm": 0.0,
      "learning_rate": 2.984634712021646e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8343
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.15625,
      "epoch": 0.40316969462698105,
      "grad_norm": 0.0,
      "learning_rate": 2.984393119443371e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8344
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.8125,
      "epoch": 0.4032180131426363,
      "grad_norm": 0.0,
      "learning_rate": 2.9841515268650944e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8345
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.59375,
      "epoch": 0.40326633165829145,
      "grad_norm": 0.0,
      "learning_rate": 2.9839099342868186e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8346
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.28125,
      "epoch": 0.40331465017394663,
      "grad_norm": 0.0,
      "learning_rate": 2.983668341708542e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8347
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.875,
      "epoch": 0.40336296868960186,
      "grad_norm": 0.0,
      "learning_rate": 2.983426749130267e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 8348
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.5625,
      "epoch": 0.40341128720525704,
      "grad_norm": 0.0,
      "learning_rate": 2.9831851565519904e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 8349
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.0,
      "epoch": 0.4034596057209123,
      "grad_norm": 0.0,
      "learning_rate": 2.9829435639737146e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8350
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.78125,
      "epoch": 0.40350792423656745,
      "grad_norm": 0.0,
      "learning_rate": 2.9827019713954387e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8351
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.25,
      "epoch": 0.4035562427522226,
      "grad_norm": 0.0,
      "learning_rate": 2.982460378817163e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8352
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.53125,
      "epoch": 0.40360456126787786,
      "grad_norm": 0.0,
      "learning_rate": 2.9822187862388864e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8353
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.8125,
      "epoch": 0.40365287978353304,
      "grad_norm": 0.0,
      "learning_rate": 2.9819771936606106e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8354
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.5,
      "epoch": 0.40370119829918827,
      "grad_norm": 0.0,
      "learning_rate": 2.9817356010823347e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8355
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.9375,
      "epoch": 0.40374951681484345,
      "grad_norm": 0.0,
      "learning_rate": 2.9814940085040583e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8356
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.375,
      "epoch": 0.4037978353304986,
      "grad_norm": 0.0,
      "learning_rate": 2.9812524159257825e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8357
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 91.53125,
      "epoch": 0.40384615384615385,
      "grad_norm": 60.352557594237474,
      "learning_rate": 2.9810108233475066e-07,
      "loss": -0.0,
      "reward": 1.5,
      "reward_std": 0.2529362142086029,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8358
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.96875,
      "epoch": 0.40389447236180903,
      "grad_norm": 0.0,
      "learning_rate": 2.980769230769231e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8359
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.53125,
      "epoch": 0.40394279087746426,
      "grad_norm": 0.0,
      "learning_rate": 2.9805276381909543e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8360
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.6875,
      "epoch": 0.40399110939311944,
      "grad_norm": 0.0,
      "learning_rate": 2.980286045612679e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8361
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.1875,
      "epoch": 0.4040394279087746,
      "grad_norm": 0.0,
      "learning_rate": 2.9800444530344026e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8362
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.28125,
      "epoch": 0.40408774642442985,
      "grad_norm": 0.0,
      "learning_rate": 2.979802860456127e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8363
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.375,
      "epoch": 0.404136064940085,
      "grad_norm": 0.0,
      "learning_rate": 2.979561267877851e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8364
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.4375,
      "epoch": 0.40418438345574026,
      "grad_norm": 0.0,
      "learning_rate": 2.9793196752995745e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8365
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.15625,
      "epoch": 0.40423270197139544,
      "grad_norm": 0.0,
      "learning_rate": 2.9790780827212986e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8366
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.5,
      "epoch": 0.4042810204870506,
      "grad_norm": 0.0,
      "learning_rate": 2.978836490143022e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8367
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.59375,
      "epoch": 0.40432933900270585,
      "grad_norm": 0.0,
      "learning_rate": 2.978594897564747e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8368
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.125,
      "epoch": 0.404377657518361,
      "grad_norm": 0.0,
      "learning_rate": 2.9783533049864705e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8369
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.75,
      "epoch": 0.40442597603401625,
      "grad_norm": 0.0,
      "learning_rate": 2.9781117124081946e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8370
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.6875,
      "epoch": 0.40447429454967143,
      "grad_norm": 0.0,
      "learning_rate": 2.977870119829919e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8371
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.8125,
      "epoch": 0.4045226130653266,
      "grad_norm": 0.0,
      "learning_rate": 2.977628527251643e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8372
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.59375,
      "epoch": 0.40457093158098184,
      "grad_norm": 0.0,
      "learning_rate": 2.9773869346733665e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8373
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.375,
      "epoch": 0.404619250096637,
      "grad_norm": 0.0,
      "learning_rate": 2.977145342095091e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8374
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.5625,
      "epoch": 0.40466756861229225,
      "grad_norm": 0.0,
      "learning_rate": 2.976903749516815e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8375
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.53125,
      "epoch": 0.4047158871279474,
      "grad_norm": 0.0,
      "learning_rate": 2.9766621569385384e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2130674123764038,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8376
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.40625,
      "epoch": 0.4047642056436026,
      "grad_norm": 0.0,
      "learning_rate": 2.9764205643602625e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 8377
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.84375,
      "epoch": 0.40481252415925784,
      "grad_norm": 0.0,
      "learning_rate": 2.9761789717819867e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8378
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.9375,
      "epoch": 0.404860842674913,
      "grad_norm": 0.0,
      "learning_rate": 2.975937379203711e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8379
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.9375,
      "epoch": 0.40490916119056825,
      "grad_norm": 0.0,
      "learning_rate": 2.9756957866254344e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8380
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.90625,
      "epoch": 0.4049574797062234,
      "grad_norm": 0.0,
      "learning_rate": 2.975454194047159e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8381
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.9375,
      "epoch": 0.4050057982218786,
      "grad_norm": 0.0,
      "learning_rate": 2.9752126014688827e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8382
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 84.6875,
      "epoch": 0.40505411673753383,
      "grad_norm": 0.0,
      "learning_rate": 2.974971008890607e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 8383
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.03125,
      "epoch": 0.405102435253189,
      "grad_norm": 0.0,
      "learning_rate": 2.974729416312331e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8384
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.15625,
      "epoch": 0.40515075376884424,
      "grad_norm": 0.0,
      "learning_rate": 2.9744878237340546e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8385
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.9375,
      "epoch": 0.4051990722844994,
      "grad_norm": 0.0,
      "learning_rate": 2.9742462311557787e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8386
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.40625,
      "epoch": 0.4052473908001546,
      "grad_norm": 0.0,
      "learning_rate": 2.9740046385775023e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8387
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.40625,
      "epoch": 0.4052957093158098,
      "grad_norm": 0.0,
      "learning_rate": 2.973763045999227e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8388
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.71875,
      "epoch": 0.405344027831465,
      "grad_norm": 0.0,
      "learning_rate": 2.9735214534209506e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8389
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.6875,
      "epoch": 0.40539234634712024,
      "grad_norm": 0.0,
      "learning_rate": 2.9732798608426747e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8390
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.28125,
      "epoch": 0.4054406648627754,
      "grad_norm": 0.0,
      "learning_rate": 2.973038268264399e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8391
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.28125,
      "epoch": 0.4054889833784306,
      "grad_norm": 0.0,
      "learning_rate": 2.972796675686123e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8392
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.71875,
      "epoch": 0.4055373018940858,
      "grad_norm": 0.0,
      "learning_rate": 2.9725550831078466e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8393
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.25,
      "epoch": 0.405585620409741,
      "grad_norm": 0.0,
      "learning_rate": 2.9723134905295713e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8394
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.78125,
      "epoch": 0.40563393892539623,
      "grad_norm": 0.0,
      "learning_rate": 2.972071897951295e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8395
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.1875,
      "epoch": 0.4056822574410514,
      "grad_norm": 0.0,
      "learning_rate": 2.9718303053730185e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8396
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.3125,
      "epoch": 0.4057305759567066,
      "grad_norm": 0.0,
      "learning_rate": 2.9715887127947426e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8397
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.40625,
      "epoch": 0.4057788944723618,
      "grad_norm": 0.0,
      "learning_rate": 2.971347120216467e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8398
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.9375,
      "epoch": 0.405827212988017,
      "grad_norm": 0.0,
      "learning_rate": 2.971105527638191e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8399
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.6875,
      "epoch": 0.4058755315036722,
      "grad_norm": 0.0,
      "learning_rate": 2.9708639350599145e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8400
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.96875,
      "epoch": 0.4059238500193274,
      "grad_norm": 0.0,
      "learning_rate": 2.970622342481639e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8401
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.21875,
      "epoch": 0.4059721685349826,
      "grad_norm": 0.0,
      "learning_rate": 2.970380749903363e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8402
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.59375,
      "epoch": 0.4060204870506378,
      "grad_norm": 0.0,
      "learning_rate": 2.970139157325087e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8403
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.8125,
      "epoch": 0.406068805566293,
      "grad_norm": 0.0,
      "learning_rate": 2.969897564746811e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8404
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.21875,
      "epoch": 0.4061171240819482,
      "grad_norm": 0.0,
      "learning_rate": 2.969655972168535e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8405
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.875,
      "epoch": 0.4061654425976034,
      "grad_norm": 0.0,
      "learning_rate": 2.969414379590259e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8406
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.8125,
      "epoch": 0.4062137611132586,
      "grad_norm": 0.0,
      "learning_rate": 2.9691727870119824e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8407
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.90625,
      "epoch": 0.4062620796289138,
      "grad_norm": 0.0,
      "learning_rate": 2.968931194433707e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8408
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.96875,
      "epoch": 0.406310398144569,
      "grad_norm": 0.0,
      "learning_rate": 2.9686896018554307e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8409
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.78125,
      "epoch": 0.4063587166602242,
      "grad_norm": 0.0,
      "learning_rate": 2.968448009277155e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8410
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.09375,
      "epoch": 0.4064070351758794,
      "grad_norm": 0.0,
      "learning_rate": 2.968206416698879e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8411
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.3125,
      "epoch": 0.40645535369153457,
      "grad_norm": 0.0,
      "learning_rate": 2.967964824120603e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8412
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.8125,
      "epoch": 0.4065036722071898,
      "grad_norm": 0.0,
      "learning_rate": 2.9677232315423267e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8413
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.875,
      "epoch": 0.406551990722845,
      "grad_norm": 0.0,
      "learning_rate": 2.9674816389640513e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8414
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.0625,
      "epoch": 0.4066003092385002,
      "grad_norm": 0.0,
      "learning_rate": 2.967240046385775e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 8415
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 86.84375,
      "epoch": 0.4066486277541554,
      "grad_norm": 31.193029415097662,
      "learning_rate": 2.9669984538074986e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8416
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.6875,
      "epoch": 0.40669694626981057,
      "grad_norm": 0.0,
      "learning_rate": 2.9667568612292227e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8417
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.40625,
      "epoch": 0.4067452647854658,
      "grad_norm": 0.0,
      "learning_rate": 2.966515268650947e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8418
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.34375,
      "epoch": 0.406793583301121,
      "grad_norm": 0.0,
      "learning_rate": 2.966273676072671e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8419
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.59375,
      "epoch": 0.4068419018167762,
      "grad_norm": 0.0,
      "learning_rate": 2.9660320834943946e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880052089691,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8420
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.34375,
      "epoch": 0.4068902203324314,
      "grad_norm": 10.018317824885344,
      "learning_rate": 2.965790490916119e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8421
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.09375,
      "epoch": 0.40693853884808656,
      "grad_norm": 0.0,
      "learning_rate": 2.965548898337843e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8422
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.78125,
      "epoch": 0.4069868573637418,
      "grad_norm": 0.0,
      "learning_rate": 2.965307305759567e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8423
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 79.53125,
      "epoch": 0.40703517587939697,
      "grad_norm": 0.0,
      "learning_rate": 2.965065713181291e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 8424
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.59375,
      "epoch": 0.4070834943950522,
      "grad_norm": 0.0,
      "learning_rate": 2.964824120603015e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24251842498779297,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8425
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.59375,
      "epoch": 0.4071318129107074,
      "grad_norm": 0.0,
      "learning_rate": 2.964582528024739e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8426
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.03125,
      "epoch": 0.40718013142636256,
      "grad_norm": 0.0,
      "learning_rate": 2.9643409354464625e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8427
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.5625,
      "epoch": 0.4072284499420178,
      "grad_norm": 0.0,
      "learning_rate": 2.964099342868187e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8428
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.0625,
      "epoch": 0.40727676845767297,
      "grad_norm": 0.0,
      "learning_rate": 2.963857750289911e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8429
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.03125,
      "epoch": 0.4073250869733282,
      "grad_norm": 0.0,
      "learning_rate": 2.963616157711635e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8430
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.28125,
      "epoch": 0.4073734054889834,
      "grad_norm": 0.0,
      "learning_rate": 2.963374565133359e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8431
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.875,
      "epoch": 0.40742172400463855,
      "grad_norm": 0.0,
      "learning_rate": 2.963132972555083e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8432
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.59375,
      "epoch": 0.4074700425202938,
      "grad_norm": 0.0,
      "learning_rate": 2.962891379976807e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8433
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.21875,
      "epoch": 0.40751836103594896,
      "grad_norm": 0.0,
      "learning_rate": 2.9626497873985314e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 8434
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.21875,
      "epoch": 0.4075666795516042,
      "grad_norm": 0.0,
      "learning_rate": 2.962408194820255e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8435
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.34375,
      "epoch": 0.40761499806725937,
      "grad_norm": 0.0,
      "learning_rate": 2.9621666022419786e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 8436
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.0625,
      "epoch": 0.40766331658291455,
      "grad_norm": 0.0,
      "learning_rate": 2.961925009663703e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 8437
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.8125,
      "epoch": 0.4077116350985698,
      "grad_norm": 0.0,
      "learning_rate": 2.961683417085427e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8438
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.375,
      "epoch": 0.40775995361422496,
      "grad_norm": 0.0,
      "learning_rate": 2.961441824507151e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8439
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.65625,
      "epoch": 0.4078082721298802,
      "grad_norm": 0.0,
      "learning_rate": 2.9612002319288747e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.24464009702205658,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8440
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.6875,
      "epoch": 0.40785659064553537,
      "grad_norm": 0.0,
      "learning_rate": 2.9609586393505993e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8441
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.3125,
      "epoch": 0.40790490916119054,
      "grad_norm": 0.0,
      "learning_rate": 2.960717046772323e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8442
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.09375,
      "epoch": 0.4079532276768458,
      "grad_norm": 0.0,
      "learning_rate": 2.960475454194047e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8443
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.75,
      "epoch": 0.40800154619250095,
      "grad_norm": 0.0,
      "learning_rate": 2.960233861615771e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8444
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.75,
      "epoch": 0.4080498647081562,
      "grad_norm": 0.0,
      "learning_rate": 2.9599922690374953e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8445
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.375,
      "epoch": 0.40809818322381136,
      "grad_norm": 0.0,
      "learning_rate": 2.959750676459219e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8446
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.53125,
      "epoch": 0.40814650173946654,
      "grad_norm": 0.0,
      "learning_rate": 2.9595090838809426e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8447
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.125,
      "epoch": 0.40819482025512177,
      "grad_norm": 0.0,
      "learning_rate": 2.959267491302667e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8448
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.15625,
      "epoch": 0.40824313877077695,
      "grad_norm": 0.0,
      "learning_rate": 2.959025898724391e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8449
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.875,
      "epoch": 0.4082914572864322,
      "grad_norm": 0.0,
      "learning_rate": 2.958784306146115e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8450
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.15625,
      "epoch": 0.40833977580208736,
      "grad_norm": 0.0,
      "learning_rate": 2.958542713567839e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8451
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.40625,
      "epoch": 0.40838809431774253,
      "grad_norm": 0.0,
      "learning_rate": 2.958301120989563e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8452
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.6875,
      "epoch": 0.40843641283339777,
      "grad_norm": 17.180555469253573,
      "learning_rate": 2.958059528411287e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.3061639666557312,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8453
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 82.1875,
      "epoch": 0.40848473134905294,
      "grad_norm": 13.738946774304717,
      "learning_rate": 2.9578179358330115e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8454
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.4375,
      "epoch": 0.4085330498647082,
      "grad_norm": 0.0,
      "learning_rate": 2.957576343254735e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8455
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.09375,
      "epoch": 0.40858136838036335,
      "grad_norm": 0.0,
      "learning_rate": 2.957334750676459e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8456
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.5,
      "epoch": 0.40862968689601853,
      "grad_norm": 0.0,
      "learning_rate": 2.957093158098183e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8457
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.0,
      "epoch": 0.40867800541167376,
      "grad_norm": 0.0,
      "learning_rate": 2.956851565519907e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8458
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.21875,
      "epoch": 0.40872632392732894,
      "grad_norm": 0.0,
      "learning_rate": 2.956609972941631e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 8459
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.8125,
      "epoch": 0.40877464244298417,
      "grad_norm": 0.0,
      "learning_rate": 2.956368380363355e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8460
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.3125,
      "epoch": 0.40882296095863935,
      "grad_norm": 0.0,
      "learning_rate": 2.9561267877850794e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8461
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.875,
      "epoch": 0.4088712794742945,
      "grad_norm": 0.0,
      "learning_rate": 2.955885195206803e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8462
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.15625,
      "epoch": 0.40891959798994976,
      "grad_norm": 0.0,
      "learning_rate": 2.955643602628527e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8463
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.9375,
      "epoch": 0.40896791650560493,
      "grad_norm": 0.0,
      "learning_rate": 2.9554020100502513e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8464
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.6875,
      "epoch": 0.40901623502126017,
      "grad_norm": 0.0,
      "learning_rate": 2.9551604174719754e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8465
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.6875,
      "epoch": 0.40906455353691534,
      "grad_norm": 0.0,
      "learning_rate": 2.954918824893699e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8466
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.0,
      "epoch": 0.4091128720525705,
      "grad_norm": 0.0,
      "learning_rate": 2.9546772323154226e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8467
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.28125,
      "epoch": 0.40916119056822575,
      "grad_norm": 0.0,
      "learning_rate": 2.9544356397371473e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8468
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.25,
      "epoch": 0.40920950908388093,
      "grad_norm": 0.0,
      "learning_rate": 2.954194047158871e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8469
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.0,
      "epoch": 0.40925782759953616,
      "grad_norm": 0.0,
      "learning_rate": 2.953952454580595e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8470
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.28125,
      "epoch": 0.40930614611519134,
      "grad_norm": 0.0,
      "learning_rate": 2.953710862002319e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8471
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.28125,
      "epoch": 0.4093544646308465,
      "grad_norm": 0.0,
      "learning_rate": 2.9534692694240433e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 8472
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.40625,
      "epoch": 0.40940278314650175,
      "grad_norm": 0.0,
      "learning_rate": 2.953227676845767e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8473
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.0,
      "epoch": 0.4094511016621569,
      "grad_norm": 0.0,
      "learning_rate": 2.9529860842674916e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8474
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.0625,
      "epoch": 0.40949942017781216,
      "grad_norm": 0.0,
      "learning_rate": 2.952744491689215e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8475
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.1875,
      "epoch": 0.40954773869346733,
      "grad_norm": 0.0,
      "learning_rate": 2.9525028991109393e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8476
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.78125,
      "epoch": 0.4095960572091225,
      "grad_norm": 0.0,
      "learning_rate": 2.952261306532663e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 8477
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.8125,
      "epoch": 0.40964437572477774,
      "grad_norm": 0.0,
      "learning_rate": 2.952019713954387e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8478
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.59375,
      "epoch": 0.4096926942404329,
      "grad_norm": 0.0,
      "learning_rate": 2.951778121376111e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8479
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.78125,
      "epoch": 0.40974101275608815,
      "grad_norm": 0.0,
      "learning_rate": 2.951536528797835e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8480
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.09375,
      "epoch": 0.40978933127174333,
      "grad_norm": 0.0,
      "learning_rate": 2.9512949362195595e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8481
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.09375,
      "epoch": 0.4098376497873985,
      "grad_norm": 0.0,
      "learning_rate": 2.951053343641283e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8482
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.5625,
      "epoch": 0.40988596830305374,
      "grad_norm": 0.0,
      "learning_rate": 2.950811751063007e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8483
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.65625,
      "epoch": 0.4099342868187089,
      "grad_norm": 0.0,
      "learning_rate": 2.9505701584847314e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8484
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.15625,
      "epoch": 0.40998260533436415,
      "grad_norm": 0.0,
      "learning_rate": 2.9503285659064555e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8485
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.09375,
      "epoch": 0.4100309238500193,
      "grad_norm": 0.0,
      "learning_rate": 2.950086973328179e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8486
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.71875,
      "epoch": 0.4100792423656745,
      "grad_norm": 0.0,
      "learning_rate": 2.9498453807499027e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8487
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.34375,
      "epoch": 0.41012756088132973,
      "grad_norm": 0.0,
      "learning_rate": 2.9496037881716274e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8488
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.71875,
      "epoch": 0.4101758793969849,
      "grad_norm": 0.0,
      "learning_rate": 2.949362195593351e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8489
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.59375,
      "epoch": 0.41022419791264014,
      "grad_norm": 0.0,
      "learning_rate": 2.949120603015075e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8490
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.09375,
      "epoch": 0.4102725164282953,
      "grad_norm": 0.0,
      "learning_rate": 2.948879010436799e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8491
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.53125,
      "epoch": 0.4103208349439505,
      "grad_norm": 0.0,
      "learning_rate": 2.9486374178585234e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8492
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.3125,
      "epoch": 0.41036915345960573,
      "grad_norm": 0.0,
      "learning_rate": 2.948395825280247e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 8493
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.53125,
      "epoch": 0.4104174719752609,
      "grad_norm": 0.0,
      "learning_rate": 2.9481542327019717e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8494
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.9375,
      "epoch": 0.41046579049091614,
      "grad_norm": 0.0,
      "learning_rate": 2.9479126401236953e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8495
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.65625,
      "epoch": 0.4105141090065713,
      "grad_norm": 0.0,
      "learning_rate": 2.9476710475454194e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8496
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.5,
      "epoch": 0.4105624275222265,
      "grad_norm": 0.0,
      "learning_rate": 2.947429454967143e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8497
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.75,
      "epoch": 0.4106107460378817,
      "grad_norm": 37.17944503518193,
      "learning_rate": 2.947187862388867e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.2715832591056824,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8498
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.78125,
      "epoch": 0.4106590645535369,
      "grad_norm": 0.0,
      "learning_rate": 2.9469462698105913e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8499
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.9375,
      "epoch": 0.41070738306919213,
      "grad_norm": 0.0,
      "learning_rate": 2.946704677232315e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8500
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.9375,
      "epoch": 0.4107557015848473,
      "grad_norm": 0.0,
      "learning_rate": 2.9464630846540396e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8501
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 71.65625,
      "epoch": 0.4108040201005025,
      "grad_norm": 0.0,
      "learning_rate": 2.946221492075763e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8502
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.21875,
      "epoch": 0.4108523386161577,
      "grad_norm": 0.0,
      "learning_rate": 2.9459798994974873e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8503
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.125,
      "epoch": 0.4109006571318129,
      "grad_norm": 0.0,
      "learning_rate": 2.9457383069192114e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8504
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.03125,
      "epoch": 0.41094897564746813,
      "grad_norm": 0.0,
      "learning_rate": 2.9454967143409356e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8505
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.40625,
      "epoch": 0.4109972941631233,
      "grad_norm": 0.0,
      "learning_rate": 2.945255121762659e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8506
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.65625,
      "epoch": 0.4110456126787785,
      "grad_norm": 0.0,
      "learning_rate": 2.9450135291843833e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8507
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.15625,
      "epoch": 0.4110939311944337,
      "grad_norm": 0.0,
      "learning_rate": 2.9447719366061075e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8508
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.53125,
      "epoch": 0.4111422497100889,
      "grad_norm": 0.0,
      "learning_rate": 2.944530344027831e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8509
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.0,
      "epoch": 0.4111905682257441,
      "grad_norm": 0.0,
      "learning_rate": 2.944288751449555e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8510
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.53125,
      "epoch": 0.4112388867413993,
      "grad_norm": 0.0,
      "learning_rate": 2.9440471588712793e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8511
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.6875,
      "epoch": 0.4112872052570545,
      "grad_norm": 0.0,
      "learning_rate": 2.9438055662930035e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8512
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.625,
      "epoch": 0.4113355237727097,
      "grad_norm": 0.0,
      "learning_rate": 2.943563973714727e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8513
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.875,
      "epoch": 0.4113838422883649,
      "grad_norm": 0.0,
      "learning_rate": 2.943322381136452e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 8514
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.53125,
      "epoch": 0.4114321608040201,
      "grad_norm": 0.0,
      "learning_rate": 2.9430807885581753e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8515
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.34375,
      "epoch": 0.4114804793196753,
      "grad_norm": 0.0,
      "learning_rate": 2.9428391959798995e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8516
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.28125,
      "epoch": 0.4115287978353305,
      "grad_norm": 0.0,
      "learning_rate": 2.942597603401623e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8517
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.21875,
      "epoch": 0.4115771163509857,
      "grad_norm": 0.0,
      "learning_rate": 2.942356010823348e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8518
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.78125,
      "epoch": 0.4116254348666409,
      "grad_norm": 0.0,
      "learning_rate": 2.9421144182450714e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8519
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.5625,
      "epoch": 0.4116737533822961,
      "grad_norm": 0.0,
      "learning_rate": 2.941872825666795e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8520
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.71875,
      "epoch": 0.4117220718979513,
      "grad_norm": 0.0,
      "learning_rate": 2.9416312330885196e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8521
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.3125,
      "epoch": 0.41177039041360647,
      "grad_norm": 0.0,
      "learning_rate": 2.941389640510243e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8522
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.5,
      "epoch": 0.4118187089292617,
      "grad_norm": 0.0,
      "learning_rate": 2.9411480479319674e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8523
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.875,
      "epoch": 0.4118670274449169,
      "grad_norm": 0.0,
      "learning_rate": 2.9409064553536915e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8524
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.4375,
      "epoch": 0.4119153459605721,
      "grad_norm": 0.0,
      "learning_rate": 2.9406648627754157e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8525
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.09375,
      "epoch": 0.4119636644762273,
      "grad_norm": 0.0,
      "learning_rate": 2.940423270197139e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8526
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.15625,
      "epoch": 0.41201198299188246,
      "grad_norm": 0.0,
      "learning_rate": 2.9401816776188634e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8527
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.21875,
      "epoch": 0.4120603015075377,
      "grad_norm": 0.0,
      "learning_rate": 2.9399400850405875e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8528
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.0625,
      "epoch": 0.4121086200231929,
      "grad_norm": 0.0,
      "learning_rate": 2.939698492462311e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8529
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.15625,
      "epoch": 0.4121569385388481,
      "grad_norm": 0.0,
      "learning_rate": 2.9394568998840353e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8530
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.125,
      "epoch": 0.4122052570545033,
      "grad_norm": 0.0,
      "learning_rate": 2.9392153073057594e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8531
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.625,
      "epoch": 0.41225357557015846,
      "grad_norm": 0.0,
      "learning_rate": 2.9389737147274835e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8532
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.0,
      "epoch": 0.4123018940858137,
      "grad_norm": 0.0,
      "learning_rate": 2.938732122149207e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8533
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.28125,
      "epoch": 0.41235021260146887,
      "grad_norm": 0.0,
      "learning_rate": 2.938490529570932e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8534
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.59375,
      "epoch": 0.4123985311171241,
      "grad_norm": 0.0,
      "learning_rate": 2.9382489369926554e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8535
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.96875,
      "epoch": 0.4124468496327793,
      "grad_norm": 0.0,
      "learning_rate": 2.9380073444143796e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8536
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.625,
      "epoch": 0.41249516814843445,
      "grad_norm": 0.0,
      "learning_rate": 2.937765751836103e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8537
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.875,
      "epoch": 0.4125434866640897,
      "grad_norm": 0.0,
      "learning_rate": 2.937524159257828e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8538
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.875,
      "epoch": 0.41259180517974486,
      "grad_norm": 0.0,
      "learning_rate": 2.9372825666795514e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 8539
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.96875,
      "epoch": 0.4126401236954001,
      "grad_norm": 0.0,
      "learning_rate": 2.937040974101275e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8540
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.0625,
      "epoch": 0.4126884422110553,
      "grad_norm": 0.0,
      "learning_rate": 2.9367993815229997e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8541
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.875,
      "epoch": 0.41273676072671045,
      "grad_norm": 0.0,
      "learning_rate": 2.9365577889447233e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8542
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.84375,
      "epoch": 0.4127850792423657,
      "grad_norm": 0.0,
      "learning_rate": 2.9363161963664475e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8543
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.375,
      "epoch": 0.41283339775802086,
      "grad_norm": 0.0,
      "learning_rate": 2.9360746037881716e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8544
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.75,
      "epoch": 0.4128817162736761,
      "grad_norm": 0.0,
      "learning_rate": 2.9358330112098957e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8545
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.0625,
      "epoch": 0.41293003478933127,
      "grad_norm": 0.0,
      "learning_rate": 2.9355914186316193e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8546
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.5625,
      "epoch": 0.41297835330498645,
      "grad_norm": 0.0,
      "learning_rate": 2.9353498260533435e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8547
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.15625,
      "epoch": 0.4130266718206417,
      "grad_norm": 0.0,
      "learning_rate": 2.9351082334750676e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8548
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.25,
      "epoch": 0.41307499033629685,
      "grad_norm": 0.0,
      "learning_rate": 2.934866640896791e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8549
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.8125,
      "epoch": 0.4131233088519521,
      "grad_norm": 0.0,
      "learning_rate": 2.9346250483185154e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 8550
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.09375,
      "epoch": 0.41317162736760726,
      "grad_norm": 0.0,
      "learning_rate": 2.9343834557402395e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8551
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.15625,
      "epoch": 0.41321994588326244,
      "grad_norm": 0.0,
      "learning_rate": 2.9341418631619636e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8552
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.1875,
      "epoch": 0.4132682643989177,
      "grad_norm": 0.0,
      "learning_rate": 2.933900270583687e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8553
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.375,
      "epoch": 0.41331658291457285,
      "grad_norm": 0.0,
      "learning_rate": 2.933658678005412e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 8554
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 74.96875,
      "epoch": 0.4133649014302281,
      "grad_norm": 31.051639290965408,
      "learning_rate": 2.9334170854271355e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8555
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 85.96875,
      "epoch": 0.41341321994588326,
      "grad_norm": 8.765958273436848,
      "learning_rate": 2.9331754928488596e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8556
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.78125,
      "epoch": 0.41346153846153844,
      "grad_norm": 0.0,
      "learning_rate": 2.932933900270583e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8557
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.96875,
      "epoch": 0.41350985697719367,
      "grad_norm": 0.0,
      "learning_rate": 2.932692307692308e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8558
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.90625,
      "epoch": 0.41355817549284885,
      "grad_norm": 0.0,
      "learning_rate": 2.9324507151140315e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8559
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.03125,
      "epoch": 0.4136064940085041,
      "grad_norm": 0.0,
      "learning_rate": 2.932209122535755e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8560
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.875,
      "epoch": 0.41365481252415925,
      "grad_norm": 0.0,
      "learning_rate": 2.93196752995748e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8561
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.65625,
      "epoch": 0.41370313103981443,
      "grad_norm": 0.0,
      "learning_rate": 2.9317259373792034e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8562
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.59375,
      "epoch": 0.41375144955546966,
      "grad_norm": 0.0,
      "learning_rate": 2.9314843448009275e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8563
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.21875,
      "epoch": 0.41379976807112484,
      "grad_norm": 0.0,
      "learning_rate": 2.9312427522226517e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8564
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.21875,
      "epoch": 0.4138480865867801,
      "grad_norm": 0.0,
      "learning_rate": 2.931001159644376e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8565
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.125,
      "epoch": 0.41389640510243525,
      "grad_norm": 0.0,
      "learning_rate": 2.9307595670660994e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8566
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.03125,
      "epoch": 0.4139447236180904,
      "grad_norm": 0.0,
      "learning_rate": 2.9305179744878236e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8567
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.28125,
      "epoch": 0.41399304213374566,
      "grad_norm": 0.0,
      "learning_rate": 2.9302763819095477e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8568
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.5625,
      "epoch": 0.41404136064940084,
      "grad_norm": 0.0,
      "learning_rate": 2.930034789331272e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8569
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.03125,
      "epoch": 0.41408967916505607,
      "grad_norm": 0.0,
      "learning_rate": 2.9297931967529954e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 8570
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.28125,
      "epoch": 0.41413799768071125,
      "grad_norm": 0.0,
      "learning_rate": 2.9295516041747196e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8571
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.5,
      "epoch": 0.4141863161963664,
      "grad_norm": 0.0,
      "learning_rate": 2.9293100115964437e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8572
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.75,
      "epoch": 0.41423463471202165,
      "grad_norm": 0.0,
      "learning_rate": 2.9290684190181673e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 8573
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.875,
      "epoch": 0.41428295322767683,
      "grad_norm": 0.0,
      "learning_rate": 2.928826826439892e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8574
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.40625,
      "epoch": 0.41433127174333206,
      "grad_norm": 0.0,
      "learning_rate": 2.9285852338616156e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8575
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.0,
      "epoch": 0.41437959025898724,
      "grad_norm": 0.0,
      "learning_rate": 2.9283436412833397e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8576
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.125,
      "epoch": 0.4144279087746424,
      "grad_norm": 0.0,
      "learning_rate": 2.9281020487050633e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8577
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.84375,
      "epoch": 0.41447622729029765,
      "grad_norm": 0.0,
      "learning_rate": 2.927860456126788e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8578
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.15625,
      "epoch": 0.4145245458059528,
      "grad_norm": 0.0,
      "learning_rate": 2.9276188635485116e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8579
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.75,
      "epoch": 0.41457286432160806,
      "grad_norm": 0.0,
      "learning_rate": 2.927377270970235e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8580
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.71875,
      "epoch": 0.41462118283726324,
      "grad_norm": 0.0,
      "learning_rate": 2.92713567839196e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8581
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 87.40625,
      "epoch": 0.4146695013529184,
      "grad_norm": 0.0,
      "learning_rate": 2.9268940858136835e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 8582
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.625,
      "epoch": 0.41471781986857364,
      "grad_norm": 0.0,
      "learning_rate": 2.9266524932354076e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8583
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.6875,
      "epoch": 0.4147661383842288,
      "grad_norm": 0.0,
      "learning_rate": 2.926410900657132e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8584
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.3125,
      "epoch": 0.41481445689988405,
      "grad_norm": 0.0,
      "learning_rate": 2.926169308078856e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8585
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.9375,
      "epoch": 0.41486277541553923,
      "grad_norm": 0.0,
      "learning_rate": 2.9259277155005795e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8586
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.5,
      "epoch": 0.4149110939311944,
      "grad_norm": 0.0,
      "learning_rate": 2.9256861229223036e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8587
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 78.09375,
      "epoch": 0.41495941244684964,
      "grad_norm": 0.0,
      "learning_rate": 2.925444530344028e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 8588
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.75,
      "epoch": 0.4150077309625048,
      "grad_norm": 0.0,
      "learning_rate": 2.925202937765752e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8589
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.0,
      "epoch": 0.41505604947816005,
      "grad_norm": 0.0,
      "learning_rate": 2.9249613451874755e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8590
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.65625,
      "epoch": 0.4151043679938152,
      "grad_norm": 0.0,
      "learning_rate": 2.9247197526091996e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8591
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.53125,
      "epoch": 0.4151526865094704,
      "grad_norm": 0.0,
      "learning_rate": 2.924478160030924e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8592
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.8125,
      "epoch": 0.41520100502512564,
      "grad_norm": 0.0,
      "learning_rate": 2.9242365674526474e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8593
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.5625,
      "epoch": 0.4152493235407808,
      "grad_norm": 0.0,
      "learning_rate": 2.923994974874372e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8594
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.46875,
      "epoch": 0.41529764205643604,
      "grad_norm": 0.0,
      "learning_rate": 2.9237533822960957e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8595
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.03125,
      "epoch": 0.4153459605720912,
      "grad_norm": 0.0,
      "learning_rate": 2.92351178971782e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8596
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.40625,
      "epoch": 0.4153942790877464,
      "grad_norm": 0.0,
      "learning_rate": 2.9232701971395434e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8597
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.0,
      "epoch": 0.41544259760340163,
      "grad_norm": 0.0,
      "learning_rate": 2.923028604561268e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8598
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.71875,
      "epoch": 0.4154909161190568,
      "grad_norm": 0.0,
      "learning_rate": 2.9227870119829917e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8599
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.46875,
      "epoch": 0.41553923463471204,
      "grad_norm": 0.0,
      "learning_rate": 2.9225454194047153e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8600
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.46875,
      "epoch": 0.4155875531503672,
      "grad_norm": 0.0,
      "learning_rate": 2.92230382682644e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8601
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.375,
      "epoch": 0.4156358716660224,
      "grad_norm": 0.0,
      "learning_rate": 2.9220622342481636e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8602
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.3125,
      "epoch": 0.4156841901816776,
      "grad_norm": 0.0,
      "learning_rate": 2.9218206416698877e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8603
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.46875,
      "epoch": 0.4157325086973328,
      "grad_norm": 0.0,
      "learning_rate": 2.921579049091612e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8604
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.96875,
      "epoch": 0.41578082721298804,
      "grad_norm": 0.0,
      "learning_rate": 2.921337456513336e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 8605
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.15625,
      "epoch": 0.4158291457286432,
      "grad_norm": 0.0,
      "learning_rate": 2.9210958639350596e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8606
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.625,
      "epoch": 0.4158774642442984,
      "grad_norm": 0.0,
      "learning_rate": 2.9208542713567837e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8607
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.71875,
      "epoch": 0.4159257827599536,
      "grad_norm": 0.0,
      "learning_rate": 2.920612678778508e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8608
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.28125,
      "epoch": 0.4159741012756088,
      "grad_norm": 0.0,
      "learning_rate": 2.920371086200232e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8609
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.4375,
      "epoch": 0.41602241979126403,
      "grad_norm": 0.0,
      "learning_rate": 2.9201294936219556e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8610
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.71875,
      "epoch": 0.4160707383069192,
      "grad_norm": 0.0,
      "learning_rate": 2.9198879010436797e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8611
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.75,
      "epoch": 0.4161190568225744,
      "grad_norm": 0.0,
      "learning_rate": 2.919646308465404e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8612
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.84375,
      "epoch": 0.4161673753382296,
      "grad_norm": 0.0,
      "learning_rate": 2.9194047158871275e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8613
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.4375,
      "epoch": 0.4162156938538848,
      "grad_norm": 0.0,
      "learning_rate": 2.919163123308852e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8614
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.40625,
      "epoch": 0.41626401236954,
      "grad_norm": 0.0,
      "learning_rate": 2.918921530730576e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8615
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.84375,
      "epoch": 0.4163123308851952,
      "grad_norm": 0.0,
      "learning_rate": 2.9186799381523e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8616
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.6875,
      "epoch": 0.4163606494008504,
      "grad_norm": 0.0,
      "learning_rate": 2.9184383455740235e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8617
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.875,
      "epoch": 0.4164089679165056,
      "grad_norm": 0.0,
      "learning_rate": 2.918196752995748e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 8618
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.59375,
      "epoch": 0.4164572864321608,
      "grad_norm": 0.0,
      "learning_rate": 2.917955160417472e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8619
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.25,
      "epoch": 0.416505604947816,
      "grad_norm": 0.0,
      "learning_rate": 2.917713567839196e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8620
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.28125,
      "epoch": 0.4165539234634712,
      "grad_norm": 0.0,
      "learning_rate": 2.91747197526092e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8621
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.21875,
      "epoch": 0.4166022419791264,
      "grad_norm": 0.0,
      "learning_rate": 2.9172303826826436e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8622
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.96875,
      "epoch": 0.4166505604947816,
      "grad_norm": 0.0,
      "learning_rate": 2.916988790104368e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8623
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.21875,
      "epoch": 0.4166988790104368,
      "grad_norm": 0.0,
      "learning_rate": 2.916747197526092e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8624
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.1875,
      "epoch": 0.416747197526092,
      "grad_norm": 0.0,
      "learning_rate": 2.916505604947816e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8625
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.4375,
      "epoch": 0.4167955160417472,
      "grad_norm": 0.0,
      "learning_rate": 2.9162640123695397e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 8626
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.34375,
      "epoch": 0.41684383455740237,
      "grad_norm": 0.0,
      "learning_rate": 2.916022419791264e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8627
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.96875,
      "epoch": 0.4168921530730576,
      "grad_norm": 0.0,
      "learning_rate": 2.915780827212988e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8628
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.03125,
      "epoch": 0.4169404715887128,
      "grad_norm": 0.0,
      "learning_rate": 2.915539234634712e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8629
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.3125,
      "epoch": 0.416988790104368,
      "grad_norm": 0.0,
      "learning_rate": 2.9152976420564357e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8630
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.40625,
      "epoch": 0.4170371086200232,
      "grad_norm": 0.0,
      "learning_rate": 2.9150560494781603e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8631
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.03125,
      "epoch": 0.41708542713567837,
      "grad_norm": 0.0,
      "learning_rate": 2.914814456899884e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8632
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.8125,
      "epoch": 0.4171337456513336,
      "grad_norm": 0.0,
      "learning_rate": 2.9145728643216075e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8633
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.4375,
      "epoch": 0.4171820641669888,
      "grad_norm": 0.0,
      "learning_rate": 2.914331271743332e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8634
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.0,
      "epoch": 0.417230382682644,
      "grad_norm": 0.0,
      "learning_rate": 2.914089679165056e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8635
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 79.28125,
      "epoch": 0.4172787011982992,
      "grad_norm": 0.0,
      "learning_rate": 2.91384808658678e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 8636
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.4375,
      "epoch": 0.41732701971395436,
      "grad_norm": 0.0,
      "learning_rate": 2.9136064940085036e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8637
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.59375,
      "epoch": 0.4173753382296096,
      "grad_norm": 0.0,
      "learning_rate": 2.913364901430228e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8638
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.25,
      "epoch": 0.41742365674526477,
      "grad_norm": 0.0,
      "learning_rate": 2.913123308851952e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8639
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.5625,
      "epoch": 0.41747197526092,
      "grad_norm": 0.0,
      "learning_rate": 2.912881716273676e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8640
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.78125,
      "epoch": 0.4175202937765752,
      "grad_norm": 0.0,
      "learning_rate": 2.9126401236954e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8641
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.25,
      "epoch": 0.41756861229223036,
      "grad_norm": 0.0,
      "learning_rate": 2.9123985311171237e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8642
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.375,
      "epoch": 0.4176169308078856,
      "grad_norm": 0.0,
      "learning_rate": 2.912156938538848e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8643
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.40625,
      "epoch": 0.41766524932354077,
      "grad_norm": 0.0,
      "learning_rate": 2.911915345960572e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8644
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.0625,
      "epoch": 0.417713567839196,
      "grad_norm": 0.0,
      "learning_rate": 2.911673753382296e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8645
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.4375,
      "epoch": 0.4177618863548512,
      "grad_norm": 0.0,
      "learning_rate": 2.9114321608040197e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8646
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.09375,
      "epoch": 0.41781020487050635,
      "grad_norm": 0.0,
      "learning_rate": 2.911190568225744e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8647
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.21875,
      "epoch": 0.4178585233861616,
      "grad_norm": 0.0,
      "learning_rate": 2.910948975647468e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8648
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.1875,
      "epoch": 0.41790684190181676,
      "grad_norm": 0.0,
      "learning_rate": 2.910707383069192e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8649
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 60.84375,
      "epoch": 0.417955160417472,
      "grad_norm": 12.476894395509058,
      "learning_rate": 2.910465790490916e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8650
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.4375,
      "epoch": 0.41800347893312717,
      "grad_norm": 0.0,
      "learning_rate": 2.9102241979126404e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8651
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.78125,
      "epoch": 0.41805179744878235,
      "grad_norm": 0.0,
      "learning_rate": 2.909982605334364e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8652
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.53125,
      "epoch": 0.4181001159644376,
      "grad_norm": 0.0,
      "learning_rate": 2.9097410127560876e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8653
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.0625,
      "epoch": 0.41814843448009276,
      "grad_norm": 0.0,
      "learning_rate": 2.9094994201778123e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8654
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.46875,
      "epoch": 0.418196752995748,
      "grad_norm": 0.0,
      "learning_rate": 2.909257827599536e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8655
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.34375,
      "epoch": 0.41824507151140317,
      "grad_norm": 0.0,
      "learning_rate": 2.90901623502126e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8656
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.25,
      "epoch": 0.41829339002705834,
      "grad_norm": 0.0,
      "learning_rate": 2.9087746424429836e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8657
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.59375,
      "epoch": 0.4183417085427136,
      "grad_norm": 0.0,
      "learning_rate": 2.9085330498647083e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8658
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.40625,
      "epoch": 0.41839002705836875,
      "grad_norm": 0.0,
      "learning_rate": 2.908291457286432e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8659
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.0625,
      "epoch": 0.418438345574024,
      "grad_norm": 0.0,
      "learning_rate": 2.908049864708156e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8660
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.125,
      "epoch": 0.41848666408967916,
      "grad_norm": 0.0,
      "learning_rate": 2.90780827212988e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8661
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.84375,
      "epoch": 0.41853498260533434,
      "grad_norm": 0.0,
      "learning_rate": 2.9075666795516043e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8662
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.25,
      "epoch": 0.41858330112098957,
      "grad_norm": 0.0,
      "learning_rate": 2.907325086973328e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8663
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.59375,
      "epoch": 0.41863161963664475,
      "grad_norm": 0.0,
      "learning_rate": 2.907083494395052e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8664
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.3125,
      "epoch": 0.4186799381523,
      "grad_norm": 0.0,
      "learning_rate": 2.906841901816776e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8665
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.78125,
      "epoch": 0.41872825666795516,
      "grad_norm": 0.0,
      "learning_rate": 2.9066003092385e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8666
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.375,
      "epoch": 0.41877657518361033,
      "grad_norm": 0.0,
      "learning_rate": 2.906358716660224e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8667
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.78125,
      "epoch": 0.41882489369926557,
      "grad_norm": 0.0,
      "learning_rate": 2.906117124081948e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8668
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.46875,
      "epoch": 0.41887321221492074,
      "grad_norm": 0.0,
      "learning_rate": 2.905875531503672e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8669
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.6875,
      "epoch": 0.418921530730576,
      "grad_norm": 0.0,
      "learning_rate": 2.905633938925396e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8670
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.78125,
      "epoch": 0.41896984924623115,
      "grad_norm": 0.0,
      "learning_rate": 2.9053923463471205e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8671
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.84375,
      "epoch": 0.41901816776188633,
      "grad_norm": 0.0,
      "learning_rate": 2.905150753768844e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 8672
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.0,
      "epoch": 0.41906648627754156,
      "grad_norm": 0.0,
      "learning_rate": 2.9049091611905677e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8673
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.5,
      "epoch": 0.41911480479319674,
      "grad_norm": 0.0,
      "learning_rate": 2.9046675686122924e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 8674
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.8125,
      "epoch": 0.41916312330885197,
      "grad_norm": 0.0,
      "learning_rate": 2.904425976034016e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8675
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.96875,
      "epoch": 0.41921144182450715,
      "grad_norm": 0.0,
      "learning_rate": 2.90418438345574e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8676
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.1875,
      "epoch": 0.4192597603401623,
      "grad_norm": 0.0,
      "learning_rate": 2.9039427908774637e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8677
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.875,
      "epoch": 0.41930807885581756,
      "grad_norm": 0.0,
      "learning_rate": 2.9037011982991884e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8678
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.09375,
      "epoch": 0.41935639737147273,
      "grad_norm": 0.0,
      "learning_rate": 2.903459605720912e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8679
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.125,
      "epoch": 0.41940471588712797,
      "grad_norm": 0.0,
      "learning_rate": 2.903218013142636e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8680
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.3125,
      "epoch": 0.41945303440278314,
      "grad_norm": 10.725872000557624,
      "learning_rate": 2.9029764205643603e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8681
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.0,
      "epoch": 0.4195013529184383,
      "grad_norm": 0.0,
      "learning_rate": 2.9027348279860844e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8682
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.25,
      "epoch": 0.41954967143409355,
      "grad_norm": 0.0,
      "learning_rate": 2.902493235407808e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8683
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.34375,
      "epoch": 0.41959798994974873,
      "grad_norm": 0.0,
      "learning_rate": 2.902251642829532e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8684
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.96875,
      "epoch": 0.41964630846540396,
      "grad_norm": 0.0,
      "learning_rate": 2.9020100502512563e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8685
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.125,
      "epoch": 0.41969462698105914,
      "grad_norm": 0.0,
      "learning_rate": 2.90176845767298e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8686
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.21875,
      "epoch": 0.4197429454967143,
      "grad_norm": 0.0,
      "learning_rate": 2.901526865094704e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8687
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.78125,
      "epoch": 0.41979126401236955,
      "grad_norm": 0.0,
      "learning_rate": 2.901285272516428e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2630178928375244,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8688
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.4375,
      "epoch": 0.4198395825280247,
      "grad_norm": 0.0,
      "learning_rate": 2.9010436799381523e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 8689
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.71875,
      "epoch": 0.41988790104367996,
      "grad_norm": 0.0,
      "learning_rate": 2.900802087359876e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 8690
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.1875,
      "epoch": 0.41993621955933513,
      "grad_norm": 0.0,
      "learning_rate": 2.9005604947816006e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8691
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.3125,
      "epoch": 0.4199845380749903,
      "grad_norm": 0.0,
      "learning_rate": 2.900318902203324e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8692
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.4375,
      "epoch": 0.42003285659064554,
      "grad_norm": 0.0,
      "learning_rate": 2.900077309625048e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8693
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.9375,
      "epoch": 0.4200811751063007,
      "grad_norm": 0.0,
      "learning_rate": 2.8998357170467725e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8694
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.6875,
      "epoch": 0.42012949362195595,
      "grad_norm": 0.0,
      "learning_rate": 2.899594124468496e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8695
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.34375,
      "epoch": 0.42017781213761113,
      "grad_norm": 0.0,
      "learning_rate": 2.89935253189022e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8696
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.71875,
      "epoch": 0.4202261306532663,
      "grad_norm": 0.0,
      "learning_rate": 2.899110939311944e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8697
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.90625,
      "epoch": 0.42027444916892154,
      "grad_norm": 0.0,
      "learning_rate": 2.8988693467336685e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8698
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.21875,
      "epoch": 0.4203227676845767,
      "grad_norm": 0.0,
      "learning_rate": 2.898627754155392e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8699
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.4375,
      "epoch": 0.42037108620023195,
      "grad_norm": 0.0,
      "learning_rate": 2.898386161577116e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 8700
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.90625,
      "epoch": 0.4204194047158871,
      "grad_norm": 0.0,
      "learning_rate": 2.8981445689988403e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8701
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.1875,
      "epoch": 0.4204677232315423,
      "grad_norm": 0.0,
      "learning_rate": 2.8979029764205645e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8702
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.59375,
      "epoch": 0.42051604174719753,
      "grad_norm": 0.0,
      "learning_rate": 2.897661383842288e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8703
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.625,
      "epoch": 0.4205643602628527,
      "grad_norm": 0.0,
      "learning_rate": 2.897419791264012e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8704
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.21875,
      "epoch": 0.42061267877850794,
      "grad_norm": 0.0,
      "learning_rate": 2.8971781986857364e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8705
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.625,
      "epoch": 0.4206609972941631,
      "grad_norm": 0.0,
      "learning_rate": 2.89693660610746e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8706
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.5,
      "epoch": 0.4207093158098183,
      "grad_norm": 0.0,
      "learning_rate": 2.896695013529184e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8707
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.6875,
      "epoch": 0.42075763432547353,
      "grad_norm": 0.0,
      "learning_rate": 2.896453420950908e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8708
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.125,
      "epoch": 0.4208059528411287,
      "grad_norm": 0.0,
      "learning_rate": 2.8962118283726324e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 8709
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.5,
      "epoch": 0.42085427135678394,
      "grad_norm": 0.0,
      "learning_rate": 2.895970235794356e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8710
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.65625,
      "epoch": 0.4209025898724391,
      "grad_norm": 0.0,
      "learning_rate": 2.8957286432160807e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8711
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.34375,
      "epoch": 0.4209509083880943,
      "grad_norm": 0.0,
      "learning_rate": 2.895487050637804e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8712
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.71875,
      "epoch": 0.4209992269037495,
      "grad_norm": 0.0,
      "learning_rate": 2.8952454580595284e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 8713
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.8125,
      "epoch": 0.4210475454194047,
      "grad_norm": 0.0,
      "learning_rate": 2.8950038654812525e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8714
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.78125,
      "epoch": 0.42109586393505993,
      "grad_norm": 0.0,
      "learning_rate": 2.894762272902976e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8715
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.8125,
      "epoch": 0.4211441824507151,
      "grad_norm": 0.0,
      "learning_rate": 2.8945206803247003e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8716
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.65625,
      "epoch": 0.4211925009663703,
      "grad_norm": 0.0,
      "learning_rate": 2.894279087746424e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8717
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.5625,
      "epoch": 0.4212408194820255,
      "grad_norm": 0.0,
      "learning_rate": 2.8940374951681485e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8718
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.0625,
      "epoch": 0.4212891379976807,
      "grad_norm": 0.0,
      "learning_rate": 2.893795902589872e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8719
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.4375,
      "epoch": 0.42133745651333593,
      "grad_norm": 0.0,
      "learning_rate": 2.8935543100115963e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 8720
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.34375,
      "epoch": 0.4213857750289911,
      "grad_norm": 0.0,
      "learning_rate": 2.8933127174333204e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8721
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.96875,
      "epoch": 0.4214340935446463,
      "grad_norm": 0.0,
      "learning_rate": 2.8930711248550446e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8722
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.78125,
      "epoch": 0.4214824120603015,
      "grad_norm": 0.0,
      "learning_rate": 2.892829532276768e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2130674123764038,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8723
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.59375,
      "epoch": 0.4215307305759567,
      "grad_norm": 0.0,
      "learning_rate": 2.892587939698493e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8724
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.03125,
      "epoch": 0.4215790490916119,
      "grad_norm": 0.0,
      "learning_rate": 2.8923463471202164e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8725
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.65625,
      "epoch": 0.4216273676072671,
      "grad_norm": 0.0,
      "learning_rate": 2.89210475454194e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8726
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.90625,
      "epoch": 0.4216756861229223,
      "grad_norm": 0.0,
      "learning_rate": 2.891863161963664e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8727
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.78125,
      "epoch": 0.4217240046385775,
      "grad_norm": 0.0,
      "learning_rate": 2.8916215693853883e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8728
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.0625,
      "epoch": 0.4217723231542327,
      "grad_norm": 0.0,
      "learning_rate": 2.8913799768071125e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8729
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.1875,
      "epoch": 0.4218206416698879,
      "grad_norm": 0.0,
      "learning_rate": 2.891138384228836e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8730
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.0625,
      "epoch": 0.4218689601855431,
      "grad_norm": 0.0,
      "learning_rate": 2.8908967916505607e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8731
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.71875,
      "epoch": 0.4219172787011983,
      "grad_norm": 0.0,
      "learning_rate": 2.8906551990722843e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8732
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.46875,
      "epoch": 0.4219655972168535,
      "grad_norm": 0.0,
      "learning_rate": 2.8904136064940085e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8733
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.90625,
      "epoch": 0.4220139157325087,
      "grad_norm": 0.0,
      "learning_rate": 2.8901720139157326e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8734
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.71875,
      "epoch": 0.4220622342481639,
      "grad_norm": 0.0,
      "learning_rate": 2.889930421337456e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8735
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.78125,
      "epoch": 0.4221105527638191,
      "grad_norm": 0.0,
      "learning_rate": 2.8896888287591804e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8736
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.03125,
      "epoch": 0.42215887127947427,
      "grad_norm": 0.0,
      "learning_rate": 2.889447236180904e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8737
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.96875,
      "epoch": 0.4222071897951295,
      "grad_norm": 0.0,
      "learning_rate": 2.8892056436026286e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8738
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.46875,
      "epoch": 0.4222555083107847,
      "grad_norm": 0.0,
      "learning_rate": 2.888964051024352e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8739
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.875,
      "epoch": 0.4223038268264399,
      "grad_norm": 0.0,
      "learning_rate": 2.8887224584460764e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8740
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.53125,
      "epoch": 0.4223521453420951,
      "grad_norm": 0.0,
      "learning_rate": 2.8884808658678005e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8741
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.34375,
      "epoch": 0.42240046385775026,
      "grad_norm": 0.0,
      "learning_rate": 2.8882392732895246e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8742
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.59375,
      "epoch": 0.4224487823734055,
      "grad_norm": 0.0,
      "learning_rate": 2.887997680711248e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8743
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.1875,
      "epoch": 0.4224971008890607,
      "grad_norm": 0.0,
      "learning_rate": 2.887756088132973e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8744
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.0,
      "epoch": 0.4225454194047159,
      "grad_norm": 0.0,
      "learning_rate": 2.8875144955546965e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8745
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.34375,
      "epoch": 0.4225937379203711,
      "grad_norm": 0.0,
      "learning_rate": 2.88727290297642e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8746
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.71875,
      "epoch": 0.42264205643602626,
      "grad_norm": 0.0,
      "learning_rate": 2.887031310398144e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8747
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.71875,
      "epoch": 0.4226903749516815,
      "grad_norm": 0.0,
      "learning_rate": 2.8867897178198684e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8748
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.5,
      "epoch": 0.42273869346733667,
      "grad_norm": 0.0,
      "learning_rate": 2.8865481252415925e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8749
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.78125,
      "epoch": 0.4227870119829919,
      "grad_norm": 0.0,
      "learning_rate": 2.886306532663316e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 8750
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.84375,
      "epoch": 0.4228353304986471,
      "grad_norm": 0.0,
      "learning_rate": 2.886064940085041e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8751
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.25,
      "epoch": 0.42288364901430225,
      "grad_norm": 0.0,
      "learning_rate": 2.8858233475067644e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8752
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.875,
      "epoch": 0.4229319675299575,
      "grad_norm": 0.0,
      "learning_rate": 2.8855817549284886e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8753
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.46875,
      "epoch": 0.42298028604561266,
      "grad_norm": 0.0,
      "learning_rate": 2.8853401623502127e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8754
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.03125,
      "epoch": 0.4230286045612679,
      "grad_norm": 0.0,
      "learning_rate": 2.8850985697719363e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 8755
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 80.9375,
      "epoch": 0.4230769230769231,
      "grad_norm": 10.376265734942667,
      "learning_rate": 2.8848569771936604e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8756
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.71875,
      "epoch": 0.42312524159257825,
      "grad_norm": 0.0,
      "learning_rate": 2.884615384615384e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8757
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.5,
      "epoch": 0.4231735601082335,
      "grad_norm": 0.0,
      "learning_rate": 2.8843737920371087e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8758
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.25,
      "epoch": 0.42322187862388866,
      "grad_norm": 0.0,
      "learning_rate": 2.8841321994588323e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8759
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 81.125,
      "epoch": 0.4232701971395439,
      "grad_norm": 33.69984693838328,
      "learning_rate": 2.8838906068805564e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8760
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.65625,
      "epoch": 0.42331851565519907,
      "grad_norm": 0.0,
      "learning_rate": 2.8836490143022806e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8761
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.0625,
      "epoch": 0.42336683417085424,
      "grad_norm": 0.0,
      "learning_rate": 2.8834074217240047e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8762
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.4375,
      "epoch": 0.4234151526865095,
      "grad_norm": 0.0,
      "learning_rate": 2.8831658291457283e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8763
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.3125,
      "epoch": 0.42346347120216465,
      "grad_norm": 0.0,
      "learning_rate": 2.882924236567453e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8764
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.0,
      "epoch": 0.4235117897178199,
      "grad_norm": 0.0,
      "learning_rate": 2.8826826439891766e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8765
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.46875,
      "epoch": 0.42356010823347506,
      "grad_norm": 0.0,
      "learning_rate": 2.8824410514109e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8766
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.125,
      "epoch": 0.42360842674913024,
      "grad_norm": 0.0,
      "learning_rate": 2.8821994588326243e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8767
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.0625,
      "epoch": 0.4236567452647855,
      "grad_norm": 0.0,
      "learning_rate": 2.8819578662543485e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8768
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.1875,
      "epoch": 0.42370506378044065,
      "grad_norm": 0.0,
      "learning_rate": 2.8817162736760726e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8769
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.3125,
      "epoch": 0.4237533822960959,
      "grad_norm": 0.0,
      "learning_rate": 2.881474681097796e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8770
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.15625,
      "epoch": 0.42380170081175106,
      "grad_norm": 0.0,
      "learning_rate": 2.881233088519521e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8771
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.0,
      "epoch": 0.42385001932740624,
      "grad_norm": 0.0,
      "learning_rate": 2.8809914959412445e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8772
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.65625,
      "epoch": 0.42389833784306147,
      "grad_norm": 0.0,
      "learning_rate": 2.8807499033629686e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8773
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.1875,
      "epoch": 0.42394665635871664,
      "grad_norm": 0.0,
      "learning_rate": 2.880508310784693e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8774
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.6875,
      "epoch": 0.4239949748743719,
      "grad_norm": 0.0,
      "learning_rate": 2.880266718206417e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8775
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.28125,
      "epoch": 0.42404329339002705,
      "grad_norm": 0.0,
      "learning_rate": 2.8800251256281405e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.19727617502212524,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8776
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.21875,
      "epoch": 0.42409161190568223,
      "grad_norm": 0.0,
      "learning_rate": 2.879783533049864e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 8777
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.03125,
      "epoch": 0.42413993042133746,
      "grad_norm": 0.0,
      "learning_rate": 2.879541940471589e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 8778
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.125,
      "epoch": 0.42418824893699264,
      "grad_norm": 0.0,
      "learning_rate": 2.8793003478933124e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8779
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.71875,
      "epoch": 0.4242365674526479,
      "grad_norm": 0.0,
      "learning_rate": 2.8790587553150365e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8780
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.0,
      "epoch": 0.42428488596830305,
      "grad_norm": 0.0,
      "learning_rate": 2.8788171627367607e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8781
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.21875,
      "epoch": 0.4243332044839582,
      "grad_norm": 0.0,
      "learning_rate": 2.878575570158485e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8782
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.09375,
      "epoch": 0.42438152299961346,
      "grad_norm": 0.0,
      "learning_rate": 2.8783339775802084e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8783
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.25,
      "epoch": 0.42442984151526864,
      "grad_norm": 0.0,
      "learning_rate": 2.878092385001933e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8784
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.375,
      "epoch": 0.42447816003092387,
      "grad_norm": 0.0,
      "learning_rate": 2.8778507924236567e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8785
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.625,
      "epoch": 0.42452647854657904,
      "grad_norm": 0.0,
      "learning_rate": 2.8776091998453803e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8786
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.84375,
      "epoch": 0.4245747970622342,
      "grad_norm": 0.0,
      "learning_rate": 2.8773676072671044e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8787
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.46875,
      "epoch": 0.42462311557788945,
      "grad_norm": 0.0,
      "learning_rate": 2.8771260146888286e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8788
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.5,
      "epoch": 0.42467143409354463,
      "grad_norm": 0.0,
      "learning_rate": 2.8768844221105527e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8789
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.1875,
      "epoch": 0.42471975260919986,
      "grad_norm": 0.0,
      "learning_rate": 2.8766428295322763e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8790
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.03125,
      "epoch": 0.42476807112485504,
      "grad_norm": 0.0,
      "learning_rate": 2.876401236954001e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8791
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.3125,
      "epoch": 0.4248163896405102,
      "grad_norm": 0.0,
      "learning_rate": 2.8761596443757246e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8792
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.71875,
      "epoch": 0.42486470815616545,
      "grad_norm": 0.0,
      "learning_rate": 2.8759180517974487e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8793
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.0625,
      "epoch": 0.4249130266718206,
      "grad_norm": 0.0,
      "learning_rate": 2.875676459219173e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8794
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 91.46875,
      "epoch": 0.42496134518747586,
      "grad_norm": 19.642992711118275,
      "learning_rate": 2.875434866640897e-07,
      "loss": -0.0,
      "reward": 2.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8795
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.78125,
      "epoch": 0.42500966370313104,
      "grad_norm": 0.0,
      "learning_rate": 2.8751932740626206e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8796
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.1875,
      "epoch": 0.4250579822187862,
      "grad_norm": 0.0,
      "learning_rate": 2.874951681484344e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8797
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.1875,
      "epoch": 0.42510630073444144,
      "grad_norm": 0.0,
      "learning_rate": 2.874710088906069e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 8798
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.46875,
      "epoch": 0.4251546192500966,
      "grad_norm": 0.0,
      "learning_rate": 2.8744684963277925e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8799
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.75,
      "epoch": 0.42520293776575185,
      "grad_norm": 0.0,
      "learning_rate": 2.8742269037495166e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8800
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.90625,
      "epoch": 0.42525125628140703,
      "grad_norm": 0.0,
      "learning_rate": 2.873985311171241e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8801
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.75,
      "epoch": 0.4252995747970622,
      "grad_norm": 0.0,
      "learning_rate": 2.873743718592965e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8802
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.25,
      "epoch": 0.42534789331271744,
      "grad_norm": 0.0,
      "learning_rate": 2.8735021260146885e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8803
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.65625,
      "epoch": 0.4253962118283726,
      "grad_norm": 0.0,
      "learning_rate": 2.873260533436413e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8804
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.8125,
      "epoch": 0.42544453034402785,
      "grad_norm": 0.0,
      "learning_rate": 2.873018940858137e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8805
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.03125,
      "epoch": 0.425492848859683,
      "grad_norm": 0.0,
      "learning_rate": 2.8727773482798604e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8806
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.71875,
      "epoch": 0.4255411673753382,
      "grad_norm": 0.0,
      "learning_rate": 2.8725357557015845e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8807
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.9375,
      "epoch": 0.42558948589099344,
      "grad_norm": 0.0,
      "learning_rate": 2.8722941631233086e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 8808
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.875,
      "epoch": 0.4256378044066486,
      "grad_norm": 0.0,
      "learning_rate": 2.872052570545033e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8809
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.9375,
      "epoch": 0.42568612292230384,
      "grad_norm": 0.0,
      "learning_rate": 2.8718109779667564e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8810
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.09375,
      "epoch": 0.425734441437959,
      "grad_norm": 0.0,
      "learning_rate": 2.871569385388481e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8811
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.59375,
      "epoch": 0.4257827599536142,
      "grad_norm": 0.0,
      "learning_rate": 2.8713277928102047e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8812
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.5625,
      "epoch": 0.42583107846926943,
      "grad_norm": 0.0,
      "learning_rate": 2.871086200231929e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8813
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.84375,
      "epoch": 0.4258793969849246,
      "grad_norm": 0.0,
      "learning_rate": 2.870844607653653e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8814
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.5625,
      "epoch": 0.42592771550057984,
      "grad_norm": 0.0,
      "learning_rate": 2.870603015075377e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8815
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.65625,
      "epoch": 0.425976034016235,
      "grad_norm": 0.0,
      "learning_rate": 2.8703614224971007e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8816
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.28125,
      "epoch": 0.4260243525318902,
      "grad_norm": 0.0,
      "learning_rate": 2.8701198299188243e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8817
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.125,
      "epoch": 0.4260726710475454,
      "grad_norm": 0.0,
      "learning_rate": 2.869878237340549e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8818
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.375,
      "epoch": 0.4261209895632006,
      "grad_norm": 0.0,
      "learning_rate": 2.8696366447622725e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8819
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.0,
      "epoch": 0.42616930807885584,
      "grad_norm": 0.0,
      "learning_rate": 2.8693950521839967e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8820
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.46875,
      "epoch": 0.426217626594511,
      "grad_norm": 0.0,
      "learning_rate": 2.869153459605721e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 8821
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.46875,
      "epoch": 0.4262659451101662,
      "grad_norm": 0.0,
      "learning_rate": 2.868911867027445e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8822
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.59375,
      "epoch": 0.4263142636258214,
      "grad_norm": 0.0,
      "learning_rate": 2.8686702744491686e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8823
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.90625,
      "epoch": 0.4263625821414766,
      "grad_norm": 0.0,
      "learning_rate": 2.868428681870893e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8824
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.5,
      "epoch": 0.42641090065713183,
      "grad_norm": 0.0,
      "learning_rate": 2.868187089292617e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8825
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.84375,
      "epoch": 0.426459219172787,
      "grad_norm": 0.0,
      "learning_rate": 2.867945496714341e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8826
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.625,
      "epoch": 0.4265075376884422,
      "grad_norm": 0.0,
      "learning_rate": 2.8677039041360646e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8827
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.03125,
      "epoch": 0.4265558562040974,
      "grad_norm": 0.0,
      "learning_rate": 2.8674623115577887e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8828
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.34375,
      "epoch": 0.4266041747197526,
      "grad_norm": 0.0,
      "learning_rate": 2.867220718979513e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 8829
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.71875,
      "epoch": 0.4266524932354078,
      "grad_norm": 0.0,
      "learning_rate": 2.8669791264012365e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8830
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.4375,
      "epoch": 0.426700811751063,
      "grad_norm": 0.0,
      "learning_rate": 2.866737533822961e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8831
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.125,
      "epoch": 0.4267491302667182,
      "grad_norm": 0.0,
      "learning_rate": 2.8664959412446847e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8832
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.53125,
      "epoch": 0.4267974487823734,
      "grad_norm": 0.0,
      "learning_rate": 2.866254348666409e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8833
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.09375,
      "epoch": 0.4268457672980286,
      "grad_norm": 0.0,
      "learning_rate": 2.866012756088133e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8834
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.25,
      "epoch": 0.4268940858136838,
      "grad_norm": 0.0,
      "learning_rate": 2.865771163509857e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8835
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.5,
      "epoch": 0.426942404329339,
      "grad_norm": 0.0,
      "learning_rate": 2.865529570931581e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8836
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.53125,
      "epoch": 0.4269907228449942,
      "grad_norm": 0.0,
      "learning_rate": 2.8652879783533044e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 8837
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.15625,
      "epoch": 0.4270390413606494,
      "grad_norm": 0.0,
      "learning_rate": 2.865046385775029e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8838
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.4375,
      "epoch": 0.4270873598763046,
      "grad_norm": 0.0,
      "learning_rate": 2.8648047931967526e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8839
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 94.96875,
      "epoch": 0.4271356783919598,
      "grad_norm": 0.0,
      "learning_rate": 2.864563200618477e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 8840
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.96875,
      "epoch": 0.427183996907615,
      "grad_norm": 0.0,
      "learning_rate": 2.864321608040201e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8841
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.9375,
      "epoch": 0.42723231542327017,
      "grad_norm": 0.0,
      "learning_rate": 2.864080015461925e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8842
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.375,
      "epoch": 0.4272806339389254,
      "grad_norm": 0.0,
      "learning_rate": 2.8638384228836486e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8843
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.9375,
      "epoch": 0.4273289524545806,
      "grad_norm": 0.0,
      "learning_rate": 2.8635968303053733e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8844
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.09375,
      "epoch": 0.4273772709702358,
      "grad_norm": 0.0,
      "learning_rate": 2.863355237727097e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8845
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.375,
      "epoch": 0.427425589485891,
      "grad_norm": 216.5741375954538,
      "learning_rate": 2.863113645148821e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.31663835048675537,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8846
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.875,
      "epoch": 0.42747390800154617,
      "grad_norm": 0.0,
      "learning_rate": 2.8628720525705447e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 8847
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.4375,
      "epoch": 0.4275222265172014,
      "grad_norm": 0.0,
      "learning_rate": 2.862630459992269e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8848
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.40625,
      "epoch": 0.4275705450328566,
      "grad_norm": 0.0,
      "learning_rate": 2.862388867413993e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8849
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.8125,
      "epoch": 0.4276188635485118,
      "grad_norm": 0.0,
      "learning_rate": 2.8621472748357165e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8850
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.15625,
      "epoch": 0.427667182064167,
      "grad_norm": 0.0,
      "learning_rate": 2.861905682257441e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8851
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.90625,
      "epoch": 0.42771550057982216,
      "grad_norm": 0.0,
      "learning_rate": 2.861664089679165e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8852
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.5625,
      "epoch": 0.4277638190954774,
      "grad_norm": 0.0,
      "learning_rate": 2.861422497100889e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8853
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.78125,
      "epoch": 0.42781213761113257,
      "grad_norm": 0.0,
      "learning_rate": 2.861180904522613e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8854
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.75,
      "epoch": 0.4278604561267878,
      "grad_norm": 0.0,
      "learning_rate": 2.860939311944337e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8855
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.5,
      "epoch": 0.427908774642443,
      "grad_norm": 0.0,
      "learning_rate": 2.860697719366061e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8856
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.90625,
      "epoch": 0.42795709315809816,
      "grad_norm": 0.0,
      "learning_rate": 2.860456126787785e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8857
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.875,
      "epoch": 0.4280054116737534,
      "grad_norm": 41.577370238104116,
      "learning_rate": 2.860214534209509e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.3014557361602783,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8858
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.5,
      "epoch": 0.42805373018940857,
      "grad_norm": 0.0,
      "learning_rate": 2.8599729416312327e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8859
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.125,
      "epoch": 0.4281020487050638,
      "grad_norm": 0.0,
      "learning_rate": 2.859731349052957e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8860
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.0625,
      "epoch": 0.428150367220719,
      "grad_norm": 0.0,
      "learning_rate": 2.859489756474681e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8861
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.625,
      "epoch": 0.42819868573637415,
      "grad_norm": 0.0,
      "learning_rate": 2.859248163896405e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8862
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.75,
      "epoch": 0.4282470042520294,
      "grad_norm": 0.0,
      "learning_rate": 2.8590065713181287e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8863
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.09375,
      "epoch": 0.42829532276768456,
      "grad_norm": 0.0,
      "learning_rate": 2.8587649787398534e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8864
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.0625,
      "epoch": 0.4283436412833398,
      "grad_norm": 0.0,
      "learning_rate": 2.858523386161577e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8865
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.375,
      "epoch": 0.42839195979899497,
      "grad_norm": 0.0,
      "learning_rate": 2.858281793583301e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8866
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.96875,
      "epoch": 0.42844027831465015,
      "grad_norm": 0.0,
      "learning_rate": 2.8580402010050247e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8867
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.25,
      "epoch": 0.4284885968303054,
      "grad_norm": 0.0,
      "learning_rate": 2.8577986084267494e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8868
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.78125,
      "epoch": 0.42853691534596056,
      "grad_norm": 0.0,
      "learning_rate": 2.857557015848473e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 8869
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.1875,
      "epoch": 0.4285852338616158,
      "grad_norm": 0.0,
      "learning_rate": 2.8573154232701966e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8870
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.78125,
      "epoch": 0.42863355237727097,
      "grad_norm": 0.0,
      "learning_rate": 2.8570738306919213e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8871
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.625,
      "epoch": 0.42868187089292614,
      "grad_norm": 0.0,
      "learning_rate": 2.856832238113645e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8872
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.46875,
      "epoch": 0.4287301894085814,
      "grad_norm": 0.0,
      "learning_rate": 2.856590645535369e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8873
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.125,
      "epoch": 0.42877850792423655,
      "grad_norm": 0.0,
      "learning_rate": 2.856349052957093e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8874
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.28125,
      "epoch": 0.4288268264398918,
      "grad_norm": 0.0,
      "learning_rate": 2.8561074603788173e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8875
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.46875,
      "epoch": 0.42887514495554696,
      "grad_norm": 0.0,
      "learning_rate": 2.855865867800541e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8876
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.4375,
      "epoch": 0.42892346347120214,
      "grad_norm": 0.0,
      "learning_rate": 2.855624275222265e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8877
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.84375,
      "epoch": 0.42897178198685737,
      "grad_norm": 0.0,
      "learning_rate": 2.855382682643989e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8878
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.5,
      "epoch": 0.42902010050251255,
      "grad_norm": 0.0,
      "learning_rate": 2.855141090065713e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8879
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.53125,
      "epoch": 0.4290684190181678,
      "grad_norm": 0.0,
      "learning_rate": 2.854899497487437e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8880
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.75,
      "epoch": 0.42911673753382296,
      "grad_norm": 0.0,
      "learning_rate": 2.854657904909161e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8881
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.46875,
      "epoch": 0.42916505604947813,
      "grad_norm": 0.0,
      "learning_rate": 2.854416312330885e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8882
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.625,
      "epoch": 0.42921337456513337,
      "grad_norm": 0.0,
      "learning_rate": 2.854174719752609e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8883
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.25,
      "epoch": 0.42926169308078854,
      "grad_norm": 0.0,
      "learning_rate": 2.8539331271743335e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8884
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.0625,
      "epoch": 0.4293100115964438,
      "grad_norm": 0.0,
      "learning_rate": 2.853691534596057e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8885
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.625,
      "epoch": 0.42935833011209895,
      "grad_norm": 0.0,
      "learning_rate": 2.853449942017781e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 8886
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.4375,
      "epoch": 0.42940664862775413,
      "grad_norm": 0.0,
      "learning_rate": 2.853208349439505e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8887
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.3125,
      "epoch": 0.42945496714340936,
      "grad_norm": 0.0,
      "learning_rate": 2.8529667568612295e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8888
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.3125,
      "epoch": 0.42950328565906454,
      "grad_norm": 0.0,
      "learning_rate": 2.852725164282953e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24251842498779297,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8889
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.96875,
      "epoch": 0.42955160417471977,
      "grad_norm": 0.0,
      "learning_rate": 2.8524835717046767e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8890
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.71875,
      "epoch": 0.42959992269037495,
      "grad_norm": 0.0,
      "learning_rate": 2.8522419791264014e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8891
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.3125,
      "epoch": 0.4296482412060301,
      "grad_norm": 0.0,
      "learning_rate": 2.852000386548125e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8892
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.15625,
      "epoch": 0.42969655972168536,
      "grad_norm": 0.0,
      "learning_rate": 2.851758793969849e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8893
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.25,
      "epoch": 0.42974487823734053,
      "grad_norm": 0.0,
      "learning_rate": 2.851517201391573e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8894
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.25,
      "epoch": 0.42979319675299577,
      "grad_norm": 0.0,
      "learning_rate": 2.8512756088132974e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8895
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.3125,
      "epoch": 0.42984151526865094,
      "grad_norm": 0.0,
      "learning_rate": 2.851034016235021e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8896
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.9375,
      "epoch": 0.4298898337843061,
      "grad_norm": 0.0,
      "learning_rate": 2.850792423656745e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8897
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.46875,
      "epoch": 0.42993815229996135,
      "grad_norm": 0.0,
      "learning_rate": 2.850550831078469e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8898
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.8125,
      "epoch": 0.42998647081561653,
      "grad_norm": 0.0,
      "learning_rate": 2.850309238500193e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 8899
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.21875,
      "epoch": 0.43003478933127176,
      "grad_norm": 0.0,
      "learning_rate": 2.850067645921917e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8900
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.4375,
      "epoch": 0.43008310784692694,
      "grad_norm": 0.0,
      "learning_rate": 2.849826053343641e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8901
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.59375,
      "epoch": 0.4301314263625821,
      "grad_norm": 0.0,
      "learning_rate": 2.8495844607653653e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8902
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.1875,
      "epoch": 0.43017974487823735,
      "grad_norm": 0.0,
      "learning_rate": 2.849342868187089e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 8903
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 80.0,
      "epoch": 0.4302280633938925,
      "grad_norm": 0.0,
      "learning_rate": 2.8491012756088135e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 8904
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.34375,
      "epoch": 0.43027638190954776,
      "grad_norm": 0.0,
      "learning_rate": 2.848859683030537e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8905
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.65625,
      "epoch": 0.43032470042520293,
      "grad_norm": 0.0,
      "learning_rate": 2.8486180904522613e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8906
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.75,
      "epoch": 0.4303730189408581,
      "grad_norm": 0.0,
      "learning_rate": 2.848376497873985e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8907
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.09375,
      "epoch": 0.43042133745651334,
      "grad_norm": 0.0,
      "learning_rate": 2.8481349052957096e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8908
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.5,
      "epoch": 0.4304696559721685,
      "grad_norm": 0.0,
      "learning_rate": 2.847893312717433e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8909
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.5,
      "epoch": 0.43051797448782375,
      "grad_norm": 0.0,
      "learning_rate": 2.847651720139157e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8910
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.1875,
      "epoch": 0.43056629300347893,
      "grad_norm": 0.0,
      "learning_rate": 2.8474101275608814e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8911
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.84375,
      "epoch": 0.4306146115191341,
      "grad_norm": 0.0,
      "learning_rate": 2.847168534982605e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8912
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.53125,
      "epoch": 0.43066293003478934,
      "grad_norm": 0.0,
      "learning_rate": 2.846926942404329e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8913
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.9375,
      "epoch": 0.4307112485504445,
      "grad_norm": 0.0,
      "learning_rate": 2.8466853498260533e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 8914
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.1875,
      "epoch": 0.43075956706609975,
      "grad_norm": 0.0,
      "learning_rate": 2.8464437572477775e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8915
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.1875,
      "epoch": 0.4308078855817549,
      "grad_norm": 0.0,
      "learning_rate": 2.846202164669501e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8916
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.1875,
      "epoch": 0.4308562040974101,
      "grad_norm": 0.0,
      "learning_rate": 2.845960572091225e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8917
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.25,
      "epoch": 0.43090452261306533,
      "grad_norm": 0.0,
      "learning_rate": 2.8457189795129493e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8918
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.28125,
      "epoch": 0.4309528411287205,
      "grad_norm": 0.0,
      "learning_rate": 2.8454773869346735e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8919
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.5625,
      "epoch": 0.43100115964437574,
      "grad_norm": 29.221345709255257,
      "learning_rate": 2.845235794356397e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 8920
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.8125,
      "epoch": 0.4310494781600309,
      "grad_norm": 0.0,
      "learning_rate": 2.844994201778121e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8921
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.15625,
      "epoch": 0.4310977966756861,
      "grad_norm": 0.0,
      "learning_rate": 2.8447526091998453e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8922
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.59375,
      "epoch": 0.43114611519134133,
      "grad_norm": 0.0,
      "learning_rate": 2.844511016621569e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8923
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.84375,
      "epoch": 0.4311944337069965,
      "grad_norm": 0.0,
      "learning_rate": 2.8442694240432936e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8924
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.78125,
      "epoch": 0.43124275222265174,
      "grad_norm": 0.0,
      "learning_rate": 2.844027831465017e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8925
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.6875,
      "epoch": 0.4312910707383069,
      "grad_norm": 0.0,
      "learning_rate": 2.8437862388867414e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8926
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.6875,
      "epoch": 0.4313393892539621,
      "grad_norm": 0.0,
      "learning_rate": 2.843544646308465e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8927
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.21875,
      "epoch": 0.4313877077696173,
      "grad_norm": 0.0,
      "learning_rate": 2.8433030537301896e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8928
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.34375,
      "epoch": 0.4314360262852725,
      "grad_norm": 0.0,
      "learning_rate": 2.843061461151913e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8929
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.1875,
      "epoch": 0.43148434480092773,
      "grad_norm": 0.0,
      "learning_rate": 2.842819868573637e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 8930
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.59375,
      "epoch": 0.4315326633165829,
      "grad_norm": 0.0,
      "learning_rate": 2.8425782759953615e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8931
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.78125,
      "epoch": 0.4315809818322381,
      "grad_norm": 0.0,
      "learning_rate": 2.842336683417085e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8932
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.03125,
      "epoch": 0.4316293003478933,
      "grad_norm": 0.0,
      "learning_rate": 2.842095090838809e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8933
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.78125,
      "epoch": 0.4316776188635485,
      "grad_norm": 0.0,
      "learning_rate": 2.8418534982605334e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8934
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 82.65625,
      "epoch": 0.43172593737920373,
      "grad_norm": 0.0,
      "learning_rate": 2.8416119056822575e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 8935
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.03125,
      "epoch": 0.4317742558948589,
      "grad_norm": 0.0,
      "learning_rate": 2.841370313103981e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8936
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.625,
      "epoch": 0.4318225744105141,
      "grad_norm": 0.0,
      "learning_rate": 2.8411287205257053e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8937
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.96875,
      "epoch": 0.4318708929261693,
      "grad_norm": 0.0,
      "learning_rate": 2.8408871279474294e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 8938
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.34375,
      "epoch": 0.4319192114418245,
      "grad_norm": 0.0,
      "learning_rate": 2.8406455353691535e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8939
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.90625,
      "epoch": 0.4319675299574797,
      "grad_norm": 0.0,
      "learning_rate": 2.840403942790877e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 8940
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.90625,
      "epoch": 0.4320158484731349,
      "grad_norm": 0.0,
      "learning_rate": 2.8401623502126013e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8941
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.0,
      "epoch": 0.4320641669887901,
      "grad_norm": 0.0,
      "learning_rate": 2.8399207576343254e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8942
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.0,
      "epoch": 0.4321124855044453,
      "grad_norm": 0.0,
      "learning_rate": 2.839679165056049e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8943
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.25,
      "epoch": 0.4321608040201005,
      "grad_norm": 0.0,
      "learning_rate": 2.8394375724777737e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8944
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.875,
      "epoch": 0.4322091225357557,
      "grad_norm": 0.0,
      "learning_rate": 2.8391959798994973e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8945
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.0625,
      "epoch": 0.4322574410514109,
      "grad_norm": 0.0,
      "learning_rate": 2.8389543873212214e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8946
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.75,
      "epoch": 0.4323057595670661,
      "grad_norm": 0.0,
      "learning_rate": 2.838712794742945e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8947
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.75,
      "epoch": 0.4323540780827213,
      "grad_norm": 0.0,
      "learning_rate": 2.8384712021646697e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8948
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.65625,
      "epoch": 0.4324023965983765,
      "grad_norm": 0.0,
      "learning_rate": 2.8382296095863933e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8949
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.15625,
      "epoch": 0.4324507151140317,
      "grad_norm": 0.0,
      "learning_rate": 2.837988017008117e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 8950
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.84375,
      "epoch": 0.4324990336296869,
      "grad_norm": 0.0,
      "learning_rate": 2.8377464244298416e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 8951
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.4375,
      "epoch": 0.43254735214534207,
      "grad_norm": 0.0,
      "learning_rate": 2.837504831851565e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8952
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.03125,
      "epoch": 0.4325956706609973,
      "grad_norm": 0.0,
      "learning_rate": 2.8372632392732893e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8953
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.34375,
      "epoch": 0.4326439891766525,
      "grad_norm": 0.0,
      "learning_rate": 2.8370216466950135e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 8954
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.84375,
      "epoch": 0.4326923076923077,
      "grad_norm": 0.0,
      "learning_rate": 2.8367800541167376e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8955
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.5625,
      "epoch": 0.4327406262079629,
      "grad_norm": 0.0,
      "learning_rate": 2.836538461538461e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8956
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 91.0,
      "epoch": 0.43278894472361806,
      "grad_norm": 10.469344751496395,
      "learning_rate": 2.8362968689601854e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8957
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.21875,
      "epoch": 0.4328372632392733,
      "grad_norm": 0.0,
      "learning_rate": 2.8360552763819095e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8958
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.90625,
      "epoch": 0.4328855817549285,
      "grad_norm": 0.0,
      "learning_rate": 2.8358136838036336e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8959
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.09375,
      "epoch": 0.4329339002705837,
      "grad_norm": 0.0,
      "learning_rate": 2.835572091225357e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 8960
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.28125,
      "epoch": 0.4329822187862389,
      "grad_norm": 0.0,
      "learning_rate": 2.8353304986470814e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8961
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.96875,
      "epoch": 0.43303053730189406,
      "grad_norm": 0.0,
      "learning_rate": 2.8350889060688055e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8962
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.71875,
      "epoch": 0.4330788558175493,
      "grad_norm": 0.0,
      "learning_rate": 2.834847313490529e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880052089691,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8963
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.375,
      "epoch": 0.43312717433320447,
      "grad_norm": 0.0,
      "learning_rate": 2.834605720912254e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8964
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.40625,
      "epoch": 0.4331754928488597,
      "grad_norm": 0.0,
      "learning_rate": 2.8343641283339774e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8965
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.03125,
      "epoch": 0.4332238113645149,
      "grad_norm": 0.0,
      "learning_rate": 2.8341225357557015e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8966
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.40625,
      "epoch": 0.43327212988017005,
      "grad_norm": 0.0,
      "learning_rate": 2.833880943177425e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8967
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.28125,
      "epoch": 0.4333204483958253,
      "grad_norm": 0.0,
      "learning_rate": 2.83363935059915e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8968
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.71875,
      "epoch": 0.43336876691148046,
      "grad_norm": 0.0,
      "learning_rate": 2.8333977580208734e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8969
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.09375,
      "epoch": 0.4334170854271357,
      "grad_norm": 0.0,
      "learning_rate": 2.8331561654425975e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 8970
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.25,
      "epoch": 0.4334654039427909,
      "grad_norm": 0.0,
      "learning_rate": 2.8329145728643217e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8971
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.96875,
      "epoch": 0.43351372245844605,
      "grad_norm": 0.0,
      "learning_rate": 2.8326729802860453e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880052089691,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8972
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.53125,
      "epoch": 0.4335620409741013,
      "grad_norm": 0.0,
      "learning_rate": 2.8324313877077694e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8973
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.4375,
      "epoch": 0.43361035948975646,
      "grad_norm": 0.0,
      "learning_rate": 2.8321897951294936e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8974
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.25,
      "epoch": 0.4336586780054117,
      "grad_norm": 0.0,
      "learning_rate": 2.8319482025512177e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8975
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.25,
      "epoch": 0.43370699652106687,
      "grad_norm": 0.0,
      "learning_rate": 2.8317066099729413e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8976
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.21875,
      "epoch": 0.43375531503672204,
      "grad_norm": 0.0,
      "learning_rate": 2.8314650173946654e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8977
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.6875,
      "epoch": 0.4338036335523773,
      "grad_norm": 0.0,
      "learning_rate": 2.8312234248163896e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 8978
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.28125,
      "epoch": 0.43385195206803245,
      "grad_norm": 0.0,
      "learning_rate": 2.8309818322381137e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8979
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.21875,
      "epoch": 0.4339002705836877,
      "grad_norm": 0.0,
      "learning_rate": 2.8307402396598373e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8980
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.59375,
      "epoch": 0.43394858909934286,
      "grad_norm": 0.0,
      "learning_rate": 2.830498647081562e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 8981
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.75,
      "epoch": 0.43399690761499804,
      "grad_norm": 0.0,
      "learning_rate": 2.8302570545032856e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 8982
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.5625,
      "epoch": 0.43404522613065327,
      "grad_norm": 62.83289205800531,
      "learning_rate": 2.830015461925009e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 8983
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.0625,
      "epoch": 0.43409354464630845,
      "grad_norm": 0.0,
      "learning_rate": 2.829773869346734e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8984
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.96875,
      "epoch": 0.4341418631619637,
      "grad_norm": 0.0,
      "learning_rate": 2.8295322767684575e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8985
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.375,
      "epoch": 0.43419018167761886,
      "grad_norm": 0.0,
      "learning_rate": 2.8292906841901816e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 8986
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.625,
      "epoch": 0.43423850019327404,
      "grad_norm": 0.0,
      "learning_rate": 2.829049091611905e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 8987
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.625,
      "epoch": 0.43428681870892927,
      "grad_norm": 0.0,
      "learning_rate": 2.82880749903363e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8988
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.21875,
      "epoch": 0.43433513722458444,
      "grad_norm": 0.0,
      "learning_rate": 2.8285659064553535e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8989
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.8125,
      "epoch": 0.4343834557402397,
      "grad_norm": 0.0,
      "learning_rate": 2.8283243138770776e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 8990
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.625,
      "epoch": 0.43443177425589485,
      "grad_norm": 0.0,
      "learning_rate": 2.828082721298802e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 8991
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.8125,
      "epoch": 0.43448009277155003,
      "grad_norm": 0.0,
      "learning_rate": 2.8278411287205254e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 8992
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.0,
      "epoch": 0.43452841128720526,
      "grad_norm": 0.0,
      "learning_rate": 2.8275995361422495e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 8993
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.09375,
      "epoch": 0.43457672980286044,
      "grad_norm": 0.0,
      "learning_rate": 2.8273579435639736e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 8994
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.96875,
      "epoch": 0.43462504831851567,
      "grad_norm": 0.0,
      "learning_rate": 2.827116350985698e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 8995
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.15625,
      "epoch": 0.43467336683417085,
      "grad_norm": 0.0,
      "learning_rate": 2.8268747584074214e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 8996
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.96875,
      "epoch": 0.434721685349826,
      "grad_norm": 0.0,
      "learning_rate": 2.8266331658291455e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 8997
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.53125,
      "epoch": 0.43477000386548126,
      "grad_norm": 0.0,
      "learning_rate": 2.8263915732508696e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 8998
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.71875,
      "epoch": 0.43481832238113644,
      "grad_norm": 0.0,
      "learning_rate": 2.826149980672594e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 8999
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.4375,
      "epoch": 0.43486664089679167,
      "grad_norm": 0.0,
      "learning_rate": 2.8259083880943174e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24251842498779297,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9000
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.5625,
      "epoch": 0.43491495941244684,
      "grad_norm": 0.0,
      "learning_rate": 2.825666795516042e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9001
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.78125,
      "epoch": 0.434963277928102,
      "grad_norm": 0.0,
      "learning_rate": 2.8254252029377657e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9002
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.15625,
      "epoch": 0.43501159644375725,
      "grad_norm": 0.0,
      "learning_rate": 2.8251836103594893e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9003
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.46875,
      "epoch": 0.43505991495941243,
      "grad_norm": 0.0,
      "learning_rate": 2.824942017781214e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9004
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.4375,
      "epoch": 0.43510823347506766,
      "grad_norm": 0.0,
      "learning_rate": 2.8247004252029375e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9005
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.4375,
      "epoch": 0.43515655199072284,
      "grad_norm": 27.40749646696424,
      "learning_rate": 2.8244588326246617e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9006
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.90625,
      "epoch": 0.435204870506378,
      "grad_norm": 0.0,
      "learning_rate": 2.8242172400463853e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9007
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.46875,
      "epoch": 0.43525318902203325,
      "grad_norm": 0.0,
      "learning_rate": 2.82397564746811e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9008
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.9375,
      "epoch": 0.4353015075376884,
      "grad_norm": 0.0,
      "learning_rate": 2.8237340548898336e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9009
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.5,
      "epoch": 0.43534982605334366,
      "grad_norm": 0.0,
      "learning_rate": 2.8234924623115577e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9010
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.46875,
      "epoch": 0.43539814456899883,
      "grad_norm": 0.0,
      "learning_rate": 2.823250869733282e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9011
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.65625,
      "epoch": 0.435446463084654,
      "grad_norm": 0.0,
      "learning_rate": 2.8230092771550054e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9012
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.96875,
      "epoch": 0.43549478160030924,
      "grad_norm": 0.0,
      "learning_rate": 2.8227676845767296e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9013
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.5,
      "epoch": 0.4355431001159644,
      "grad_norm": 0.0,
      "learning_rate": 2.8225260919984537e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9014
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.46875,
      "epoch": 0.43559141863161965,
      "grad_norm": 0.0,
      "learning_rate": 2.822284499420178e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9015
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.125,
      "epoch": 0.43563973714727483,
      "grad_norm": 0.0,
      "learning_rate": 2.8220429068419015e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9016
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.5625,
      "epoch": 0.43568805566293,
      "grad_norm": 0.0,
      "learning_rate": 2.8218013142636256e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9017
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.65625,
      "epoch": 0.43573637417858524,
      "grad_norm": 0.0,
      "learning_rate": 2.8215597216853497e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9018
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.3125,
      "epoch": 0.4357846926942404,
      "grad_norm": 0.0,
      "learning_rate": 2.821318129107074e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9019
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.65625,
      "epoch": 0.43583301120989565,
      "grad_norm": 0.0,
      "learning_rate": 2.8210765365287975e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9020
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.6875,
      "epoch": 0.4358813297255508,
      "grad_norm": 0.0,
      "learning_rate": 2.820834943950522e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9021
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.96875,
      "epoch": 0.435929648241206,
      "grad_norm": 0.0,
      "learning_rate": 2.820593351372246e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9022
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.84375,
      "epoch": 0.43597796675686123,
      "grad_norm": 0.0,
      "learning_rate": 2.8203517587939693e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9023
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.375,
      "epoch": 0.4360262852725164,
      "grad_norm": 0.0,
      "learning_rate": 2.820110166215694e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9024
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.6875,
      "epoch": 0.43607460378817164,
      "grad_norm": 0.0,
      "learning_rate": 2.8198685736374176e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9025
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.84375,
      "epoch": 0.4361229223038268,
      "grad_norm": 0.0,
      "learning_rate": 2.819626981059142e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 9026
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.3125,
      "epoch": 0.436171240819482,
      "grad_norm": 0.0,
      "learning_rate": 2.8193853884808654e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9027
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.8125,
      "epoch": 0.43621955933513723,
      "grad_norm": 0.0,
      "learning_rate": 2.81914379590259e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9028
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.90625,
      "epoch": 0.4362678778507924,
      "grad_norm": 0.0,
      "learning_rate": 2.8189022033243136e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9029
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.28125,
      "epoch": 0.43631619636644764,
      "grad_norm": 0.0,
      "learning_rate": 2.818660610746038e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.16130642592906952,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 0.96875,
      "step": 9030
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.5625,
      "epoch": 0.4363645148821028,
      "grad_norm": 0.0,
      "learning_rate": 2.818419018167762e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9031
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.5625,
      "epoch": 0.436412833397758,
      "grad_norm": 0.0,
      "learning_rate": 2.818177425589486e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9032
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.34375,
      "epoch": 0.4364611519134132,
      "grad_norm": 0.0,
      "learning_rate": 2.8179358330112097e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880052089691,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9033
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.84375,
      "epoch": 0.4365094704290684,
      "grad_norm": 0.0,
      "learning_rate": 2.817694240432934e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9034
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.0,
      "epoch": 0.43655778894472363,
      "grad_norm": 0.0,
      "learning_rate": 2.817452647854658e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9035
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.375,
      "epoch": 0.4366061074603788,
      "grad_norm": 0.0,
      "learning_rate": 2.8172110552763815e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9036
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.65625,
      "epoch": 0.436654425976034,
      "grad_norm": 0.0,
      "learning_rate": 2.8169694626981057e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9037
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.75,
      "epoch": 0.4367027444916892,
      "grad_norm": 0.0,
      "learning_rate": 2.81672787011983e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9038
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.875,
      "epoch": 0.4367510630073444,
      "grad_norm": 0.0,
      "learning_rate": 2.816486277541554e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9039
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.625,
      "epoch": 0.43679938152299963,
      "grad_norm": 0.0,
      "learning_rate": 2.8162446849632775e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9040
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.0,
      "epoch": 0.4368477000386548,
      "grad_norm": 0.0,
      "learning_rate": 2.816003092385002e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9041
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.6875,
      "epoch": 0.43689601855431,
      "grad_norm": 0.0,
      "learning_rate": 2.815761499806726e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9042
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.34375,
      "epoch": 0.4369443370699652,
      "grad_norm": 0.0,
      "learning_rate": 2.8155199072284494e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9043
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.46875,
      "epoch": 0.4369926555856204,
      "grad_norm": 0.0,
      "learning_rate": 2.815278314650174e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9044
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.53125,
      "epoch": 0.4370409741012756,
      "grad_norm": 0.0,
      "learning_rate": 2.8150367220718977e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9045
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.5,
      "epoch": 0.4370892926169308,
      "grad_norm": 0.0,
      "learning_rate": 2.814795129493622e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9046
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.3125,
      "epoch": 0.437137611132586,
      "grad_norm": 0.0,
      "learning_rate": 2.8145535369153454e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9047
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.4375,
      "epoch": 0.4371859296482412,
      "grad_norm": 0.0,
      "learning_rate": 2.81431194433707e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9048
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.5,
      "epoch": 0.4372342481638964,
      "grad_norm": 0.0,
      "learning_rate": 2.8140703517587937e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9049
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.09375,
      "epoch": 0.4372825666795516,
      "grad_norm": 0.0,
      "learning_rate": 2.813828759180518e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9050
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.90625,
      "epoch": 0.4373308851952068,
      "grad_norm": 0.0,
      "learning_rate": 2.813587166602242e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9051
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.4375,
      "epoch": 0.437379203710862,
      "grad_norm": 0.0,
      "learning_rate": 2.813345574023966e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9052
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.15625,
      "epoch": 0.4374275222265172,
      "grad_norm": 0.0,
      "learning_rate": 2.8131039814456897e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9053
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.4375,
      "epoch": 0.4374758407421724,
      "grad_norm": 0.0,
      "learning_rate": 2.812862388867414e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9054
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.71875,
      "epoch": 0.4375241592578276,
      "grad_norm": 0.0,
      "learning_rate": 2.812620796289138e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9055
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.78125,
      "epoch": 0.4375724777734828,
      "grad_norm": 0.0,
      "learning_rate": 2.8123792037108616e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9056
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.25,
      "epoch": 0.437620796289138,
      "grad_norm": 0.0,
      "learning_rate": 2.812137611132586e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9057
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.96875,
      "epoch": 0.4376691148047932,
      "grad_norm": 0.0,
      "learning_rate": 2.81189601855431e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2130674123764038,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9058
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.40625,
      "epoch": 0.4377174333204484,
      "grad_norm": 0.0,
      "learning_rate": 2.811654425976034e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9059
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.9375,
      "epoch": 0.4377657518361036,
      "grad_norm": 0.0,
      "learning_rate": 2.8114128333977576e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9060
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.25,
      "epoch": 0.4378140703517588,
      "grad_norm": 0.0,
      "learning_rate": 2.8111712408194823e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9061
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.46875,
      "epoch": 0.437862388867414,
      "grad_norm": 0.0,
      "learning_rate": 2.810929648241206e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9062
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.5625,
      "epoch": 0.4379107073830692,
      "grad_norm": 0.0,
      "learning_rate": 2.8106880556629295e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9063
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.625,
      "epoch": 0.4379590258987244,
      "grad_norm": 0.0,
      "learning_rate": 2.810446463084654e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9064
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.34375,
      "epoch": 0.4380073444143796,
      "grad_norm": 0.0,
      "learning_rate": 2.810204870506378e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9065
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.40625,
      "epoch": 0.4380556629300348,
      "grad_norm": 0.0,
      "learning_rate": 2.809963277928102e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9066
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.96875,
      "epoch": 0.43810398144569,
      "grad_norm": 0.0,
      "learning_rate": 2.8097216853498255e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9067
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.84375,
      "epoch": 0.4381522999613452,
      "grad_norm": 0.0,
      "learning_rate": 2.80948009277155e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9068
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.125,
      "epoch": 0.43820061847700037,
      "grad_norm": 0.0,
      "learning_rate": 2.809238500193274e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9069
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.03125,
      "epoch": 0.4382489369926556,
      "grad_norm": 0.0,
      "learning_rate": 2.808996907614998e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9070
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.6875,
      "epoch": 0.4382972555083108,
      "grad_norm": 0.0,
      "learning_rate": 2.808755315036722e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9071
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.46875,
      "epoch": 0.438345574023966,
      "grad_norm": 0.0,
      "learning_rate": 2.808513722458446e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9072
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.4375,
      "epoch": 0.4383938925396212,
      "grad_norm": 0.0,
      "learning_rate": 2.80827212988017e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9073
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.8125,
      "epoch": 0.43844221105527637,
      "grad_norm": 0.0,
      "learning_rate": 2.808030537301894e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9074
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.875,
      "epoch": 0.4384905295709316,
      "grad_norm": 0.0,
      "learning_rate": 2.807788944723618e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9075
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.96875,
      "epoch": 0.4385388480865868,
      "grad_norm": 0.0,
      "learning_rate": 2.8075473521453417e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.249358132481575,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9076
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.21875,
      "epoch": 0.438587166602242,
      "grad_norm": 0.0,
      "learning_rate": 2.807305759567066e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9077
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.84375,
      "epoch": 0.4386354851178972,
      "grad_norm": 0.0,
      "learning_rate": 2.80706416698879e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9078
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 72.59375,
      "epoch": 0.43868380363355236,
      "grad_norm": 0.0,
      "learning_rate": 2.806822574410514e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 9079
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.5625,
      "epoch": 0.4387321221492076,
      "grad_norm": 0.0,
      "learning_rate": 2.8065809818322377e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9080
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.375,
      "epoch": 0.43878044066486277,
      "grad_norm": 0.0,
      "learning_rate": 2.8063393892539624e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9081
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.6875,
      "epoch": 0.438828759180518,
      "grad_norm": 0.0,
      "learning_rate": 2.806097796675686e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 9082
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.9375,
      "epoch": 0.4388770776961732,
      "grad_norm": 0.0,
      "learning_rate": 2.80585620409741e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9083
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.90625,
      "epoch": 0.43892539621182836,
      "grad_norm": 0.0,
      "learning_rate": 2.805614611519134e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9084
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.21875,
      "epoch": 0.4389737147274836,
      "grad_norm": 0.0,
      "learning_rate": 2.805373018940858e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9085
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.46875,
      "epoch": 0.43902203324313877,
      "grad_norm": 0.0,
      "learning_rate": 2.805131426362582e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9086
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.53125,
      "epoch": 0.439070351758794,
      "grad_norm": 0.0,
      "learning_rate": 2.8048898337843056e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9087
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.84375,
      "epoch": 0.4391186702744492,
      "grad_norm": 0.0,
      "learning_rate": 2.8046482412060303e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9088
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.65625,
      "epoch": 0.43916698879010435,
      "grad_norm": 0.0,
      "learning_rate": 2.804406648627754e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9089
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.375,
      "epoch": 0.4392153073057596,
      "grad_norm": 0.0,
      "learning_rate": 2.804165056049478e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9090
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.90625,
      "epoch": 0.43926362582141476,
      "grad_norm": 0.0,
      "learning_rate": 2.803923463471202e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9091
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.3125,
      "epoch": 0.43931194433707,
      "grad_norm": 0.0,
      "learning_rate": 2.8036818708929263e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9092
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.75,
      "epoch": 0.43936026285272517,
      "grad_norm": 0.0,
      "learning_rate": 2.80344027831465e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9093
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.3125,
      "epoch": 0.43940858136838035,
      "grad_norm": 0.0,
      "learning_rate": 2.8031986857363746e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9094
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.15625,
      "epoch": 0.4394568998840356,
      "grad_norm": 0.0,
      "learning_rate": 2.802957093158098e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9095
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.875,
      "epoch": 0.43950521839969076,
      "grad_norm": 0.0,
      "learning_rate": 2.802715500579822e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9096
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.875,
      "epoch": 0.439553536915346,
      "grad_norm": 0.0,
      "learning_rate": 2.802473908001546e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9097
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.375,
      "epoch": 0.43960185543100117,
      "grad_norm": 0.0,
      "learning_rate": 2.80223231542327e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9098
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.9375,
      "epoch": 0.43965017394665634,
      "grad_norm": 0.0,
      "learning_rate": 2.801990722844994e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9099
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.875,
      "epoch": 0.4396984924623116,
      "grad_norm": 0.0,
      "learning_rate": 2.801749130266718e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9100
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.875,
      "epoch": 0.43974681097796675,
      "grad_norm": 0.0,
      "learning_rate": 2.8015075376884424e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9101
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.40625,
      "epoch": 0.439795129493622,
      "grad_norm": 0.0,
      "learning_rate": 2.801265945110166e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9102
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.6875,
      "epoch": 0.43984344800927716,
      "grad_norm": 0.0,
      "learning_rate": 2.80102435253189e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 9103
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.96875,
      "epoch": 0.43989176652493234,
      "grad_norm": 0.0,
      "learning_rate": 2.8007827599536143e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9104
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.71875,
      "epoch": 0.43994008504058757,
      "grad_norm": 0.0,
      "learning_rate": 2.800541167375338e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9105
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.125,
      "epoch": 0.43998840355624275,
      "grad_norm": 0.0,
      "learning_rate": 2.800299574797062e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9106
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.84375,
      "epoch": 0.440036722071898,
      "grad_norm": 0.0,
      "learning_rate": 2.8000579822187857e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9107
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.78125,
      "epoch": 0.44008504058755316,
      "grad_norm": 0.0,
      "learning_rate": 2.7998163896405103e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9108
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.46875,
      "epoch": 0.44013335910320833,
      "grad_norm": 0.0,
      "learning_rate": 2.799574797062234e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 9109
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.75,
      "epoch": 0.44018167761886356,
      "grad_norm": 0.0,
      "learning_rate": 2.799333204483958e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9110
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.21875,
      "epoch": 0.44022999613451874,
      "grad_norm": 0.0,
      "learning_rate": 2.799091611905682e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9111
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.5,
      "epoch": 0.440278314650174,
      "grad_norm": 0.0,
      "learning_rate": 2.7988500193274064e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9112
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.5625,
      "epoch": 0.44032663316582915,
      "grad_norm": 0.0,
      "learning_rate": 2.79860842674913e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9113
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 75.875,
      "epoch": 0.44037495168148433,
      "grad_norm": 0.0,
      "learning_rate": 2.7983668341708546e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 9114
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 82.65625,
      "epoch": 0.44042327019713956,
      "grad_norm": 27.710910335928684,
      "learning_rate": 2.798125241592578e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9115
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.15625,
      "epoch": 0.44047158871279474,
      "grad_norm": 0.0,
      "learning_rate": 2.797883649014302e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9116
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.0625,
      "epoch": 0.44051990722844997,
      "grad_norm": 0.0,
      "learning_rate": 2.797642056436026e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24251843988895416,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9117
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.3125,
      "epoch": 0.44056822574410515,
      "grad_norm": 0.0,
      "learning_rate": 2.79740046385775e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9118
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.53125,
      "epoch": 0.4406165442597603,
      "grad_norm": 0.0,
      "learning_rate": 2.797158871279474e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 9119
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.875,
      "epoch": 0.44066486277541556,
      "grad_norm": 0.0,
      "learning_rate": 2.796917278701198e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9120
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.78125,
      "epoch": 0.44071318129107073,
      "grad_norm": 0.0,
      "learning_rate": 2.7966756861229225e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9121
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.5,
      "epoch": 0.44076149980672596,
      "grad_norm": 0.0,
      "learning_rate": 2.796434093544646e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9122
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.40625,
      "epoch": 0.44080981832238114,
      "grad_norm": 0.0,
      "learning_rate": 2.7961925009663703e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9123
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.78125,
      "epoch": 0.4408581368380363,
      "grad_norm": 0.0,
      "learning_rate": 2.7959509083880944e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9124
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.90625,
      "epoch": 0.44090645535369155,
      "grad_norm": 0.0,
      "learning_rate": 2.7957093158098185e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9125
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.96875,
      "epoch": 0.44095477386934673,
      "grad_norm": 0.0,
      "learning_rate": 2.795467723231542e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9126
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.125,
      "epoch": 0.44100309238500196,
      "grad_norm": 0.0,
      "learning_rate": 2.795226130653266e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9127
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.84375,
      "epoch": 0.44105141090065714,
      "grad_norm": 0.0,
      "learning_rate": 2.7949845380749904e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9128
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.3125,
      "epoch": 0.4410997294163123,
      "grad_norm": 0.0,
      "learning_rate": 2.794742945496714e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9129
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.75,
      "epoch": 0.44114804793196755,
      "grad_norm": 0.0,
      "learning_rate": 2.794501352918438e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9130
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.5,
      "epoch": 0.4411963664476227,
      "grad_norm": 0.0,
      "learning_rate": 2.7942597603401623e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 9131
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.4375,
      "epoch": 0.44124468496327796,
      "grad_norm": 0.0,
      "learning_rate": 2.7940181677618864e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9132
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.3125,
      "epoch": 0.44129300347893313,
      "grad_norm": 0.0,
      "learning_rate": 2.79377657518361e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9133
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.5,
      "epoch": 0.4413413219945883,
      "grad_norm": 0.0,
      "learning_rate": 2.7935349826053347e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9134
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.96875,
      "epoch": 0.44138964051024354,
      "grad_norm": 0.0,
      "learning_rate": 2.7932933900270583e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9135
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.09375,
      "epoch": 0.4414379590258987,
      "grad_norm": 0.0,
      "learning_rate": 2.793051797448782e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9136
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.53125,
      "epoch": 0.44148627754155395,
      "grad_norm": 0.0,
      "learning_rate": 2.792810204870506e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9137
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.125,
      "epoch": 0.4415345960572091,
      "grad_norm": 0.0,
      "learning_rate": 2.79256861229223e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 9138
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.375,
      "epoch": 0.4415829145728643,
      "grad_norm": 0.0,
      "learning_rate": 2.7923270197139543e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9139
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.15625,
      "epoch": 0.44163123308851954,
      "grad_norm": 0.0,
      "learning_rate": 2.792085427135678e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9140
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.78125,
      "epoch": 0.4416795516041747,
      "grad_norm": 0.0,
      "learning_rate": 2.7918438345574026e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9141
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.4375,
      "epoch": 0.44172787011982995,
      "grad_norm": 0.0,
      "learning_rate": 2.791602241979126e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9142
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.03125,
      "epoch": 0.4417761886354851,
      "grad_norm": 0.0,
      "learning_rate": 2.7913606494008503e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9143
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 87.9375,
      "epoch": 0.4418245071511403,
      "grad_norm": 0.0,
      "learning_rate": 2.7911190568225745e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 9144
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.46875,
      "epoch": 0.44187282566679553,
      "grad_norm": 0.0,
      "learning_rate": 2.7908774642442986e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9145
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.5625,
      "epoch": 0.4419211441824507,
      "grad_norm": 0.0,
      "learning_rate": 2.790635871666022e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9146
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.1875,
      "epoch": 0.44196946269810594,
      "grad_norm": 0.0,
      "learning_rate": 2.790394279087746e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9147
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.03125,
      "epoch": 0.4420177812137611,
      "grad_norm": 0.0,
      "learning_rate": 2.7901526865094705e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9148
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.71875,
      "epoch": 0.4420660997294163,
      "grad_norm": 0.0,
      "learning_rate": 2.789911093931194e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9149
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.71875,
      "epoch": 0.4421144182450715,
      "grad_norm": 0.0,
      "learning_rate": 2.789669501352918e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9150
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.3125,
      "epoch": 0.4421627367607267,
      "grad_norm": 0.0,
      "learning_rate": 2.7894279087746424e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9151
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.875,
      "epoch": 0.44221105527638194,
      "grad_norm": 0.0,
      "learning_rate": 2.7891863161963665e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9152
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.65625,
      "epoch": 0.4422593737920371,
      "grad_norm": 0.0,
      "learning_rate": 2.78894472361809e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9153
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.40625,
      "epoch": 0.4423076923076923,
      "grad_norm": 0.0,
      "learning_rate": 2.788703131039815e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9154
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.25,
      "epoch": 0.4423560108233475,
      "grad_norm": 0.0,
      "learning_rate": 2.7884615384615384e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9155
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.0,
      "epoch": 0.4424043293390027,
      "grad_norm": 0.0,
      "learning_rate": 2.788219945883262e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9156
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.5625,
      "epoch": 0.44245264785465793,
      "grad_norm": 0.0,
      "learning_rate": 2.787978353304986e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9157
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.9375,
      "epoch": 0.4425009663703131,
      "grad_norm": 0.0,
      "learning_rate": 2.7877367607267103e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9158
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.6875,
      "epoch": 0.4425492848859683,
      "grad_norm": 0.0,
      "learning_rate": 2.7874951681484344e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9159
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.875,
      "epoch": 0.4425976034016235,
      "grad_norm": 0.0,
      "learning_rate": 2.787253575570158e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9160
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.3125,
      "epoch": 0.4426459219172787,
      "grad_norm": 0.0,
      "learning_rate": 2.7870119829918827e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9161
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.71875,
      "epoch": 0.4426942404329339,
      "grad_norm": 0.0,
      "learning_rate": 2.7867703904136063e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9162
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.59375,
      "epoch": 0.4427425589485891,
      "grad_norm": 0.0,
      "learning_rate": 2.7865287978353304e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9163
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.75,
      "epoch": 0.4427908774642443,
      "grad_norm": 0.0,
      "learning_rate": 2.7862872052570546e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9164
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 91.40625,
      "epoch": 0.4428391959798995,
      "grad_norm": 0.0,
      "learning_rate": 2.7860456126787787e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 9165
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.3125,
      "epoch": 0.4428875144955547,
      "grad_norm": 0.0,
      "learning_rate": 2.7858040201005023e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9166
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.09375,
      "epoch": 0.4429358330112099,
      "grad_norm": 0.0,
      "learning_rate": 2.785562427522226e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9167
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.375,
      "epoch": 0.4429841515268651,
      "grad_norm": 0.0,
      "learning_rate": 2.7853208349439506e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9168
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.53125,
      "epoch": 0.4430324700425203,
      "grad_norm": 0.0,
      "learning_rate": 2.785079242365674e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9169
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.625,
      "epoch": 0.4430807885581755,
      "grad_norm": 0.0,
      "learning_rate": 2.7848376497873983e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9170
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.78125,
      "epoch": 0.4431291070738307,
      "grad_norm": 0.0,
      "learning_rate": 2.7845960572091225e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9171
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.4375,
      "epoch": 0.4431774255894859,
      "grad_norm": 0.0,
      "learning_rate": 2.7843544646308466e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9172
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.3125,
      "epoch": 0.4432257441051411,
      "grad_norm": 0.0,
      "learning_rate": 2.78411287205257e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9173
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.375,
      "epoch": 0.44327406262079627,
      "grad_norm": 0.0,
      "learning_rate": 2.783871279474295e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9174
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.78125,
      "epoch": 0.4433223811364515,
      "grad_norm": 0.0,
      "learning_rate": 2.7836296868960185e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9175
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.9375,
      "epoch": 0.4433706996521067,
      "grad_norm": 0.0,
      "learning_rate": 2.7833880943177426e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9176
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.4375,
      "epoch": 0.4434190181677619,
      "grad_norm": 0.0,
      "learning_rate": 2.783146501739466e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9177
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.3125,
      "epoch": 0.4434673366834171,
      "grad_norm": 0.0,
      "learning_rate": 2.7829049091611904e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9178
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.78125,
      "epoch": 0.44351565519907227,
      "grad_norm": 0.0,
      "learning_rate": 2.7826633165829145e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9179
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.96875,
      "epoch": 0.4435639737147275,
      "grad_norm": 0.0,
      "learning_rate": 2.782421724004638e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9180
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.0,
      "epoch": 0.4436122922303827,
      "grad_norm": 0.0,
      "learning_rate": 2.782180131426363e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9181
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.15625,
      "epoch": 0.4436606107460379,
      "grad_norm": 0.0,
      "learning_rate": 2.7819385388480864e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9182
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.84375,
      "epoch": 0.4437089292616931,
      "grad_norm": 0.0,
      "learning_rate": 2.7816969462698105e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9183
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.09375,
      "epoch": 0.44375724777734826,
      "grad_norm": 0.0,
      "learning_rate": 2.7814553536915346e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9184
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.3125,
      "epoch": 0.4438055662930035,
      "grad_norm": 0.0,
      "learning_rate": 2.781213761113259e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9185
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.375,
      "epoch": 0.44385388480865867,
      "grad_norm": 0.0,
      "learning_rate": 2.7809721685349824e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9186
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.4375,
      "epoch": 0.4439022033243139,
      "grad_norm": 0.0,
      "learning_rate": 2.780730575956706e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9187
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.5625,
      "epoch": 0.4439505218399691,
      "grad_norm": 0.0,
      "learning_rate": 2.7804889833784307e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9188
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.46875,
      "epoch": 0.44399884035562426,
      "grad_norm": 0.0,
      "learning_rate": 2.7802473908001543e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9189
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.53125,
      "epoch": 0.4440471588712795,
      "grad_norm": 0.0,
      "learning_rate": 2.7800057982218784e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9190
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.84375,
      "epoch": 0.44409547738693467,
      "grad_norm": 0.0,
      "learning_rate": 2.7797642056436025e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9191
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.25,
      "epoch": 0.4441437959025899,
      "grad_norm": 0.0,
      "learning_rate": 2.7795226130653267e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9192
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.0625,
      "epoch": 0.4441921144182451,
      "grad_norm": 24.0093321561628,
      "learning_rate": 2.7792810204870503e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9193
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.09375,
      "epoch": 0.44424043293390025,
      "grad_norm": 0.0,
      "learning_rate": 2.779039427908775e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9194
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.15625,
      "epoch": 0.4442887514495555,
      "grad_norm": 0.0,
      "learning_rate": 2.7787978353304986e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9195
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.28125,
      "epoch": 0.44433706996521066,
      "grad_norm": 0.0,
      "learning_rate": 2.7785562427522227e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9196
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.34375,
      "epoch": 0.4443853884808659,
      "grad_norm": 0.0,
      "learning_rate": 2.7783146501739463e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 9197
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.3125,
      "epoch": 0.44443370699652107,
      "grad_norm": 0.0,
      "learning_rate": 2.7780730575956704e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9198
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.375,
      "epoch": 0.44448202551217625,
      "grad_norm": 0.0,
      "learning_rate": 2.7778314650173946e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9199
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.40625,
      "epoch": 0.4445303440278315,
      "grad_norm": 0.0,
      "learning_rate": 2.777589872439118e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9200
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.65625,
      "epoch": 0.44457866254348666,
      "grad_norm": 0.0,
      "learning_rate": 2.777348279860843e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9201
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.34375,
      "epoch": 0.4446269810591419,
      "grad_norm": 0.0,
      "learning_rate": 2.7771066872825664e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9202
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.4375,
      "epoch": 0.44467529957479707,
      "grad_norm": 0.0,
      "learning_rate": 2.7768650947042906e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9203
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.40625,
      "epoch": 0.44472361809045224,
      "grad_norm": 0.0,
      "learning_rate": 2.7766235021260147e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9204
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.5,
      "epoch": 0.4447719366061075,
      "grad_norm": 0.0,
      "learning_rate": 2.776381909547739e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23568853735923767,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9205
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.75,
      "epoch": 0.44482025512176265,
      "grad_norm": 0.0,
      "learning_rate": 2.7761403169694625e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9206
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.59375,
      "epoch": 0.4448685736374179,
      "grad_norm": 0.0,
      "learning_rate": 2.775898724391186e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9207
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.46875,
      "epoch": 0.44491689215307306,
      "grad_norm": 0.0,
      "learning_rate": 2.775657131812911e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9208
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 87.34375,
      "epoch": 0.44496521066872824,
      "grad_norm": 58.244351409498805,
      "learning_rate": 2.7754155392346343e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.33725470304489136,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9209
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.875,
      "epoch": 0.44501352918438347,
      "grad_norm": 0.0,
      "learning_rate": 2.7751739466563585e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9210
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.1875,
      "epoch": 0.44506184770003865,
      "grad_norm": 0.0,
      "learning_rate": 2.7749323540780826e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9211
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.28125,
      "epoch": 0.4451101662156939,
      "grad_norm": 0.0,
      "learning_rate": 2.774690761499807e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9212
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.8125,
      "epoch": 0.44515848473134906,
      "grad_norm": 0.0,
      "learning_rate": 2.7744491689215304e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9213
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.1875,
      "epoch": 0.44520680324700423,
      "grad_norm": 0.0,
      "learning_rate": 2.774207576343255e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9214
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.1875,
      "epoch": 0.44525512176265947,
      "grad_norm": 0.0,
      "learning_rate": 2.7739659837649786e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9215
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.21875,
      "epoch": 0.44530344027831464,
      "grad_norm": 0.0,
      "learning_rate": 2.773724391186703e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9216
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.4375,
      "epoch": 0.4453517587939699,
      "grad_norm": 0.0,
      "learning_rate": 2.7734827986084264e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9217
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.5625,
      "epoch": 0.44540007730962505,
      "grad_norm": 0.0,
      "learning_rate": 2.7732412060301505e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9218
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.65625,
      "epoch": 0.44544839582528023,
      "grad_norm": 0.0,
      "learning_rate": 2.7729996134518746e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9219
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.25,
      "epoch": 0.44549671434093546,
      "grad_norm": 0.0,
      "learning_rate": 2.772758020873598e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9220
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.28125,
      "epoch": 0.44554503285659064,
      "grad_norm": 0.0,
      "learning_rate": 2.772516428295323e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9221
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.53125,
      "epoch": 0.44559335137224587,
      "grad_norm": 0.0,
      "learning_rate": 2.7722748357170465e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9222
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.34375,
      "epoch": 0.44564166988790105,
      "grad_norm": 0.0,
      "learning_rate": 2.7720332431387707e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 9223
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.03125,
      "epoch": 0.4456899884035562,
      "grad_norm": 0.0,
      "learning_rate": 2.771791650560495e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9224
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.71875,
      "epoch": 0.44573830691921146,
      "grad_norm": 0.0,
      "learning_rate": 2.771550057982219e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9225
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.28125,
      "epoch": 0.44578662543486663,
      "grad_norm": 0.0,
      "learning_rate": 2.7713084654039425e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.22201895713806152,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9226
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 91.625,
      "epoch": 0.44583494395052187,
      "grad_norm": 66.96423433007,
      "learning_rate": 2.7710668728256667e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.23243674635887146,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9227
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.0625,
      "epoch": 0.44588326246617704,
      "grad_norm": 0.0,
      "learning_rate": 2.770825280247391e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9228
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.25,
      "epoch": 0.4459315809818322,
      "grad_norm": 0.0,
      "learning_rate": 2.7705836876691144e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9229
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.65625,
      "epoch": 0.44597989949748745,
      "grad_norm": 0.0,
      "learning_rate": 2.7703420950908386e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9230
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.875,
      "epoch": 0.44602821801314263,
      "grad_norm": 0.0,
      "learning_rate": 2.7701005025125627e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 9231
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.84375,
      "epoch": 0.44607653652879786,
      "grad_norm": 0.0,
      "learning_rate": 2.769858909934287e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9232
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.84375,
      "epoch": 0.44612485504445304,
      "grad_norm": 0.0,
      "learning_rate": 2.7696173173560104e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9233
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.5625,
      "epoch": 0.4461731735601082,
      "grad_norm": 0.0,
      "learning_rate": 2.769375724777735e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9234
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.28125,
      "epoch": 0.44622149207576345,
      "grad_norm": 0.0,
      "learning_rate": 2.7691341321994587e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9235
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.21875,
      "epoch": 0.4462698105914186,
      "grad_norm": 0.0,
      "learning_rate": 2.768892539621183e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9236
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.90625,
      "epoch": 0.44631812910707386,
      "grad_norm": 0.0,
      "learning_rate": 2.7686509470429065e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9237
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.90625,
      "epoch": 0.44636644762272903,
      "grad_norm": 0.0,
      "learning_rate": 2.768409354464631e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9238
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.0625,
      "epoch": 0.4464147661383842,
      "grad_norm": 0.0,
      "learning_rate": 2.7681677618863547e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9239
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 85.84375,
      "epoch": 0.44646308465403944,
      "grad_norm": 70.32490540524996,
      "learning_rate": 2.7679261693080783e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9240
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.75,
      "epoch": 0.4465114031696946,
      "grad_norm": 0.0,
      "learning_rate": 2.767684576729803e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9241
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.4375,
      "epoch": 0.44655972168534985,
      "grad_norm": 0.0,
      "learning_rate": 2.7674429841515266e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9242
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.03125,
      "epoch": 0.44660804020100503,
      "grad_norm": 0.0,
      "learning_rate": 2.767201391573251e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9243
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.96875,
      "epoch": 0.4466563587166602,
      "grad_norm": 0.0,
      "learning_rate": 2.766959798994975e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 9244
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.21875,
      "epoch": 0.44670467723231544,
      "grad_norm": 0.0,
      "learning_rate": 2.766718206416699e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9245
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.15625,
      "epoch": 0.4467529957479706,
      "grad_norm": 0.0,
      "learning_rate": 2.7664766138384226e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9246
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.3125,
      "epoch": 0.44680131426362585,
      "grad_norm": 0.0,
      "learning_rate": 2.766235021260147e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9247
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.0,
      "epoch": 0.446849632779281,
      "grad_norm": 0.0,
      "learning_rate": 2.765993428681871e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9248
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.03125,
      "epoch": 0.4468979512949362,
      "grad_norm": 0.0,
      "learning_rate": 2.7657518361035945e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9249
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.53125,
      "epoch": 0.44694626981059143,
      "grad_norm": 0.0,
      "learning_rate": 2.7655102435253186e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9250
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.71875,
      "epoch": 0.4469945883262466,
      "grad_norm": 0.0,
      "learning_rate": 2.765268650947043e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9251
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.71875,
      "epoch": 0.44704290684190184,
      "grad_norm": 16.773695457633067,
      "learning_rate": 2.765027058368767e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.2893187999725342,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9252
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.6875,
      "epoch": 0.447091225357557,
      "grad_norm": 45.69664342949498,
      "learning_rate": 2.7647854657904905e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.2877861559391022,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9253
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.5,
      "epoch": 0.4471395438732122,
      "grad_norm": 0.0,
      "learning_rate": 2.764543873212215e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9254
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.84375,
      "epoch": 0.44718786238886743,
      "grad_norm": 0.0,
      "learning_rate": 2.764302280633939e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9255
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.78125,
      "epoch": 0.4472361809045226,
      "grad_norm": 0.0,
      "learning_rate": 2.764060688055663e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9256
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.9375,
      "epoch": 0.44728449942017784,
      "grad_norm": 0.0,
      "learning_rate": 2.7638190954773865e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9257
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.1875,
      "epoch": 0.447332817935833,
      "grad_norm": 0.0,
      "learning_rate": 2.763577502899111e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 9258
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.5625,
      "epoch": 0.4473811364514882,
      "grad_norm": 0.0,
      "learning_rate": 2.763335910320835e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9259
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.40625,
      "epoch": 0.4474294549671434,
      "grad_norm": 0.0,
      "learning_rate": 2.7630943177425584e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9260
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.59375,
      "epoch": 0.4474777734827986,
      "grad_norm": 0.0,
      "learning_rate": 2.762852725164283e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9261
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.03125,
      "epoch": 0.44752609199845383,
      "grad_norm": 0.0,
      "learning_rate": 2.7626111325860067e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9262
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.3125,
      "epoch": 0.447574410514109,
      "grad_norm": 0.0,
      "learning_rate": 2.762369540007731e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9263
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 94.5625,
      "epoch": 0.4476227290297642,
      "grad_norm": 46.92723032956583,
      "learning_rate": 2.762127947429455e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.2982703447341919,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9264
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.53125,
      "epoch": 0.4476710475454194,
      "grad_norm": 0.0,
      "learning_rate": 2.761886354851179e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9265
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.8125,
      "epoch": 0.4477193660610746,
      "grad_norm": 0.0,
      "learning_rate": 2.7616447622729027e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9266
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.21875,
      "epoch": 0.44776768457672983,
      "grad_norm": 0.0,
      "learning_rate": 2.761403169694627e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9267
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.5625,
      "epoch": 0.447816003092385,
      "grad_norm": 0.0,
      "learning_rate": 2.761161577116351e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9268
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.03125,
      "epoch": 0.4478643216080402,
      "grad_norm": 0.0,
      "learning_rate": 2.7609199845380746e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9269
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.9375,
      "epoch": 0.4479126401236954,
      "grad_norm": 0.0,
      "learning_rate": 2.7606783919597987e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9270
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.25,
      "epoch": 0.4479609586393506,
      "grad_norm": 0.0,
      "learning_rate": 2.760436799381523e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9271
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.40625,
      "epoch": 0.4480092771550058,
      "grad_norm": 0.0,
      "learning_rate": 2.760195206803247e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9272
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.375,
      "epoch": 0.448057595670661,
      "grad_norm": 0.0,
      "learning_rate": 2.7599536142249706e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9273
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.1875,
      "epoch": 0.4481059141863162,
      "grad_norm": 0.0,
      "learning_rate": 2.759712021646695e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 9274
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.875,
      "epoch": 0.4481542327019714,
      "grad_norm": 0.0,
      "learning_rate": 2.759470429068419e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9275
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.40625,
      "epoch": 0.4482025512176266,
      "grad_norm": 0.0,
      "learning_rate": 2.759228836490143e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9276
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.78125,
      "epoch": 0.4482508697332818,
      "grad_norm": 0.0,
      "learning_rate": 2.7589872439118666e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9277
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.0625,
      "epoch": 0.448299188248937,
      "grad_norm": 0.0,
      "learning_rate": 2.7587456513335913e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9278
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.53125,
      "epoch": 0.4483475067645922,
      "grad_norm": 0.0,
      "learning_rate": 2.758504058755315e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9279
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.53125,
      "epoch": 0.4483958252802474,
      "grad_norm": 0.0,
      "learning_rate": 2.7582624661770385e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9280
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.0625,
      "epoch": 0.4484441437959026,
      "grad_norm": 0.0,
      "learning_rate": 2.758020873598763e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9281
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.84375,
      "epoch": 0.4484924623115578,
      "grad_norm": 0.0,
      "learning_rate": 2.757779281020487e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9282
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 84.0,
      "epoch": 0.448540780827213,
      "grad_norm": 9.206653933059032,
      "learning_rate": 2.757537688442211e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9283
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.3125,
      "epoch": 0.44858909934286817,
      "grad_norm": 0.0,
      "learning_rate": 2.757296095863935e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9284
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.875,
      "epoch": 0.4486374178585234,
      "grad_norm": 0.0,
      "learning_rate": 2.757054503285659e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9285
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.15625,
      "epoch": 0.4486857363741786,
      "grad_norm": 0.0,
      "learning_rate": 2.756812910707383e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9286
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.09375,
      "epoch": 0.4487340548898338,
      "grad_norm": 0.0,
      "learning_rate": 2.756571318129107e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9287
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.8125,
      "epoch": 0.448782373405489,
      "grad_norm": 0.0,
      "learning_rate": 2.756329725550831e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9288
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.625,
      "epoch": 0.44883069192114416,
      "grad_norm": 0.0,
      "learning_rate": 2.756088132972555e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9289
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.875,
      "epoch": 0.4488790104367994,
      "grad_norm": 0.0,
      "learning_rate": 2.755846540394279e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9290
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.3125,
      "epoch": 0.4489273289524546,
      "grad_norm": 0.0,
      "learning_rate": 2.755604947816003e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 9291
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.03125,
      "epoch": 0.4489756474681098,
      "grad_norm": 0.0,
      "learning_rate": 2.755363355237727e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9292
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.1875,
      "epoch": 0.449023965983765,
      "grad_norm": 0.0,
      "learning_rate": 2.7551217626594507e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9293
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 91.71875,
      "epoch": 0.44907228449942016,
      "grad_norm": 12.830401671390032,
      "learning_rate": 2.7548801700811753e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9294
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.71875,
      "epoch": 0.4491206030150754,
      "grad_norm": 0.0,
      "learning_rate": 2.754638577502899e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9295
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.03125,
      "epoch": 0.44916892153073057,
      "grad_norm": 0.0,
      "learning_rate": 2.754396984924623e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9296
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.875,
      "epoch": 0.4492172400463858,
      "grad_norm": 0.0,
      "learning_rate": 2.7541553923463467e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9297
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.5625,
      "epoch": 0.449265558562041,
      "grad_norm": 0.0,
      "learning_rate": 2.7539137997680714e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9298
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.5625,
      "epoch": 0.44931387707769616,
      "grad_norm": 0.0,
      "learning_rate": 2.753672207189795e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9299
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.96875,
      "epoch": 0.4493621955933514,
      "grad_norm": 0.0,
      "learning_rate": 2.7534306146115186e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 9300
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.875,
      "epoch": 0.44941051410900656,
      "grad_norm": 0.0,
      "learning_rate": 2.753189022033243e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 9301
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.09375,
      "epoch": 0.4494588326246618,
      "grad_norm": 0.0,
      "learning_rate": 2.752947429454967e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9302
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.3125,
      "epoch": 0.449507151140317,
      "grad_norm": 0.0,
      "learning_rate": 2.752705836876691e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 0.96875,
      "step": 9303
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.84375,
      "epoch": 0.44955546965597215,
      "grad_norm": 0.0,
      "learning_rate": 2.752464244298415e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9304
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.4375,
      "epoch": 0.4496037881716274,
      "grad_norm": 0.0,
      "learning_rate": 2.752222651720139e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9305
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.1875,
      "epoch": 0.44965210668728256,
      "grad_norm": 0.0,
      "learning_rate": 2.751981059141863e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9306
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.5,
      "epoch": 0.4497004252029378,
      "grad_norm": 0.0,
      "learning_rate": 2.751739466563587e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9307
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.0,
      "epoch": 0.44974874371859297,
      "grad_norm": 0.0,
      "learning_rate": 2.751497873985311e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9308
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.5625,
      "epoch": 0.44979706223424815,
      "grad_norm": 0.0,
      "learning_rate": 2.7512562814070353e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 9309
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.46875,
      "epoch": 0.4498453807499034,
      "grad_norm": 0.0,
      "learning_rate": 2.751014688828759e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9310
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.8125,
      "epoch": 0.44989369926555856,
      "grad_norm": 0.0,
      "learning_rate": 2.750773096250483e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9311
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.90625,
      "epoch": 0.4499420177812138,
      "grad_norm": 0.0,
      "learning_rate": 2.750531503672207e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9312
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.46875,
      "epoch": 0.44999033629686896,
      "grad_norm": 0.0,
      "learning_rate": 2.750289911093931e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9313
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.28125,
      "epoch": 0.45003865481252414,
      "grad_norm": 0.0,
      "learning_rate": 2.7500483185156554e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24251842498779297,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9314
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.1875,
      "epoch": 0.4500869733281794,
      "grad_norm": 0.0,
      "learning_rate": 2.749806725937379e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9315
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 87.28125,
      "epoch": 0.45013529184383455,
      "grad_norm": 23.317217695285564,
      "learning_rate": 2.749565133359103e-07,
      "loss": -0.0,
      "reward": 1.5625,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9316
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.28125,
      "epoch": 0.4501836103594898,
      "grad_norm": 0.0,
      "learning_rate": 2.749323540780827e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9317
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.84375,
      "epoch": 0.45023192887514496,
      "grad_norm": 0.0,
      "learning_rate": 2.7490819482025514e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9318
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.8125,
      "epoch": 0.45028024739080014,
      "grad_norm": 0.0,
      "learning_rate": 2.748840355624275e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9319
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.25,
      "epoch": 0.45032856590645537,
      "grad_norm": 0.0,
      "learning_rate": 2.748598763045999e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9320
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.375,
      "epoch": 0.45037688442211055,
      "grad_norm": 0.0,
      "learning_rate": 2.7483571704677233e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9321
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.03125,
      "epoch": 0.4504252029377658,
      "grad_norm": 0.0,
      "learning_rate": 2.748115577889447e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9322
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.78125,
      "epoch": 0.45047352145342096,
      "grad_norm": 0.0,
      "learning_rate": 2.747873985311171e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9323
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.71875,
      "epoch": 0.45052183996907613,
      "grad_norm": 0.0,
      "learning_rate": 2.747632392732895e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9324
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.21875,
      "epoch": 0.45057015848473136,
      "grad_norm": 0.0,
      "learning_rate": 2.7473908001546193e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 9325
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.3125,
      "epoch": 0.45061847700038654,
      "grad_norm": 0.0,
      "learning_rate": 2.747149207576343e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9326
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.34375,
      "epoch": 0.4506667955160418,
      "grad_norm": 0.0,
      "learning_rate": 2.746907614998067e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9327
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.90625,
      "epoch": 0.45071511403169695,
      "grad_norm": 0.0,
      "learning_rate": 2.746666022419791e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9328
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.90625,
      "epoch": 0.4507634325473521,
      "grad_norm": 0.0,
      "learning_rate": 2.7464244298415153e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9329
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.03125,
      "epoch": 0.45081175106300736,
      "grad_norm": 54.7435534372823,
      "learning_rate": 2.746182837263239e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.3567763864994049,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9330
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.09375,
      "epoch": 0.45086006957866254,
      "grad_norm": 0.0,
      "learning_rate": 2.7459412446849636e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9331
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.90625,
      "epoch": 0.45090838809431777,
      "grad_norm": 0.0,
      "learning_rate": 2.745699652106687e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9332
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.0625,
      "epoch": 0.45095670660997295,
      "grad_norm": 0.0,
      "learning_rate": 2.745458059528411e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9333
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.9375,
      "epoch": 0.4510050251256281,
      "grad_norm": 0.0,
      "learning_rate": 2.7452164669501355e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9334
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.375,
      "epoch": 0.45105334364128336,
      "grad_norm": 0.0,
      "learning_rate": 2.744974874371859e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9335
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.21875,
      "epoch": 0.45110166215693853,
      "grad_norm": 0.0,
      "learning_rate": 2.744733281793583e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9336
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.8125,
      "epoch": 0.45114998067259376,
      "grad_norm": 0.0,
      "learning_rate": 2.744491689215307e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9337
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.6875,
      "epoch": 0.45119829918824894,
      "grad_norm": 0.0,
      "learning_rate": 2.7442500966370315e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9338
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.5,
      "epoch": 0.4512466177039041,
      "grad_norm": 0.0,
      "learning_rate": 2.744008504058755e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9339
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.875,
      "epoch": 0.45129493621955935,
      "grad_norm": 0.0,
      "learning_rate": 2.743766911480479e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9340
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.21875,
      "epoch": 0.4513432547352145,
      "grad_norm": 0.0,
      "learning_rate": 2.7435253189022034e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9341
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.21875,
      "epoch": 0.45139157325086976,
      "grad_norm": 0.0,
      "learning_rate": 2.743283726323927e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9342
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.59375,
      "epoch": 0.45143989176652494,
      "grad_norm": 0.0,
      "learning_rate": 2.743042133745651e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9343
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 83.5625,
      "epoch": 0.4514882102821801,
      "grad_norm": 0.0,
      "learning_rate": 2.7428005411673753e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 9344
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.6875,
      "epoch": 0.45153652879783535,
      "grad_norm": 0.0,
      "learning_rate": 2.7425589485890994e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9345
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.5625,
      "epoch": 0.4515848473134905,
      "grad_norm": 0.0,
      "learning_rate": 2.742317356010823e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9346
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.9375,
      "epoch": 0.45163316582914576,
      "grad_norm": 0.0,
      "learning_rate": 2.742075763432547e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9347
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.3125,
      "epoch": 0.45168148434480093,
      "grad_norm": 0.0,
      "learning_rate": 2.7418341708542713e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9348
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.625,
      "epoch": 0.4517298028604561,
      "grad_norm": 0.0,
      "learning_rate": 2.7415925782759954e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9349
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.8125,
      "epoch": 0.45177812137611134,
      "grad_norm": 0.0,
      "learning_rate": 2.741350985697719e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9350
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.5,
      "epoch": 0.4518264398917665,
      "grad_norm": 0.0,
      "learning_rate": 2.7411093931194437e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9351
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.78125,
      "epoch": 0.45187475840742175,
      "grad_norm": 0.0,
      "learning_rate": 2.7408678005411673e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9352
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.5,
      "epoch": 0.4519230769230769,
      "grad_norm": 0.0,
      "learning_rate": 2.740626207962891e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9353
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.8125,
      "epoch": 0.4519713954387321,
      "grad_norm": 0.0,
      "learning_rate": 2.7403846153846156e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9354
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.25,
      "epoch": 0.45201971395438734,
      "grad_norm": 0.0,
      "learning_rate": 2.740143022806339e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9355
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.875,
      "epoch": 0.4520680324700425,
      "grad_norm": 0.0,
      "learning_rate": 2.7399014302280633e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9356
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.75,
      "epoch": 0.45211635098569775,
      "grad_norm": 0.0,
      "learning_rate": 2.739659837649787e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9357
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.375,
      "epoch": 0.4521646695013529,
      "grad_norm": 0.0,
      "learning_rate": 2.7394182450715116e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9358
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.90625,
      "epoch": 0.4522129880170081,
      "grad_norm": 0.0,
      "learning_rate": 2.739176652493235e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9359
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.625,
      "epoch": 0.45226130653266333,
      "grad_norm": 0.0,
      "learning_rate": 2.7389350599149593e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9360
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.96875,
      "epoch": 0.4523096250483185,
      "grad_norm": 0.0,
      "learning_rate": 2.7386934673366835e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9361
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.59375,
      "epoch": 0.45235794356397374,
      "grad_norm": 0.0,
      "learning_rate": 2.738451874758407e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9362
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.4375,
      "epoch": 0.4524062620796289,
      "grad_norm": 0.0,
      "learning_rate": 2.738210282180131e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9363
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.0625,
      "epoch": 0.4524545805952841,
      "grad_norm": 0.0,
      "learning_rate": 2.7379686896018554e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9364
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.46875,
      "epoch": 0.4525028991109393,
      "grad_norm": 0.0,
      "learning_rate": 2.7377270970235795e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9365
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 92.125,
      "epoch": 0.4525512176265945,
      "grad_norm": 0.0,
      "learning_rate": 2.737485504445303e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 9366
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.46875,
      "epoch": 0.45259953614224974,
      "grad_norm": 0.0,
      "learning_rate": 2.737243911867027e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9367
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.40625,
      "epoch": 0.4526478546579049,
      "grad_norm": 0.0,
      "learning_rate": 2.7370023192887514e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9368
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.78125,
      "epoch": 0.4526961731735601,
      "grad_norm": 0.0,
      "learning_rate": 2.7367607267104755e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 9369
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.40625,
      "epoch": 0.4527444916892153,
      "grad_norm": 0.0,
      "learning_rate": 2.736519134132199e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9370
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.0,
      "epoch": 0.4527928102048705,
      "grad_norm": 0.0,
      "learning_rate": 2.736277541553924e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9371
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.34375,
      "epoch": 0.45284112872052573,
      "grad_norm": 0.0,
      "learning_rate": 2.7360359489756474e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9372
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.8125,
      "epoch": 0.4528894472361809,
      "grad_norm": 0.0,
      "learning_rate": 2.735794356397371e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9373
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.9375,
      "epoch": 0.4529377657518361,
      "grad_norm": 0.0,
      "learning_rate": 2.7355527638190957e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9374
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.84375,
      "epoch": 0.4529860842674913,
      "grad_norm": 0.0,
      "learning_rate": 2.735311171240819e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9375
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.6875,
      "epoch": 0.4530344027831465,
      "grad_norm": 0.0,
      "learning_rate": 2.7350695786625434e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9376
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.15625,
      "epoch": 0.4530827212988017,
      "grad_norm": 0.0,
      "learning_rate": 2.734827986084267e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9377
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 88.15625,
      "epoch": 0.4531310398144569,
      "grad_norm": 52.8373175748201,
      "learning_rate": 2.7345863935059917e-07,
      "loss": 0.0,
      "reward": 1.5625,
      "reward_std": 0.24386970698833466,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9378
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.25,
      "epoch": 0.4531793583301121,
      "grad_norm": 0.0,
      "learning_rate": 2.7343448009277153e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9379
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.625,
      "epoch": 0.4532276768457673,
      "grad_norm": 0.0,
      "learning_rate": 2.7341032083494394e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9380
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.5625,
      "epoch": 0.4532759953614225,
      "grad_norm": 0.0,
      "learning_rate": 2.7338616157711636e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9381
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.53125,
      "epoch": 0.4533243138770777,
      "grad_norm": 0.0,
      "learning_rate": 2.7336200231928877e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9382
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.5,
      "epoch": 0.4533726323927329,
      "grad_norm": 0.0,
      "learning_rate": 2.7333784306146113e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 9383
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.65625,
      "epoch": 0.4534209509083881,
      "grad_norm": 0.0,
      "learning_rate": 2.7331368380363354e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9384
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.125,
      "epoch": 0.4534692694240433,
      "grad_norm": 0.0,
      "learning_rate": 2.7328952454580596e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9385
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.90625,
      "epoch": 0.4535175879396985,
      "grad_norm": 0.0,
      "learning_rate": 2.732653652879783e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 9386
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.09375,
      "epoch": 0.4535659064553537,
      "grad_norm": 0.0,
      "learning_rate": 2.7324120603015073e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9387
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.40625,
      "epoch": 0.4536142249710089,
      "grad_norm": 0.0,
      "learning_rate": 2.7321704677232314e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.2041158676147461,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9388
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.625,
      "epoch": 0.45366254348666407,
      "grad_norm": 0.0,
      "learning_rate": 2.7319288751449556e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9389
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.09375,
      "epoch": 0.4537108620023193,
      "grad_norm": 0.0,
      "learning_rate": 2.731687282566679e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 9390
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.75,
      "epoch": 0.4537591805179745,
      "grad_norm": 103.10996959465012,
      "learning_rate": 2.731445689988404e-07,
      "loss": -0.0,
      "reward": 2.359375,
      "reward_std": 0.2993341088294983,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9391
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.71875,
      "epoch": 0.4538074990336297,
      "grad_norm": 0.0,
      "learning_rate": 2.7312040974101275e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 9392
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.09375,
      "epoch": 0.4538558175492849,
      "grad_norm": 0.0,
      "learning_rate": 2.730962504831851e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 9393
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.78125,
      "epoch": 0.45390413606494007,
      "grad_norm": 0.0,
      "learning_rate": 2.7307209122535757e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9394
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.46875,
      "epoch": 0.4539524545805953,
      "grad_norm": 0.0,
      "learning_rate": 2.7304793196752993e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9395
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.375,
      "epoch": 0.4540007730962505,
      "grad_norm": 0.0,
      "learning_rate": 2.7302377270970235e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9396
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.15625,
      "epoch": 0.4540490916119057,
      "grad_norm": 0.0,
      "learning_rate": 2.729996134518747e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9397
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.53125,
      "epoch": 0.4540974101275609,
      "grad_norm": 0.0,
      "learning_rate": 2.729754541940472e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9398
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.46875,
      "epoch": 0.45414572864321606,
      "grad_norm": 0.0,
      "learning_rate": 2.7295129493621954e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9399
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.6875,
      "epoch": 0.4541940471588713,
      "grad_norm": 0.0,
      "learning_rate": 2.7292713567839195e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9400
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.4375,
      "epoch": 0.45424236567452647,
      "grad_norm": 0.0,
      "learning_rate": 2.7290297642056436e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9401
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.9375,
      "epoch": 0.4542906841901817,
      "grad_norm": 0.0,
      "learning_rate": 2.728788171627368e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9402
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.4375,
      "epoch": 0.4543390027058369,
      "grad_norm": 0.0,
      "learning_rate": 2.7285465790490914e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9403
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.4375,
      "epoch": 0.45438732122149206,
      "grad_norm": 0.0,
      "learning_rate": 2.7283049864708155e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 9404
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.28125,
      "epoch": 0.4544356397371473,
      "grad_norm": 0.0,
      "learning_rate": 2.7280633938925396e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9405
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.875,
      "epoch": 0.45448395825280247,
      "grad_norm": 0.0,
      "learning_rate": 2.727821801314263e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9406
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.90625,
      "epoch": 0.4545322767684577,
      "grad_norm": 0.0,
      "learning_rate": 2.7275802087359874e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9407
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.125,
      "epoch": 0.4545805952841129,
      "grad_norm": 0.0,
      "learning_rate": 2.7273386161577115e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9408
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.5625,
      "epoch": 0.45462891379976805,
      "grad_norm": 0.0,
      "learning_rate": 2.7270970235794357e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9409
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.03125,
      "epoch": 0.4546772323154233,
      "grad_norm": 0.0,
      "learning_rate": 2.7268554310011593e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9410
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.375,
      "epoch": 0.45472555083107846,
      "grad_norm": 0.0,
      "learning_rate": 2.726613838422884e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9411
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.03125,
      "epoch": 0.4547738693467337,
      "grad_norm": 0.0,
      "learning_rate": 2.7263722458446075e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9412
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.21875,
      "epoch": 0.45482218786238887,
      "grad_norm": 0.0,
      "learning_rate": 2.726130653266331e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9413
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.78125,
      "epoch": 0.45487050637804405,
      "grad_norm": 0.0,
      "learning_rate": 2.725889060688056e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9414
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.8125,
      "epoch": 0.4549188248936993,
      "grad_norm": 0.0,
      "learning_rate": 2.7256474681097794e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9415
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.6875,
      "epoch": 0.45496714340935446,
      "grad_norm": 0.0,
      "learning_rate": 2.7254058755315036e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9416
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.84375,
      "epoch": 0.4550154619250097,
      "grad_norm": 0.0,
      "learning_rate": 2.725164282953227e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9417
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.5625,
      "epoch": 0.45506378044066487,
      "grad_norm": 0.0,
      "learning_rate": 2.724922690374952e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9418
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.375,
      "epoch": 0.45511209895632004,
      "grad_norm": 0.0,
      "learning_rate": 2.7246810977966754e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9419
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.25,
      "epoch": 0.4551604174719753,
      "grad_norm": 0.0,
      "learning_rate": 2.7244395052183996e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9420
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.25,
      "epoch": 0.45520873598763045,
      "grad_norm": 0.0,
      "learning_rate": 2.7241979126401237e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9421
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.5,
      "epoch": 0.4552570545032857,
      "grad_norm": 0.0,
      "learning_rate": 2.723956320061848e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9422
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.09375,
      "epoch": 0.45530537301894086,
      "grad_norm": 0.0,
      "learning_rate": 2.7237147274835715e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9423
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.84375,
      "epoch": 0.45535369153459604,
      "grad_norm": 0.0,
      "learning_rate": 2.7234731349052956e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9424
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.03125,
      "epoch": 0.45540201005025127,
      "grad_norm": 0.0,
      "learning_rate": 2.7232315423270197e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9425
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.84375,
      "epoch": 0.45545032856590645,
      "grad_norm": 0.0,
      "learning_rate": 2.7229899497487433e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2630178928375244,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9426
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.0,
      "epoch": 0.4554986470815617,
      "grad_norm": 0.0,
      "learning_rate": 2.7227483571704675e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9427
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.78125,
      "epoch": 0.45554696559721686,
      "grad_norm": 0.0,
      "learning_rate": 2.7225067645921916e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9428
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.5,
      "epoch": 0.45559528411287203,
      "grad_norm": 0.0,
      "learning_rate": 2.722265172013916e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9429
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.6875,
      "epoch": 0.45564360262852727,
      "grad_norm": 0.0,
      "learning_rate": 2.7220235794356393e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9430
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.625,
      "epoch": 0.45569192114418244,
      "grad_norm": 0.0,
      "learning_rate": 2.721781986857364e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9431
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.625,
      "epoch": 0.4557402396598377,
      "grad_norm": 0.0,
      "learning_rate": 2.7215403942790876e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9432
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.4375,
      "epoch": 0.45578855817549285,
      "grad_norm": 0.0,
      "learning_rate": 2.721298801700812e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9433
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.75,
      "epoch": 0.45583687669114803,
      "grad_norm": 0.0,
      "learning_rate": 2.721057209122536e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9434
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.6875,
      "epoch": 0.45588519520680326,
      "grad_norm": 0.0,
      "learning_rate": 2.7208156165442595e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9435
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.65625,
      "epoch": 0.45593351372245844,
      "grad_norm": 0.0,
      "learning_rate": 2.7205740239659836e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9436
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 91.1875,
      "epoch": 0.45598183223811367,
      "grad_norm": 0.0,
      "learning_rate": 2.720332431387707e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 9437
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.6875,
      "epoch": 0.45603015075376885,
      "grad_norm": 0.0,
      "learning_rate": 2.720090838809432e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9438
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.96875,
      "epoch": 0.456078469269424,
      "grad_norm": 0.0,
      "learning_rate": 2.7198492462311555e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9439
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.75,
      "epoch": 0.45612678778507926,
      "grad_norm": 0.0,
      "learning_rate": 2.7196076536528796e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9440
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.71875,
      "epoch": 0.45617510630073443,
      "grad_norm": 0.0,
      "learning_rate": 2.719366061074604e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9441
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.40625,
      "epoch": 0.45622342481638967,
      "grad_norm": 0.0,
      "learning_rate": 2.719124468496328e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23568853735923767,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9442
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.28125,
      "epoch": 0.45627174333204484,
      "grad_norm": 0.0,
      "learning_rate": 2.7188828759180515e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9443
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.59375,
      "epoch": 0.4563200618477,
      "grad_norm": 0.0,
      "learning_rate": 2.718641283339776e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9444
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.34375,
      "epoch": 0.45636838036335525,
      "grad_norm": 0.0,
      "learning_rate": 2.7183996907615e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9445
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.9375,
      "epoch": 0.45641669887901043,
      "grad_norm": 0.0,
      "learning_rate": 2.7181580981832234e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9446
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.125,
      "epoch": 0.45646501739466566,
      "grad_norm": 0.0,
      "learning_rate": 2.7179165056049475e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9447
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.75,
      "epoch": 0.45651333591032084,
      "grad_norm": 0.0,
      "learning_rate": 2.7176749130266717e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9448
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.28125,
      "epoch": 0.456561654425976,
      "grad_norm": 0.0,
      "learning_rate": 2.717433320448396e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9449
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.71875,
      "epoch": 0.45660997294163125,
      "grad_norm": 168.9401232630829,
      "learning_rate": 2.7171917278701194e-07,
      "loss": -0.0,
      "reward": 1.875,
      "reward_std": 0.2529362142086029,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9450
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.78125,
      "epoch": 0.4566582914572864,
      "grad_norm": 0.0,
      "learning_rate": 2.716950135291844e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9451
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.9375,
      "epoch": 0.45670660997294166,
      "grad_norm": 0.0,
      "learning_rate": 2.7167085427135677e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9452
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.75,
      "epoch": 0.45675492848859683,
      "grad_norm": 0.0,
      "learning_rate": 2.716466950135292e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9453
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.4375,
      "epoch": 0.456803247004252,
      "grad_norm": 0.0,
      "learning_rate": 2.716225357557016e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9454
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.4375,
      "epoch": 0.45685156551990724,
      "grad_norm": 0.0,
      "learning_rate": 2.7159837649787396e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9455
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.03125,
      "epoch": 0.4568998840355624,
      "grad_norm": 0.0,
      "learning_rate": 2.7157421724004637e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9456
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.03125,
      "epoch": 0.45694820255121765,
      "grad_norm": 0.0,
      "learning_rate": 2.7155005798221873e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9457
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.8125,
      "epoch": 0.45699652106687283,
      "grad_norm": 0.0,
      "learning_rate": 2.715258987243912e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9458
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.96875,
      "epoch": 0.457044839582528,
      "grad_norm": 0.0,
      "learning_rate": 2.7150173946656356e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9459
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.125,
      "epoch": 0.45709315809818324,
      "grad_norm": 0.0,
      "learning_rate": 2.7147758020873597e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9460
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.25,
      "epoch": 0.4571414766138384,
      "grad_norm": 0.0,
      "learning_rate": 2.714534209509084e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9461
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.59375,
      "epoch": 0.45718979512949365,
      "grad_norm": 0.0,
      "learning_rate": 2.714292616930808e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9462
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.40625,
      "epoch": 0.4572381136451488,
      "grad_norm": 0.0,
      "learning_rate": 2.7140510243525316e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9463
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.5,
      "epoch": 0.457286432160804,
      "grad_norm": 0.0,
      "learning_rate": 2.7138094317742563e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9464
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.5625,
      "epoch": 0.45733475067645923,
      "grad_norm": 0.0,
      "learning_rate": 2.71356783919598e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9465
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.4375,
      "epoch": 0.4573830691921144,
      "grad_norm": 0.0,
      "learning_rate": 2.7133262466177035e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9466
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.4375,
      "epoch": 0.45743138770776964,
      "grad_norm": 0.0,
      "learning_rate": 2.7130846540394276e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9467
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.0,
      "epoch": 0.4574797062234248,
      "grad_norm": 0.0,
      "learning_rate": 2.712843061461152e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9468
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.21875,
      "epoch": 0.45752802473908,
      "grad_norm": 0.0,
      "learning_rate": 2.712601468882876e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9469
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.28125,
      "epoch": 0.45757634325473523,
      "grad_norm": 0.0,
      "learning_rate": 2.7123598763045995e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9470
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.53125,
      "epoch": 0.4576246617703904,
      "grad_norm": 0.0,
      "learning_rate": 2.712118283726324e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9471
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.25,
      "epoch": 0.45767298028604564,
      "grad_norm": 0.0,
      "learning_rate": 2.711876691148048e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9472
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.0,
      "epoch": 0.4577212988017008,
      "grad_norm": 0.0,
      "learning_rate": 2.711635098569772e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9473
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.84375,
      "epoch": 0.457769617317356,
      "grad_norm": 0.0,
      "learning_rate": 2.711393505991496e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9474
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.25,
      "epoch": 0.4578179358330112,
      "grad_norm": 0.0,
      "learning_rate": 2.7111519134132197e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9475
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.21875,
      "epoch": 0.4578662543486664,
      "grad_norm": 0.0,
      "learning_rate": 2.710910320834944e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9476
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.5625,
      "epoch": 0.45791457286432163,
      "grad_norm": 0.0,
      "learning_rate": 2.7106687282566674e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9477
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.3125,
      "epoch": 0.4579628913799768,
      "grad_norm": 0.0,
      "learning_rate": 2.710427135678392e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9478
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.75,
      "epoch": 0.458011209895632,
      "grad_norm": 0.0,
      "learning_rate": 2.7101855431001157e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9479
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.625,
      "epoch": 0.4580595284112872,
      "grad_norm": 0.0,
      "learning_rate": 2.70994395052184e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9480
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.78125,
      "epoch": 0.4581078469269424,
      "grad_norm": 0.0,
      "learning_rate": 2.709702357943564e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9481
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.59375,
      "epoch": 0.45815616544259763,
      "grad_norm": 0.0,
      "learning_rate": 2.709460765365288e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9482
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.6875,
      "epoch": 0.4582044839582528,
      "grad_norm": 0.0,
      "learning_rate": 2.7092191727870117e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 9483
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.09375,
      "epoch": 0.458252802473908,
      "grad_norm": 0.0,
      "learning_rate": 2.7089775802087364e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9484
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.03125,
      "epoch": 0.4583011209895632,
      "grad_norm": 0.0,
      "learning_rate": 2.70873598763046e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9485
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.625,
      "epoch": 0.4583494395052184,
      "grad_norm": 0.0,
      "learning_rate": 2.7084943950521836e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9486
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.65625,
      "epoch": 0.4583977580208736,
      "grad_norm": 0.0,
      "learning_rate": 2.7082528024739077e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9487
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.90625,
      "epoch": 0.4584460765365288,
      "grad_norm": 0.0,
      "learning_rate": 2.708011209895632e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9488
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.5,
      "epoch": 0.458494395052184,
      "grad_norm": 0.0,
      "learning_rate": 2.707769617317356e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9489
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.6875,
      "epoch": 0.4585427135678392,
      "grad_norm": 0.0,
      "learning_rate": 2.7075280247390796e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9490
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.125,
      "epoch": 0.4585910320834944,
      "grad_norm": 0.0,
      "learning_rate": 2.707286432160804e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9491
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.6875,
      "epoch": 0.4586393505991496,
      "grad_norm": 0.0,
      "learning_rate": 2.707044839582528e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9492
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.53125,
      "epoch": 0.4586876691148048,
      "grad_norm": 0.0,
      "learning_rate": 2.706803247004252e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9493
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.9375,
      "epoch": 0.45873598763046,
      "grad_norm": 0.0,
      "learning_rate": 2.706561654425976e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9494
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.625,
      "epoch": 0.4587843061461152,
      "grad_norm": 0.0,
      "learning_rate": 2.7063200618477e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9495
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.46875,
      "epoch": 0.4588326246617704,
      "grad_norm": 0.0,
      "learning_rate": 2.706078469269424e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9496
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.0625,
      "epoch": 0.4588809431774256,
      "grad_norm": 0.0,
      "learning_rate": 2.7058368766911475e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9497
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.8125,
      "epoch": 0.4589292616930808,
      "grad_norm": 0.0,
      "learning_rate": 2.705595284112872e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9498
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.40625,
      "epoch": 0.45897758020873597,
      "grad_norm": 0.0,
      "learning_rate": 2.705353691534596e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9499
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.5625,
      "epoch": 0.4590258987243912,
      "grad_norm": 0.0,
      "learning_rate": 2.70511209895632e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9500
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.96875,
      "epoch": 0.4590742172400464,
      "grad_norm": 0.0,
      "learning_rate": 2.704870506378044e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9501
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.65625,
      "epoch": 0.4591225357557016,
      "grad_norm": 0.0,
      "learning_rate": 2.704628913799768e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9502
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.125,
      "epoch": 0.4591708542713568,
      "grad_norm": 0.0,
      "learning_rate": 2.704387321221492e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9503
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.75,
      "epoch": 0.45921917278701196,
      "grad_norm": 0.0,
      "learning_rate": 2.7041457286432164e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9504
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.6875,
      "epoch": 0.4592674913026672,
      "grad_norm": 0.0,
      "learning_rate": 2.70390413606494e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9505
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.875,
      "epoch": 0.4593158098183224,
      "grad_norm": 0.0,
      "learning_rate": 2.7036625434866636e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9506
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.5625,
      "epoch": 0.4593641283339776,
      "grad_norm": 0.0,
      "learning_rate": 2.703420950908388e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9507
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.5,
      "epoch": 0.4594124468496328,
      "grad_norm": 0.0,
      "learning_rate": 2.703179358330112e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9508
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.71875,
      "epoch": 0.45946076536528796,
      "grad_norm": 0.0,
      "learning_rate": 2.702937765751836e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9509
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.53125,
      "epoch": 0.4595090838809432,
      "grad_norm": 0.0,
      "learning_rate": 2.7026961731735597e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9510
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.8125,
      "epoch": 0.45955740239659837,
      "grad_norm": 0.0,
      "learning_rate": 2.7024545805952843e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9511
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.375,
      "epoch": 0.4596057209122536,
      "grad_norm": 0.0,
      "learning_rate": 2.702212988017008e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9512
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.46875,
      "epoch": 0.4596540394279088,
      "grad_norm": 0.0,
      "learning_rate": 2.701971395438732e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9513
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.25,
      "epoch": 0.45970235794356396,
      "grad_norm": 0.0,
      "learning_rate": 2.701729802860456e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9514
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.65625,
      "epoch": 0.4597506764592192,
      "grad_norm": 0.0,
      "learning_rate": 2.7014882102821803e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9515
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.5625,
      "epoch": 0.45979899497487436,
      "grad_norm": 0.0,
      "learning_rate": 2.701246617703904e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9516
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.40625,
      "epoch": 0.4598473134905296,
      "grad_norm": 0.0,
      "learning_rate": 2.7010050251256276e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9517
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.625,
      "epoch": 0.4598956320061848,
      "grad_norm": 0.0,
      "learning_rate": 2.700763432547352e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9518
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.4375,
      "epoch": 0.45994395052183995,
      "grad_norm": 0.0,
      "learning_rate": 2.700521839969076e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9519
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.875,
      "epoch": 0.4599922690374952,
      "grad_norm": 0.0,
      "learning_rate": 2.7002802473908e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9520
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.125,
      "epoch": 0.46004058755315036,
      "grad_norm": 0.0,
      "learning_rate": 2.700038654812524e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9521
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.28125,
      "epoch": 0.4600889060688056,
      "grad_norm": 0.0,
      "learning_rate": 2.699797062234248e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9522
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.1875,
      "epoch": 0.46013722458446077,
      "grad_norm": 0.0,
      "learning_rate": 2.699555469655972e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9523
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.5,
      "epoch": 0.46018554310011595,
      "grad_norm": 0.0,
      "learning_rate": 2.6993138770776965e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9524
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.84375,
      "epoch": 0.4602338616157712,
      "grad_norm": 0.0,
      "learning_rate": 2.69907228449942e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9525
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.75,
      "epoch": 0.46028218013142636,
      "grad_norm": 0.0,
      "learning_rate": 2.6988306919211437e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9526
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.375,
      "epoch": 0.4603304986470816,
      "grad_norm": 0.0,
      "learning_rate": 2.698589099342868e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9527
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.40625,
      "epoch": 0.46037881716273676,
      "grad_norm": 0.0,
      "learning_rate": 2.698347506764592e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9528
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.53125,
      "epoch": 0.46042713567839194,
      "grad_norm": 0.0,
      "learning_rate": 2.698105914186316e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9529
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.28125,
      "epoch": 0.4604754541940472,
      "grad_norm": 0.0,
      "learning_rate": 2.69786432160804e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9530
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.1875,
      "epoch": 0.46052377270970235,
      "grad_norm": 0.0,
      "learning_rate": 2.6976227290297644e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 9531
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.28125,
      "epoch": 0.4605720912253576,
      "grad_norm": 0.0,
      "learning_rate": 2.697381136451488e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9532
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.84375,
      "epoch": 0.46062040974101276,
      "grad_norm": 0.0,
      "learning_rate": 2.697139543873212e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9533
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.84375,
      "epoch": 0.46066872825666794,
      "grad_norm": 0.0,
      "learning_rate": 2.6968979512949363e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9534
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.84375,
      "epoch": 0.46071704677232317,
      "grad_norm": 0.0,
      "learning_rate": 2.6966563587166604e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9535
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.75,
      "epoch": 0.46076536528797835,
      "grad_norm": 0.0,
      "learning_rate": 2.696414766138384e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9536
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.46875,
      "epoch": 0.4608136838036336,
      "grad_norm": 59.98074887779461,
      "learning_rate": 2.6961731735601076e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.28247910737991333,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9537
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.65625,
      "epoch": 0.46086200231928875,
      "grad_norm": 0.0,
      "learning_rate": 2.6959315809818323e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9538
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.75,
      "epoch": 0.46091032083494393,
      "grad_norm": 0.0,
      "learning_rate": 2.695689988403556e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9539
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.0,
      "epoch": 0.46095863935059916,
      "grad_norm": 0.0,
      "learning_rate": 2.69544839582528e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9540
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.90625,
      "epoch": 0.46100695786625434,
      "grad_norm": 0.0,
      "learning_rate": 2.695206803247004e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9541
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.9375,
      "epoch": 0.4610552763819096,
      "grad_norm": 0.0,
      "learning_rate": 2.6949652106687283e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9542
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.0625,
      "epoch": 0.46110359489756475,
      "grad_norm": 0.0,
      "learning_rate": 2.694723618090452e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9543
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.25,
      "epoch": 0.4611519134132199,
      "grad_norm": 0.0,
      "learning_rate": 2.6944820255121766e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9544
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.90625,
      "epoch": 0.46120023192887516,
      "grad_norm": 0.0,
      "learning_rate": 2.6942404329339e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9545
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.125,
      "epoch": 0.46124855044453034,
      "grad_norm": 0.0,
      "learning_rate": 2.6939988403556243e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9546
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.6875,
      "epoch": 0.46129686896018557,
      "grad_norm": 0.0,
      "learning_rate": 2.693757247777348e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9547
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.59375,
      "epoch": 0.46134518747584075,
      "grad_norm": 0.0,
      "learning_rate": 2.693515655199072e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9548
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.09375,
      "epoch": 0.4613935059914959,
      "grad_norm": 0.0,
      "learning_rate": 2.693274062620796e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 9549
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.625,
      "epoch": 0.46144182450715115,
      "grad_norm": 0.0,
      "learning_rate": 2.69303247004252e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9550
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.375,
      "epoch": 0.46149014302280633,
      "grad_norm": 0.0,
      "learning_rate": 2.6927908774642445e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9551
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.71875,
      "epoch": 0.46153846153846156,
      "grad_norm": 0.0,
      "learning_rate": 2.692549284885968e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 9552
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.28125,
      "epoch": 0.46158678005411674,
      "grad_norm": 0.0,
      "learning_rate": 2.692307692307692e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9553
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.71875,
      "epoch": 0.4616350985697719,
      "grad_norm": 0.0,
      "learning_rate": 2.6920660997294164e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9554
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.59375,
      "epoch": 0.46168341708542715,
      "grad_norm": 0.0,
      "learning_rate": 2.6918245071511405e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9555
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.28125,
      "epoch": 0.4617317356010823,
      "grad_norm": 0.0,
      "learning_rate": 2.691582914572864e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9556
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.65625,
      "epoch": 0.46178005411673756,
      "grad_norm": 0.0,
      "learning_rate": 2.6913413219945877e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9557
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.125,
      "epoch": 0.46182837263239274,
      "grad_norm": 0.0,
      "learning_rate": 2.6910997294163124e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9558
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.3125,
      "epoch": 0.4618766911480479,
      "grad_norm": 0.0,
      "learning_rate": 2.690858136838036e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9559
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.3125,
      "epoch": 0.46192500966370315,
      "grad_norm": 0.0,
      "learning_rate": 2.69061654425976e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9560
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.4375,
      "epoch": 0.4619733281793583,
      "grad_norm": 0.0,
      "learning_rate": 2.690374951681484e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9561
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.5625,
      "epoch": 0.46202164669501355,
      "grad_norm": 0.0,
      "learning_rate": 2.6901333591032084e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9562
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.21875,
      "epoch": 0.46206996521066873,
      "grad_norm": 0.0,
      "learning_rate": 2.689891766524932e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9563
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.6875,
      "epoch": 0.4621182837263239,
      "grad_norm": 0.0,
      "learning_rate": 2.6896501739466567e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9564
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.46875,
      "epoch": 0.46216660224197914,
      "grad_norm": 0.0,
      "learning_rate": 2.6894085813683803e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9565
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.84375,
      "epoch": 0.4622149207576343,
      "grad_norm": 0.0,
      "learning_rate": 2.6891669887901044e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9566
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.25,
      "epoch": 0.46226323927328955,
      "grad_norm": 0.0,
      "learning_rate": 2.688925396211828e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9567
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.03125,
      "epoch": 0.4623115577889447,
      "grad_norm": 0.0,
      "learning_rate": 2.688683803633552e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9568
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 96.46875,
      "epoch": 0.4623598763045999,
      "grad_norm": 17.680273699292254,
      "learning_rate": 2.6884422110552763e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.2873119115829468,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9569
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.40625,
      "epoch": 0.46240819482025514,
      "grad_norm": 0.0,
      "learning_rate": 2.688200618477e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9570
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 84.5625,
      "epoch": 0.4624565133359103,
      "grad_norm": 0.0,
      "learning_rate": 2.6879590258987246e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 9571
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.1875,
      "epoch": 0.46250483185156555,
      "grad_norm": 0.0,
      "learning_rate": 2.687717433320448e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9572
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.84375,
      "epoch": 0.4625531503672207,
      "grad_norm": 0.0,
      "learning_rate": 2.6874758407421723e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9573
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.84375,
      "epoch": 0.4626014688828759,
      "grad_norm": 0.0,
      "learning_rate": 2.6872342481638964e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9574
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.84375,
      "epoch": 0.46264978739853113,
      "grad_norm": 0.0,
      "learning_rate": 2.6869926555856206e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9575
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.5625,
      "epoch": 0.4626981059141863,
      "grad_norm": 0.0,
      "learning_rate": 2.686751063007344e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9576
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.34375,
      "epoch": 0.46274642442984154,
      "grad_norm": 0.0,
      "learning_rate": 2.6865094704290683e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9577
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.875,
      "epoch": 0.4627947429454967,
      "grad_norm": 0.0,
      "learning_rate": 2.6862678778507925e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9578
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.71875,
      "epoch": 0.4628430614611519,
      "grad_norm": 0.0,
      "learning_rate": 2.686026285272516e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9579
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.875,
      "epoch": 0.4628913799768071,
      "grad_norm": 0.0,
      "learning_rate": 2.68578469269424e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9580
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.03125,
      "epoch": 0.4629396984924623,
      "grad_norm": 0.0,
      "learning_rate": 2.6855431001159643e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9581
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.6875,
      "epoch": 0.46298801700811754,
      "grad_norm": 0.0,
      "learning_rate": 2.6853015075376885e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9582
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.78125,
      "epoch": 0.4630363355237727,
      "grad_norm": 0.0,
      "learning_rate": 2.685059914959412e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9583
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.5625,
      "epoch": 0.4630846540394279,
      "grad_norm": 0.0,
      "learning_rate": 2.684818322381137e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9584
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.6875,
      "epoch": 0.4631329725550831,
      "grad_norm": 0.0,
      "learning_rate": 2.6845767298028604e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9585
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.78125,
      "epoch": 0.4631812910707383,
      "grad_norm": 0.0,
      "learning_rate": 2.6843351372245845e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9586
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.09375,
      "epoch": 0.46322960958639353,
      "grad_norm": 0.0,
      "learning_rate": 2.684093544646308e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9587
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.0,
      "epoch": 0.4632779281020487,
      "grad_norm": 0.0,
      "learning_rate": 2.683851952068033e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9588
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.65625,
      "epoch": 0.4633262466177039,
      "grad_norm": 0.0,
      "learning_rate": 2.6836103594897564e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9589
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.96875,
      "epoch": 0.4633745651333591,
      "grad_norm": 0.0,
      "learning_rate": 2.68336876691148e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9590
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.28125,
      "epoch": 0.4634228836490143,
      "grad_norm": 0.0,
      "learning_rate": 2.6831271743332046e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9591
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.8125,
      "epoch": 0.4634712021646695,
      "grad_norm": 0.0,
      "learning_rate": 2.682885581754928e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9592
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.96875,
      "epoch": 0.4635195206803247,
      "grad_norm": 0.0,
      "learning_rate": 2.6826439891766524e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9593
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.125,
      "epoch": 0.4635678391959799,
      "grad_norm": 0.0,
      "learning_rate": 2.6824023965983765e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9594
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.1875,
      "epoch": 0.4636161577116351,
      "grad_norm": 0.0,
      "learning_rate": 2.6821608040201007e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9595
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.78125,
      "epoch": 0.4636644762272903,
      "grad_norm": 0.0,
      "learning_rate": 2.681919211441824e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9596
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.03125,
      "epoch": 0.4637127947429455,
      "grad_norm": 36.47067854613329,
      "learning_rate": 2.6816776188635484e-07,
      "loss": -0.0,
      "reward": 2.3125,
      "reward_std": 0.2725512385368347,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9597
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.40625,
      "epoch": 0.4637611132586007,
      "grad_norm": 0.0,
      "learning_rate": 2.6814360262852725e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9598
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.8125,
      "epoch": 0.4638094317742559,
      "grad_norm": 0.0,
      "learning_rate": 2.681194433706996e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9599
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.6875,
      "epoch": 0.4638577502899111,
      "grad_norm": 0.0,
      "learning_rate": 2.6809528411287203e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9600
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.3125,
      "epoch": 0.4639060688055663,
      "grad_norm": 0.0,
      "learning_rate": 2.6807112485504444e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9601
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.96875,
      "epoch": 0.4639543873212215,
      "grad_norm": 0.0,
      "learning_rate": 2.6804696559721686e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 9602
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.75,
      "epoch": 0.4640027058368767,
      "grad_norm": 0.0,
      "learning_rate": 2.680228063393892e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9603
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.0625,
      "epoch": 0.46405102435253187,
      "grad_norm": 0.0,
      "learning_rate": 2.679986470815617e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9604
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.9375,
      "epoch": 0.4640993428681871,
      "grad_norm": 0.0,
      "learning_rate": 2.6797448782373404e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9605
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.125,
      "epoch": 0.4641476613838423,
      "grad_norm": 0.0,
      "learning_rate": 2.6795032856590646e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9606
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.6875,
      "epoch": 0.4641959798994975,
      "grad_norm": 0.0,
      "learning_rate": 2.679261693080788e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9607
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.5625,
      "epoch": 0.4642442984151527,
      "grad_norm": 0.0,
      "learning_rate": 2.679020100502513e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9608
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.625,
      "epoch": 0.46429261693080787,
      "grad_norm": 0.0,
      "learning_rate": 2.6787785079242364e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9609
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.375,
      "epoch": 0.4643409354464631,
      "grad_norm": 0.0,
      "learning_rate": 2.67853691534596e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9610
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.09375,
      "epoch": 0.4643892539621183,
      "grad_norm": 0.0,
      "learning_rate": 2.6782953227676847e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9611
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.375,
      "epoch": 0.4644375724777735,
      "grad_norm": 0.0,
      "learning_rate": 2.6780537301894083e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9612
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.0,
      "epoch": 0.4644858909934287,
      "grad_norm": 0.0,
      "learning_rate": 2.6778121376111325e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9613
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.84375,
      "epoch": 0.46453420950908386,
      "grad_norm": 0.0,
      "learning_rate": 2.6775705450328566e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9614
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.09375,
      "epoch": 0.4645825280247391,
      "grad_norm": 0.0,
      "learning_rate": 2.6773289524545807e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9615
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.375,
      "epoch": 0.46463084654039427,
      "grad_norm": 0.0,
      "learning_rate": 2.6770873598763043e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9616
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.09375,
      "epoch": 0.4646791650560495,
      "grad_norm": 0.0,
      "learning_rate": 2.6768457672980285e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9617
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.03125,
      "epoch": 0.4647274835717047,
      "grad_norm": 0.0,
      "learning_rate": 2.6766041747197526e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9618
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.6875,
      "epoch": 0.46477580208735986,
      "grad_norm": 0.0,
      "learning_rate": 2.676362582141476e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 9619
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.84375,
      "epoch": 0.4648241206030151,
      "grad_norm": 0.0,
      "learning_rate": 2.6761209895632004e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9620
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.65625,
      "epoch": 0.46487243911867027,
      "grad_norm": 0.0,
      "learning_rate": 2.6758793969849245e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9621
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 76.65625,
      "epoch": 0.4649207576343255,
      "grad_norm": 140.5315419826242,
      "learning_rate": 2.6756378044066486e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9622
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.96875,
      "epoch": 0.4649690761499807,
      "grad_norm": 0.0,
      "learning_rate": 2.675396211828372e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9623
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.96875,
      "epoch": 0.46501739466563585,
      "grad_norm": 0.0,
      "learning_rate": 2.675154619250097e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9624
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.125,
      "epoch": 0.4650657131812911,
      "grad_norm": 0.0,
      "learning_rate": 2.6749130266718205e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9625
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.5,
      "epoch": 0.46511403169694626,
      "grad_norm": 0.0,
      "learning_rate": 2.6746714340935446e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2130674123764038,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9626
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.84375,
      "epoch": 0.4651623502126015,
      "grad_norm": 0.0,
      "learning_rate": 2.674429841515268e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9627
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.90625,
      "epoch": 0.46521066872825667,
      "grad_norm": 0.0,
      "learning_rate": 2.674188248936993e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9628
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.09375,
      "epoch": 0.46525898724391185,
      "grad_norm": 0.0,
      "learning_rate": 2.6739466563587165e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9629
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.9375,
      "epoch": 0.4653073057595671,
      "grad_norm": 0.0,
      "learning_rate": 2.67370506378044e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9630
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.21875,
      "epoch": 0.46535562427522226,
      "grad_norm": 0.0,
      "learning_rate": 2.673463471202165e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9631
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.3125,
      "epoch": 0.4654039427908775,
      "grad_norm": 0.0,
      "learning_rate": 2.6732218786238884e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9632
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.84375,
      "epoch": 0.46545226130653267,
      "grad_norm": 0.0,
      "learning_rate": 2.6729802860456125e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9633
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.3125,
      "epoch": 0.46550057982218784,
      "grad_norm": 0.0,
      "learning_rate": 2.6727386934673367e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9634
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.09375,
      "epoch": 0.4655488983378431,
      "grad_norm": 0.0,
      "learning_rate": 2.672497100889061e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9635
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.78125,
      "epoch": 0.46559721685349825,
      "grad_norm": 0.0,
      "learning_rate": 2.6722555083107844e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9636
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.25,
      "epoch": 0.4656455353691535,
      "grad_norm": 0.0,
      "learning_rate": 2.6720139157325086e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9637
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.375,
      "epoch": 0.46569385388480866,
      "grad_norm": 0.0,
      "learning_rate": 2.6717723231542327e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9638
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.65625,
      "epoch": 0.46574217240046384,
      "grad_norm": 0.0,
      "learning_rate": 2.671530730575957e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9639
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.875,
      "epoch": 0.46579049091611907,
      "grad_norm": 0.0,
      "learning_rate": 2.6712891379976804e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9640
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.6875,
      "epoch": 0.46583880943177425,
      "grad_norm": 0.0,
      "learning_rate": 2.6710475454194046e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9641
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.84375,
      "epoch": 0.4658871279474295,
      "grad_norm": 0.0,
      "learning_rate": 2.6708059528411287e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9642
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.5625,
      "epoch": 0.46593544646308466,
      "grad_norm": 0.0,
      "learning_rate": 2.6705643602628523e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9643
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.40625,
      "epoch": 0.46598376497873983,
      "grad_norm": 0.0,
      "learning_rate": 2.670322767684577e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 9644
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.1875,
      "epoch": 0.46603208349439507,
      "grad_norm": 0.0,
      "learning_rate": 2.6700811751063006e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9645
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.6875,
      "epoch": 0.46608040201005024,
      "grad_norm": 0.0,
      "learning_rate": 2.6698395825280247e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9646
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.1875,
      "epoch": 0.4661287205257055,
      "grad_norm": 0.0,
      "learning_rate": 2.6695979899497483e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9647
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.75,
      "epoch": 0.46617703904136065,
      "grad_norm": 0.0,
      "learning_rate": 2.669356397371473e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9648
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.0,
      "epoch": 0.46622535755701583,
      "grad_norm": 0.0,
      "learning_rate": 2.6691148047931966e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9649
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.90625,
      "epoch": 0.46627367607267106,
      "grad_norm": 0.0,
      "learning_rate": 2.66887321221492e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9650
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.4375,
      "epoch": 0.46632199458832624,
      "grad_norm": 0.0,
      "learning_rate": 2.668631619636645e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9651
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.0625,
      "epoch": 0.46637031310398147,
      "grad_norm": 0.0,
      "learning_rate": 2.6683900270583685e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9652
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.65625,
      "epoch": 0.46641863161963665,
      "grad_norm": 0.0,
      "learning_rate": 2.6681484344800926e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 0.96875,
      "step": 9653
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.71875,
      "epoch": 0.4664669501352918,
      "grad_norm": 0.0,
      "learning_rate": 2.667906841901817e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9654
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.34375,
      "epoch": 0.46651526865094706,
      "grad_norm": 0.0,
      "learning_rate": 2.667665249323541e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9655
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.0,
      "epoch": 0.46656358716660223,
      "grad_norm": 0.0,
      "learning_rate": 2.6674236567452645e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9656
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.4375,
      "epoch": 0.46661190568225747,
      "grad_norm": 0.0,
      "learning_rate": 2.6671820641669886e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9657
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.6875,
      "epoch": 0.46666022419791264,
      "grad_norm": 0.0,
      "learning_rate": 2.666940471588713e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9658
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.1875,
      "epoch": 0.4667085427135678,
      "grad_norm": 0.0,
      "learning_rate": 2.666698879010437e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9659
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.375,
      "epoch": 0.46675686122922305,
      "grad_norm": 0.0,
      "learning_rate": 2.6664572864321605e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9660
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.1875,
      "epoch": 0.46680517974487823,
      "grad_norm": 0.0,
      "learning_rate": 2.6662156938538847e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9661
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.96875,
      "epoch": 0.46685349826053346,
      "grad_norm": 0.0,
      "learning_rate": 2.665974101275609e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9662
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.09375,
      "epoch": 0.46690181677618864,
      "grad_norm": 0.0,
      "learning_rate": 2.6657325086973324e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9663
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.4375,
      "epoch": 0.4669501352918438,
      "grad_norm": 0.0,
      "learning_rate": 2.665490916119057e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9664
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.03125,
      "epoch": 0.46699845380749905,
      "grad_norm": 0.0,
      "learning_rate": 2.6652493235407807e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9665
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.40625,
      "epoch": 0.4670467723231542,
      "grad_norm": 0.0,
      "learning_rate": 2.665007730962505e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9666
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.53125,
      "epoch": 0.46709509083880946,
      "grad_norm": 0.0,
      "learning_rate": 2.6647661383842284e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9667
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.4375,
      "epoch": 0.46714340935446463,
      "grad_norm": 0.0,
      "learning_rate": 2.664524545805953e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 0.96875,
      "step": 9668
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.6875,
      "epoch": 0.4671917278701198,
      "grad_norm": 0.0,
      "learning_rate": 2.6642829532276767e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9669
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.84375,
      "epoch": 0.46724004638577504,
      "grad_norm": 0.0,
      "learning_rate": 2.6640413606494003e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 9670
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.46875,
      "epoch": 0.4672883649014302,
      "grad_norm": 0.0,
      "learning_rate": 2.663799768071125e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9671
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.28125,
      "epoch": 0.46733668341708545,
      "grad_norm": 0.0,
      "learning_rate": 2.6635581754928486e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9672
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.28125,
      "epoch": 0.46738500193274063,
      "grad_norm": 0.0,
      "learning_rate": 2.6633165829145727e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9673
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.96875,
      "epoch": 0.4674333204483958,
      "grad_norm": 0.0,
      "learning_rate": 2.663074990336297e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9674
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.09375,
      "epoch": 0.46748163896405104,
      "grad_norm": 0.0,
      "learning_rate": 2.662833397758021e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9675
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.59375,
      "epoch": 0.4675299574797062,
      "grad_norm": 0.0,
      "learning_rate": 2.6625918051797446e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9676
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.8125,
      "epoch": 0.46757827599536145,
      "grad_norm": 0.0,
      "learning_rate": 2.6623502126014687e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9677
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.71875,
      "epoch": 0.4676265945110166,
      "grad_norm": 0.0,
      "learning_rate": 2.662108620023193e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9678
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.28125,
      "epoch": 0.4676749130266718,
      "grad_norm": 0.0,
      "learning_rate": 2.661867027444917e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9679
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.5,
      "epoch": 0.46772323154232703,
      "grad_norm": 0.0,
      "learning_rate": 2.6616254348666406e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9680
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.90625,
      "epoch": 0.4677715500579822,
      "grad_norm": 0.0,
      "learning_rate": 2.6613838422883647e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9681
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.59375,
      "epoch": 0.46781986857363744,
      "grad_norm": 0.0,
      "learning_rate": 2.661142249710089e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9682
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 89.15625,
      "epoch": 0.4678681870892926,
      "grad_norm": 0.0,
      "learning_rate": 2.6609006571318125e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 9683
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.9375,
      "epoch": 0.4679165056049478,
      "grad_norm": 0.0,
      "learning_rate": 2.660659064553537e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9684
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.15625,
      "epoch": 0.46796482412060303,
      "grad_norm": 0.0,
      "learning_rate": 2.660417471975261e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9685
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.25,
      "epoch": 0.4680131426362582,
      "grad_norm": 0.0,
      "learning_rate": 2.660175879396985e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9686
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.9375,
      "epoch": 0.46806146115191344,
      "grad_norm": 0.0,
      "learning_rate": 2.6599342868187085e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9687
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.28125,
      "epoch": 0.4681097796675686,
      "grad_norm": 0.0,
      "learning_rate": 2.659692694240433e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9688
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.125,
      "epoch": 0.4681580981832238,
      "grad_norm": 0.0,
      "learning_rate": 2.659451101662157e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9689
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.53125,
      "epoch": 0.468206416698879,
      "grad_norm": 0.0,
      "learning_rate": 2.659209509083881e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9690
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.75,
      "epoch": 0.4682547352145342,
      "grad_norm": 0.0,
      "learning_rate": 2.658967916505605e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9691
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.25,
      "epoch": 0.46830305373018943,
      "grad_norm": 0.0,
      "learning_rate": 2.6587263239273286e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9692
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.03125,
      "epoch": 0.4683513722458446,
      "grad_norm": 0.0,
      "learning_rate": 2.658484731349053e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9693
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.625,
      "epoch": 0.4683996907614998,
      "grad_norm": 0.0,
      "learning_rate": 2.658243138770777e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9694
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.09375,
      "epoch": 0.468448009277155,
      "grad_norm": 0.0,
      "learning_rate": 2.658001546192501e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9695
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 88.40625,
      "epoch": 0.4684963277928102,
      "grad_norm": 0.0,
      "learning_rate": 2.6577599536142247e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 9696
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.96875,
      "epoch": 0.46854464630846543,
      "grad_norm": 0.0,
      "learning_rate": 2.657518361035949e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9697
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.84375,
      "epoch": 0.4685929648241206,
      "grad_norm": 0.0,
      "learning_rate": 2.657276768457673e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9698
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.84375,
      "epoch": 0.4686412833397758,
      "grad_norm": 0.0,
      "learning_rate": 2.657035175879397e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9699
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.15625,
      "epoch": 0.468689601855431,
      "grad_norm": 0.0,
      "learning_rate": 2.6567935833011207e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9700
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.6875,
      "epoch": 0.4687379203710862,
      "grad_norm": 0.0,
      "learning_rate": 2.6565519907228453e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9701
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.40625,
      "epoch": 0.4687862388867414,
      "grad_norm": 0.0,
      "learning_rate": 2.656310398144569e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9702
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 82.96875,
      "epoch": 0.4688345574023966,
      "grad_norm": 45.24721741959536,
      "learning_rate": 2.6560688055662926e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9703
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.03125,
      "epoch": 0.4688828759180518,
      "grad_norm": 0.0,
      "learning_rate": 2.655827212988017e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9704
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.09375,
      "epoch": 0.468931194433707,
      "grad_norm": 0.0,
      "learning_rate": 2.655585620409741e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9705
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.09375,
      "epoch": 0.4689795129493622,
      "grad_norm": 0.0,
      "learning_rate": 2.655344027831465e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9706
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.40625,
      "epoch": 0.4690278314650174,
      "grad_norm": 0.0,
      "learning_rate": 2.6551024352531886e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9707
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.0625,
      "epoch": 0.4690761499806726,
      "grad_norm": 0.0,
      "learning_rate": 2.654860842674913e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9708
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.84375,
      "epoch": 0.4691244684963278,
      "grad_norm": 0.0,
      "learning_rate": 2.654619250096637e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9709
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.0625,
      "epoch": 0.469172787011983,
      "grad_norm": 0.0,
      "learning_rate": 2.654377657518361e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9710
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.0625,
      "epoch": 0.4692211055276382,
      "grad_norm": 0.0,
      "learning_rate": 2.654136064940085e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9711
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.5625,
      "epoch": 0.4692694240432934,
      "grad_norm": 0.0,
      "learning_rate": 2.6538944723618087e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9712
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.71875,
      "epoch": 0.4693177425589486,
      "grad_norm": 0.0,
      "learning_rate": 2.653652879783533e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9713
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.5625,
      "epoch": 0.46936606107460377,
      "grad_norm": 0.0,
      "learning_rate": 2.653411287205257e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9714
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.375,
      "epoch": 0.469414379590259,
      "grad_norm": 0.0,
      "learning_rate": 2.653169694626981e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9715
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.65625,
      "epoch": 0.4694626981059142,
      "grad_norm": 0.0,
      "learning_rate": 2.652928102048705e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9716
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.5,
      "epoch": 0.4695110166215694,
      "grad_norm": 0.0,
      "learning_rate": 2.652686509470429e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9717
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.53125,
      "epoch": 0.4695593351372246,
      "grad_norm": 0.0,
      "learning_rate": 2.652444916892153e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9718
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.21875,
      "epoch": 0.46960765365287976,
      "grad_norm": 0.0,
      "learning_rate": 2.652203324313877e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9719
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.9375,
      "epoch": 0.469655972168535,
      "grad_norm": 0.0,
      "learning_rate": 2.651961731735601e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9720
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.03125,
      "epoch": 0.4697042906841902,
      "grad_norm": 0.0,
      "learning_rate": 2.6517201391573254e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9721
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.6875,
      "epoch": 0.4697526091998454,
      "grad_norm": 0.0,
      "learning_rate": 2.651478546579049e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9722
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.25,
      "epoch": 0.4698009277155006,
      "grad_norm": 0.0,
      "learning_rate": 2.6512369540007726e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9723
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.4375,
      "epoch": 0.46984924623115576,
      "grad_norm": 0.0,
      "learning_rate": 2.6509953614224973e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9724
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.46875,
      "epoch": 0.469897564746811,
      "grad_norm": 0.0,
      "learning_rate": 2.650753768844221e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9725
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.625,
      "epoch": 0.46994588326246617,
      "grad_norm": 0.0,
      "learning_rate": 2.650512176265945e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9726
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.5625,
      "epoch": 0.4699942017781214,
      "grad_norm": 0.0,
      "learning_rate": 2.6502705836876686e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9727
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.53125,
      "epoch": 0.4700425202937766,
      "grad_norm": 0.0,
      "learning_rate": 2.6500289911093933e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9728
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.96875,
      "epoch": 0.47009083880943175,
      "grad_norm": 0.0,
      "learning_rate": 2.649787398531117e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9729
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.3125,
      "epoch": 0.470139157325087,
      "grad_norm": 0.0,
      "learning_rate": 2.649545805952841e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9730
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.4375,
      "epoch": 0.47018747584074216,
      "grad_norm": 0.0,
      "learning_rate": 2.649304213374565e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9731
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.90625,
      "epoch": 0.4702357943563974,
      "grad_norm": 0.0,
      "learning_rate": 2.649062620796289e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9732
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.375,
      "epoch": 0.4702841128720526,
      "grad_norm": 0.0,
      "learning_rate": 2.648821028218013e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9733
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.09375,
      "epoch": 0.47033243138770775,
      "grad_norm": 0.0,
      "learning_rate": 2.648579435639737e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9734
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.15625,
      "epoch": 0.470380749903363,
      "grad_norm": 0.0,
      "learning_rate": 2.648337843061461e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9735
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.28125,
      "epoch": 0.47042906841901816,
      "grad_norm": 0.0,
      "learning_rate": 2.648096250483185e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9736
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.6875,
      "epoch": 0.4704773869346734,
      "grad_norm": 0.0,
      "learning_rate": 2.647854657904909e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9737
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.40625,
      "epoch": 0.47052570545032857,
      "grad_norm": 0.0,
      "learning_rate": 2.647613065326633e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9738
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.46875,
      "epoch": 0.47057402396598375,
      "grad_norm": 0.0,
      "learning_rate": 2.647371472748357e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9739
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.15625,
      "epoch": 0.470622342481639,
      "grad_norm": 0.0,
      "learning_rate": 2.647129880170081e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9740
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.40625,
      "epoch": 0.47067066099729415,
      "grad_norm": 0.0,
      "learning_rate": 2.6468882875918055e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9741
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.59375,
      "epoch": 0.4707189795129494,
      "grad_norm": 0.0,
      "learning_rate": 2.646646695013529e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9742
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.6875,
      "epoch": 0.47076729802860456,
      "grad_norm": 0.0,
      "learning_rate": 2.6464051024352527e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9743
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.0625,
      "epoch": 0.47081561654425974,
      "grad_norm": 0.0,
      "learning_rate": 2.6461635098569774e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9744
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.9375,
      "epoch": 0.470863935059915,
      "grad_norm": 0.0,
      "learning_rate": 2.645921917278701e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9745
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.5625,
      "epoch": 0.47091225357557015,
      "grad_norm": 0.0,
      "learning_rate": 2.645680324700425e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9746
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.875,
      "epoch": 0.4709605720912254,
      "grad_norm": 0.0,
      "learning_rate": 2.6454387321221487e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9747
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.875,
      "epoch": 0.47100889060688056,
      "grad_norm": 0.0,
      "learning_rate": 2.6451971395438734e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9748
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.59375,
      "epoch": 0.47105720912253574,
      "grad_norm": 0.0,
      "learning_rate": 2.644955546965597e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9749
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.375,
      "epoch": 0.47110552763819097,
      "grad_norm": 0.0,
      "learning_rate": 2.644713954387321e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9750
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.4375,
      "epoch": 0.47115384615384615,
      "grad_norm": 0.0,
      "learning_rate": 2.6444723618090453e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9751
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.84375,
      "epoch": 0.4712021646695014,
      "grad_norm": 0.0,
      "learning_rate": 2.6442307692307694e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9752
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.28125,
      "epoch": 0.47125048318515655,
      "grad_norm": 0.0,
      "learning_rate": 2.643989176652493e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9753
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.75,
      "epoch": 0.47129880170081173,
      "grad_norm": 0.0,
      "learning_rate": 2.643747584074217e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 9754
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.21875,
      "epoch": 0.47134712021646696,
      "grad_norm": 0.0,
      "learning_rate": 2.6435059914959413e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9755
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.5,
      "epoch": 0.47139543873212214,
      "grad_norm": 0.0,
      "learning_rate": 2.643264398917665e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9756
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.03125,
      "epoch": 0.4714437572477774,
      "grad_norm": 0.0,
      "learning_rate": 2.643022806339389e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9757
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.5,
      "epoch": 0.47149207576343255,
      "grad_norm": 0.0,
      "learning_rate": 2.642781213761113e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9758
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.75,
      "epoch": 0.4715403942790877,
      "grad_norm": 74.18102669942391,
      "learning_rate": 2.6425396211828373e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9759
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.21875,
      "epoch": 0.47158871279474296,
      "grad_norm": 0.0,
      "learning_rate": 2.642298028604561e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9760
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.21875,
      "epoch": 0.47163703131039814,
      "grad_norm": 0.0,
      "learning_rate": 2.6420564360262856e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 9761
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.90625,
      "epoch": 0.47168534982605337,
      "grad_norm": 0.0,
      "learning_rate": 2.641814843448009e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9762
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.0625,
      "epoch": 0.47173366834170855,
      "grad_norm": 0.0,
      "learning_rate": 2.641573250869733e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9763
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.96875,
      "epoch": 0.4717819868573637,
      "grad_norm": 0.0,
      "learning_rate": 2.6413316582914575e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9764
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.90625,
      "epoch": 0.47183030537301895,
      "grad_norm": 0.0,
      "learning_rate": 2.641090065713181e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9765
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.125,
      "epoch": 0.47187862388867413,
      "grad_norm": 0.0,
      "learning_rate": 2.640848473134905e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9766
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.65625,
      "epoch": 0.47192694240432936,
      "grad_norm": 0.0,
      "learning_rate": 2.640606880556629e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9767
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.875,
      "epoch": 0.47197526091998454,
      "grad_norm": 0.0,
      "learning_rate": 2.6403652879783535e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9768
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.75,
      "epoch": 0.4720235794356397,
      "grad_norm": 0.0,
      "learning_rate": 2.640123695400077e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9769
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.78125,
      "epoch": 0.47207189795129495,
      "grad_norm": 0.0,
      "learning_rate": 2.639882102821801e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 9770
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.5625,
      "epoch": 0.4721202164669501,
      "grad_norm": 0.0,
      "learning_rate": 2.6396405102435253e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9771
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.40625,
      "epoch": 0.47216853498260536,
      "grad_norm": 0.0,
      "learning_rate": 2.6393989176652495e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9772
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 89.5,
      "epoch": 0.47221685349826054,
      "grad_norm": 55.88035794278045,
      "learning_rate": 2.639157325086973e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.3098084628582001,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9773
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.90625,
      "epoch": 0.4722651720139157,
      "grad_norm": 0.0,
      "learning_rate": 2.638915732508697e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9774
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.375,
      "epoch": 0.47231349052957095,
      "grad_norm": 0.0,
      "learning_rate": 2.6386741399304214e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9775
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.90625,
      "epoch": 0.4723618090452261,
      "grad_norm": 0.0,
      "learning_rate": 2.638432547352145e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9776
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.25,
      "epoch": 0.47241012756088135,
      "grad_norm": 0.0,
      "learning_rate": 2.638190954773869e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9777
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.1875,
      "epoch": 0.47245844607653653,
      "grad_norm": 0.0,
      "learning_rate": 2.637949362195593e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9778
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.15625,
      "epoch": 0.4725067645921917,
      "grad_norm": 0.0,
      "learning_rate": 2.6377077696173174e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9779
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.65625,
      "epoch": 0.47255508310784694,
      "grad_norm": 0.0,
      "learning_rate": 2.637466177039041e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9780
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.9375,
      "epoch": 0.4726034016235021,
      "grad_norm": 0.0,
      "learning_rate": 2.6372245844607657e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9781
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.875,
      "epoch": 0.47265172013915735,
      "grad_norm": 0.0,
      "learning_rate": 2.636982991882489e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9782
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.90625,
      "epoch": 0.4727000386548125,
      "grad_norm": 0.0,
      "learning_rate": 2.6367413993042134e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9783
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.53125,
      "epoch": 0.4727483571704677,
      "grad_norm": 0.0,
      "learning_rate": 2.6364998067259375e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9784
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.375,
      "epoch": 0.47279667568612294,
      "grad_norm": 0.0,
      "learning_rate": 2.636258214147661e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9785
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.75,
      "epoch": 0.4728449942017781,
      "grad_norm": 0.0,
      "learning_rate": 2.6360166215693853e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9786
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.40625,
      "epoch": 0.47289331271743335,
      "grad_norm": 0.0,
      "learning_rate": 2.635775028991109e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9787
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.96875,
      "epoch": 0.4729416312330885,
      "grad_norm": 0.0,
      "learning_rate": 2.6355334364128335e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9788
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.875,
      "epoch": 0.4729899497487437,
      "grad_norm": 0.0,
      "learning_rate": 2.635291843834557e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9789
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.375,
      "epoch": 0.47303826826439893,
      "grad_norm": 0.0,
      "learning_rate": 2.6350502512562813e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9790
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.75,
      "epoch": 0.4730865867800541,
      "grad_norm": 0.0,
      "learning_rate": 2.6348086586780054e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9791
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.875,
      "epoch": 0.47313490529570934,
      "grad_norm": 0.0,
      "learning_rate": 2.6345670660997296e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9792
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.09375,
      "epoch": 0.4731832238113645,
      "grad_norm": 0.0,
      "learning_rate": 2.634325473521453e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9793
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.75,
      "epoch": 0.4732315423270197,
      "grad_norm": 0.0,
      "learning_rate": 2.634083880943178e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9794
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.0625,
      "epoch": 0.4732798608426749,
      "grad_norm": 0.0,
      "learning_rate": 2.6338422883649014e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9795
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.9375,
      "epoch": 0.4733281793583301,
      "grad_norm": 50.83908556821807,
      "learning_rate": 2.633600695786625e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.2657213509082794,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9796
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.625,
      "epoch": 0.47337649787398534,
      "grad_norm": 0.0,
      "learning_rate": 2.633359103208349e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9797
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.125,
      "epoch": 0.4734248163896405,
      "grad_norm": 0.0,
      "learning_rate": 2.6331175106300733e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9798
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.71875,
      "epoch": 0.4734731349052957,
      "grad_norm": 0.0,
      "learning_rate": 2.6328759180517975e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9799
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 83.21875,
      "epoch": 0.4735214534209509,
      "grad_norm": 19.185894247664347,
      "learning_rate": 2.632634325473521e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9800
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.4375,
      "epoch": 0.4735697719366061,
      "grad_norm": 0.0,
      "learning_rate": 2.6323927328952457e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9801
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.875,
      "epoch": 0.47361809045226133,
      "grad_norm": 0.0,
      "learning_rate": 2.6321511403169693e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 9802
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 90.03125,
      "epoch": 0.4736664089679165,
      "grad_norm": 28.15312730564163,
      "learning_rate": 2.6319095477386935e-07,
      "loss": -0.0,
      "reward": 2.28125,
      "reward_std": 0.30509042739868164,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9803
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.4375,
      "epoch": 0.4737147274835717,
      "grad_norm": 0.0,
      "learning_rate": 2.6316679551604176e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9804
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.4375,
      "epoch": 0.4737630459992269,
      "grad_norm": 0.0,
      "learning_rate": 2.631426362582141e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9805
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.84375,
      "epoch": 0.4738113645148821,
      "grad_norm": 0.0,
      "learning_rate": 2.6311847700038654e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9806
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.5625,
      "epoch": 0.4738596830305373,
      "grad_norm": 0.0,
      "learning_rate": 2.630943177425589e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9807
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.59375,
      "epoch": 0.4739080015461925,
      "grad_norm": 0.0,
      "learning_rate": 2.6307015848473136e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9808
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.125,
      "epoch": 0.4739563200618477,
      "grad_norm": 0.0,
      "learning_rate": 2.630459992269037e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9809
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.96875,
      "epoch": 0.4740046385775029,
      "grad_norm": 0.0,
      "learning_rate": 2.6302183996907614e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9810
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.46875,
      "epoch": 0.4740529570931581,
      "grad_norm": 0.0,
      "learning_rate": 2.6299768071124855e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9811
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.65625,
      "epoch": 0.4741012756088133,
      "grad_norm": 0.0,
      "learning_rate": 2.6297352145342096e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9812
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.09375,
      "epoch": 0.4741495941244685,
      "grad_norm": 0.0,
      "learning_rate": 2.629493621955933e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9813
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.8125,
      "epoch": 0.4741979126401237,
      "grad_norm": 0.0,
      "learning_rate": 2.629252029377658e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9814
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.0625,
      "epoch": 0.4742462311557789,
      "grad_norm": 0.0,
      "learning_rate": 2.6290104367993815e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9815
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.78125,
      "epoch": 0.4742945496714341,
      "grad_norm": 0.0,
      "learning_rate": 2.628768844221105e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9816
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.75,
      "epoch": 0.4743428681870893,
      "grad_norm": 0.0,
      "learning_rate": 2.6285272516428293e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9817
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.28125,
      "epoch": 0.4743911867027445,
      "grad_norm": 0.0,
      "learning_rate": 2.6282856590645534e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9818
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.0,
      "epoch": 0.47443950521839967,
      "grad_norm": 0.0,
      "learning_rate": 2.6280440664862775e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9819
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.90625,
      "epoch": 0.4744878237340549,
      "grad_norm": 0.0,
      "learning_rate": 2.627802473908001e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9820
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 69.84375,
      "epoch": 0.4745361422497101,
      "grad_norm": 0.0,
      "learning_rate": 2.627560881329726e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9821
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.84375,
      "epoch": 0.4745844607653653,
      "grad_norm": 0.0,
      "learning_rate": 2.6273192887514494e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9822
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.34375,
      "epoch": 0.4746327792810205,
      "grad_norm": 0.0,
      "learning_rate": 2.6270776961731736e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9823
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.90625,
      "epoch": 0.47468109779667567,
      "grad_norm": 0.0,
      "learning_rate": 2.6268361035948977e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9824
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.15625,
      "epoch": 0.4747294163123309,
      "grad_norm": 0.0,
      "learning_rate": 2.6265945110166213e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9825
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.0625,
      "epoch": 0.4747777348279861,
      "grad_norm": 0.0,
      "learning_rate": 2.6263529184383454e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9826
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.0,
      "epoch": 0.4748260533436413,
      "grad_norm": 0.0,
      "learning_rate": 2.626111325860069e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9827
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.96875,
      "epoch": 0.4748743718592965,
      "grad_norm": 0.0,
      "learning_rate": 2.6258697332817937e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9828
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.0625,
      "epoch": 0.47492269037495166,
      "grad_norm": 0.0,
      "learning_rate": 2.6256281407035173e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9829
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.0625,
      "epoch": 0.4749710088906069,
      "grad_norm": 0.0,
      "learning_rate": 2.6253865481252414e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9830
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.96875,
      "epoch": 0.47501932740626207,
      "grad_norm": 0.0,
      "learning_rate": 2.6251449555469656e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9831
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.71875,
      "epoch": 0.4750676459219173,
      "grad_norm": 0.0,
      "learning_rate": 2.6249033629686897e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9832
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.59375,
      "epoch": 0.4751159644375725,
      "grad_norm": 0.0,
      "learning_rate": 2.6246617703904133e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 9833
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.59375,
      "epoch": 0.47516428295322766,
      "grad_norm": 0.0,
      "learning_rate": 2.624420177812138e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 9834
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.5625,
      "epoch": 0.4752126014688829,
      "grad_norm": 0.0,
      "learning_rate": 2.6241785852338616e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9835
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.5,
      "epoch": 0.47526091998453807,
      "grad_norm": 0.0,
      "learning_rate": 2.623936992655585e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9836
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.0625,
      "epoch": 0.4753092385001933,
      "grad_norm": 0.0,
      "learning_rate": 2.6236954000773093e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9837
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.1875,
      "epoch": 0.4753575570158485,
      "grad_norm": 0.0,
      "learning_rate": 2.6234538074990335e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9838
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.46875,
      "epoch": 0.47540587553150365,
      "grad_norm": 0.0,
      "learning_rate": 2.6232122149207576e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9839
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.9375,
      "epoch": 0.4754541940471589,
      "grad_norm": 0.0,
      "learning_rate": 2.622970622342481e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9840
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.9375,
      "epoch": 0.47550251256281406,
      "grad_norm": 0.0,
      "learning_rate": 2.622729029764206e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9841
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.75,
      "epoch": 0.4755508310784693,
      "grad_norm": 0.0,
      "learning_rate": 2.6224874371859295e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9842
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.53125,
      "epoch": 0.47559914959412447,
      "grad_norm": 0.0,
      "learning_rate": 2.6222458446076536e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9843
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.9375,
      "epoch": 0.47564746810977965,
      "grad_norm": 0.0,
      "learning_rate": 2.622004252029378e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9844
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.375,
      "epoch": 0.4756957866254349,
      "grad_norm": 0.0,
      "learning_rate": 2.621762659451102e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9845
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.0625,
      "epoch": 0.47574410514109006,
      "grad_norm": 0.0,
      "learning_rate": 2.6215210668728255e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9846
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.09375,
      "epoch": 0.4757924236567453,
      "grad_norm": 0.0,
      "learning_rate": 2.621279474294549e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9847
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.65625,
      "epoch": 0.47584074217240047,
      "grad_norm": 0.0,
      "learning_rate": 2.621037881716274e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9848
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.46875,
      "epoch": 0.47588906068805564,
      "grad_norm": 0.0,
      "learning_rate": 2.6207962891379974e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9849
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.71875,
      "epoch": 0.4759373792037109,
      "grad_norm": 0.0,
      "learning_rate": 2.6205546965597215e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9850
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.28125,
      "epoch": 0.47598569771936605,
      "grad_norm": 0.0,
      "learning_rate": 2.6203131039814457e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9851
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.5,
      "epoch": 0.4760340162350213,
      "grad_norm": 0.0,
      "learning_rate": 2.62007151140317e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9852
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.5,
      "epoch": 0.47608233475067646,
      "grad_norm": 0.0,
      "learning_rate": 2.6198299188248934e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9853
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.5625,
      "epoch": 0.47613065326633164,
      "grad_norm": 0.0,
      "learning_rate": 2.619588326246618e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9854
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.5,
      "epoch": 0.47617897178198687,
      "grad_norm": 0.0,
      "learning_rate": 2.6193467336683417e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9855
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.375,
      "epoch": 0.47622729029764205,
      "grad_norm": 0.0,
      "learning_rate": 2.6191051410900653e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9856
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.4375,
      "epoch": 0.4762756088132973,
      "grad_norm": 0.0,
      "learning_rate": 2.6188635485117894e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9857
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.15625,
      "epoch": 0.47632392732895246,
      "grad_norm": 0.0,
      "learning_rate": 2.6186219559335136e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9858
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.5,
      "epoch": 0.47637224584460763,
      "grad_norm": 0.0,
      "learning_rate": 2.6183803633552377e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9859
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.53125,
      "epoch": 0.47642056436026287,
      "grad_norm": 0.0,
      "learning_rate": 2.6181387707769613e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9860
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.5,
      "epoch": 0.47646888287591804,
      "grad_norm": 0.0,
      "learning_rate": 2.617897178198686e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 9861
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.46875,
      "epoch": 0.4765172013915733,
      "grad_norm": 0.0,
      "learning_rate": 2.6176555856204096e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9862
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.84375,
      "epoch": 0.47656551990722845,
      "grad_norm": 0.0,
      "learning_rate": 2.6174139930421337e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9863
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.4375,
      "epoch": 0.47661383842288363,
      "grad_norm": 0.0,
      "learning_rate": 2.617172400463858e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9864
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.5625,
      "epoch": 0.47666215693853886,
      "grad_norm": 0.0,
      "learning_rate": 2.616930807885582e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9865
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.125,
      "epoch": 0.47671047545419404,
      "grad_norm": 0.0,
      "learning_rate": 2.6166892153073056e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9866
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.3125,
      "epoch": 0.47675879396984927,
      "grad_norm": 0.0,
      "learning_rate": 2.616447622729029e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9867
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.34375,
      "epoch": 0.47680711248550445,
      "grad_norm": 0.0,
      "learning_rate": 2.616206030150754e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9868
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.40625,
      "epoch": 0.4768554310011596,
      "grad_norm": 0.0,
      "learning_rate": 2.6159644375724775e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9869
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.84375,
      "epoch": 0.47690374951681486,
      "grad_norm": 0.0,
      "learning_rate": 2.6157228449942016e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9870
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.84375,
      "epoch": 0.47695206803247003,
      "grad_norm": 0.0,
      "learning_rate": 2.615481252415926e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9871
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.4375,
      "epoch": 0.47700038654812527,
      "grad_norm": 0.0,
      "learning_rate": 2.61523965983765e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9872
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.78125,
      "epoch": 0.47704870506378044,
      "grad_norm": 0.0,
      "learning_rate": 2.6149980672593735e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9873
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.4375,
      "epoch": 0.4770970235794356,
      "grad_norm": 0.0,
      "learning_rate": 2.614756474681098e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9874
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.875,
      "epoch": 0.47714534209509085,
      "grad_norm": 0.0,
      "learning_rate": 2.614514882102822e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9875
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.0625,
      "epoch": 0.47719366061074603,
      "grad_norm": 0.0,
      "learning_rate": 2.6142732895245454e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9876
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.28125,
      "epoch": 0.47724197912640126,
      "grad_norm": 0.0,
      "learning_rate": 2.6140316969462695e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9877
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.09375,
      "epoch": 0.47729029764205644,
      "grad_norm": 0.0,
      "learning_rate": 2.6137901043679936e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9878
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.8125,
      "epoch": 0.4773386161577116,
      "grad_norm": 0.0,
      "learning_rate": 2.613548511789718e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9879
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.40625,
      "epoch": 0.47738693467336685,
      "grad_norm": 0.0,
      "learning_rate": 2.6133069192114414e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9880
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.34375,
      "epoch": 0.477435253189022,
      "grad_norm": 0.0,
      "learning_rate": 2.613065326633166e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9881
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.03125,
      "epoch": 0.47748357170467726,
      "grad_norm": 0.0,
      "learning_rate": 2.6128237340548897e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9882
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.40625,
      "epoch": 0.47753189022033243,
      "grad_norm": 0.0,
      "learning_rate": 2.612582141476614e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9883
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.90625,
      "epoch": 0.4775802087359876,
      "grad_norm": 0.0,
      "learning_rate": 2.612340548898338e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9884
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.1875,
      "epoch": 0.47762852725164284,
      "grad_norm": 0.0,
      "learning_rate": 2.612098956320062e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9885
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.09375,
      "epoch": 0.477676845767298,
      "grad_norm": 0.0,
      "learning_rate": 2.6118573637417857e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9886
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.28125,
      "epoch": 0.47772516428295325,
      "grad_norm": 0.0,
      "learning_rate": 2.6116157711635093e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9887
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.0,
      "epoch": 0.47777348279860843,
      "grad_norm": 0.0,
      "learning_rate": 2.611374178585234e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9888
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.15625,
      "epoch": 0.4778218013142636,
      "grad_norm": 0.0,
      "learning_rate": 2.6111325860069575e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9889
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.25,
      "epoch": 0.47787011982991884,
      "grad_norm": 0.0,
      "learning_rate": 2.6108909934286817e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9890
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.125,
      "epoch": 0.477918438345574,
      "grad_norm": 0.0,
      "learning_rate": 2.610649400850406e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9891
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.4375,
      "epoch": 0.47796675686122925,
      "grad_norm": 0.0,
      "learning_rate": 2.61040780827213e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9892
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.40625,
      "epoch": 0.4780150753768844,
      "grad_norm": 0.0,
      "learning_rate": 2.6101662156938536e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9893
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.46875,
      "epoch": 0.4780633938925396,
      "grad_norm": 0.0,
      "learning_rate": 2.609924623115578e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9894
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.71875,
      "epoch": 0.47811171240819483,
      "grad_norm": 0.0,
      "learning_rate": 2.609683030537302e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9895
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.4375,
      "epoch": 0.47816003092385,
      "grad_norm": 0.0,
      "learning_rate": 2.609441437959026e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9896
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.375,
      "epoch": 0.47820834943950524,
      "grad_norm": 0.0,
      "learning_rate": 2.6091998453807496e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9897
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.4375,
      "epoch": 0.4782566679551604,
      "grad_norm": 0.0,
      "learning_rate": 2.6089582528024737e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9898
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.71875,
      "epoch": 0.4783049864708156,
      "grad_norm": 0.0,
      "learning_rate": 2.608716660224198e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 9899
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.03125,
      "epoch": 0.47835330498647083,
      "grad_norm": 0.0,
      "learning_rate": 2.6084750676459215e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9900
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.625,
      "epoch": 0.478401623502126,
      "grad_norm": 0.0,
      "learning_rate": 2.608233475067646e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9901
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.34375,
      "epoch": 0.47844994201778124,
      "grad_norm": 0.0,
      "learning_rate": 2.6079918824893697e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9902
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.53125,
      "epoch": 0.4784982605334364,
      "grad_norm": 0.0,
      "learning_rate": 2.607750289911094e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 9903
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 78.5625,
      "epoch": 0.4785465790490916,
      "grad_norm": 172.84883252117078,
      "learning_rate": 2.607508697332818e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9904
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.4375,
      "epoch": 0.4785948975647468,
      "grad_norm": 0.0,
      "learning_rate": 2.607267104754542e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9905
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.125,
      "epoch": 0.478643216080402,
      "grad_norm": 0.0,
      "learning_rate": 2.607025512176266e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9906
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.5,
      "epoch": 0.47869153459605723,
      "grad_norm": 0.0,
      "learning_rate": 2.6067839195979894e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9907
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.8125,
      "epoch": 0.4787398531117124,
      "grad_norm": 0.0,
      "learning_rate": 2.606542327019714e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9908
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.8125,
      "epoch": 0.4787881716273676,
      "grad_norm": 62.35164125333828,
      "learning_rate": 2.6063007344414376e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.3187600374221802,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9909
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 73.96875,
      "epoch": 0.4788364901430228,
      "grad_norm": 47.29194857239291,
      "learning_rate": 2.606059141863162e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9910
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.625,
      "epoch": 0.478884808658678,
      "grad_norm": 0.0,
      "learning_rate": 2.605817549284886e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9911
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.875,
      "epoch": 0.47893312717433323,
      "grad_norm": 0.0,
      "learning_rate": 2.60557595670661e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9912
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.0625,
      "epoch": 0.4789814456899884,
      "grad_norm": 0.0,
      "learning_rate": 2.6053343641283336e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9913
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.34375,
      "epoch": 0.4790297642056436,
      "grad_norm": 0.0,
      "learning_rate": 2.6050927715500583e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9914
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.9375,
      "epoch": 0.4790780827212988,
      "grad_norm": 0.0,
      "learning_rate": 2.604851178971782e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 9915
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.8125,
      "epoch": 0.479126401236954,
      "grad_norm": 0.0,
      "learning_rate": 2.604609586393506e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9916
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.46875,
      "epoch": 0.4791747197526092,
      "grad_norm": 0.0,
      "learning_rate": 2.6043679938152297e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9917
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.84375,
      "epoch": 0.4792230382682644,
      "grad_norm": 0.0,
      "learning_rate": 2.604126401236954e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9918
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.34375,
      "epoch": 0.4792713567839196,
      "grad_norm": 0.0,
      "learning_rate": 2.603884808658678e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9919
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.59375,
      "epoch": 0.4793196752995748,
      "grad_norm": 0.0,
      "learning_rate": 2.6036432160804015e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9920
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.96875,
      "epoch": 0.47936799381523,
      "grad_norm": 0.0,
      "learning_rate": 2.603401623502126e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9921
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.3125,
      "epoch": 0.4794163123308852,
      "grad_norm": 0.0,
      "learning_rate": 2.60316003092385e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9922
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.90625,
      "epoch": 0.4794646308465404,
      "grad_norm": 0.0,
      "learning_rate": 2.602918438345574e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9923
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.6875,
      "epoch": 0.4795129493621956,
      "grad_norm": 0.0,
      "learning_rate": 2.602676845767298e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9924
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.5,
      "epoch": 0.4795612678778508,
      "grad_norm": 0.0,
      "learning_rate": 2.602435253189022e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9925
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.8125,
      "epoch": 0.479609586393506,
      "grad_norm": 0.0,
      "learning_rate": 2.602193660610746e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9926
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.46875,
      "epoch": 0.4796579049091612,
      "grad_norm": 0.0,
      "learning_rate": 2.6019520680324694e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9927
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.9375,
      "epoch": 0.4797062234248164,
      "grad_norm": 52.70887769231504,
      "learning_rate": 2.601710475454194e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.3230288028717041,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9928
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.40625,
      "epoch": 0.47975454194047157,
      "grad_norm": 0.0,
      "learning_rate": 2.6014688828759177e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9929
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.875,
      "epoch": 0.4798028604561268,
      "grad_norm": 0.0,
      "learning_rate": 2.601227290297642e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9930
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.5625,
      "epoch": 0.479851178971782,
      "grad_norm": 0.0,
      "learning_rate": 2.600985697719366e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9931
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.78125,
      "epoch": 0.4798994974874372,
      "grad_norm": 0.0,
      "learning_rate": 2.60074410514109e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9932
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.71875,
      "epoch": 0.4799478160030924,
      "grad_norm": 0.0,
      "learning_rate": 2.6005025125628137e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9933
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.84375,
      "epoch": 0.47999613451874756,
      "grad_norm": 0.0,
      "learning_rate": 2.6002609199845384e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9934
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.4375,
      "epoch": 0.4800444530344028,
      "grad_norm": 0.0,
      "learning_rate": 2.600019327406262e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9935
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.84375,
      "epoch": 0.480092771550058,
      "grad_norm": 0.0,
      "learning_rate": 2.599777734827986e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9936
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.9375,
      "epoch": 0.4801410900657132,
      "grad_norm": 0.0,
      "learning_rate": 2.59953614224971e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9937
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 84.03125,
      "epoch": 0.4801894085813684,
      "grad_norm": 23.182269654045182,
      "learning_rate": 2.599294549671434e-07,
      "loss": -0.0,
      "reward": 1.15625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.125,
      "rewards/bonus_rewards": 0.03125,
      "rewards/format_reward": 1.0,
      "step": 9938
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.28125,
      "epoch": 0.48023772709702356,
      "grad_norm": 0.0,
      "learning_rate": 2.599052957093158e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9939
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.28125,
      "epoch": 0.4802860456126788,
      "grad_norm": 0.0,
      "learning_rate": 2.5988113645148816e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9940
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.84375,
      "epoch": 0.48033436412833397,
      "grad_norm": 0.0,
      "learning_rate": 2.5985697719366063e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9941
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.15625,
      "epoch": 0.4803826826439892,
      "grad_norm": 0.0,
      "learning_rate": 2.59832817935833e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9942
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.46875,
      "epoch": 0.4804310011596444,
      "grad_norm": 0.0,
      "learning_rate": 2.598086586780054e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9943
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.5,
      "epoch": 0.48047931967529955,
      "grad_norm": 0.0,
      "learning_rate": 2.597844994201778e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9944
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.28125,
      "epoch": 0.4805276381909548,
      "grad_norm": 0.0,
      "learning_rate": 2.5976034016235023e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9945
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.34375,
      "epoch": 0.48057595670660996,
      "grad_norm": 0.0,
      "learning_rate": 2.597361809045226e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 9946
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.03125,
      "epoch": 0.4806242752222652,
      "grad_norm": 0.0,
      "learning_rate": 2.59712021646695e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 9947
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.375,
      "epoch": 0.4806725937379204,
      "grad_norm": 0.0,
      "learning_rate": 2.596878623888674e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9948
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.8125,
      "epoch": 0.48072091225357555,
      "grad_norm": 0.0,
      "learning_rate": 2.596637031310398e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 9949
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.8125,
      "epoch": 0.4807692307692308,
      "grad_norm": 0.0,
      "learning_rate": 2.596395438732122e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 9950
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.875,
      "epoch": 0.48081754928488596,
      "grad_norm": 0.0,
      "learning_rate": 2.596153846153846e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9951
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.0625,
      "epoch": 0.4808658678005412,
      "grad_norm": 0.0,
      "learning_rate": 2.59591225357557e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9952
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.59375,
      "epoch": 0.48091418631619637,
      "grad_norm": 0.0,
      "learning_rate": 2.595670660997294e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9953
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.59375,
      "epoch": 0.48096250483185155,
      "grad_norm": 0.0,
      "learning_rate": 2.5954290684190185e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 9954
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.75,
      "epoch": 0.4810108233475068,
      "grad_norm": 0.0,
      "learning_rate": 2.595187475840742e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9955
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.6875,
      "epoch": 0.48105914186316195,
      "grad_norm": 0.0,
      "learning_rate": 2.594945883262466e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9956
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.375,
      "epoch": 0.4811074603788172,
      "grad_norm": 0.0,
      "learning_rate": 2.59470429068419e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9957
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.25,
      "epoch": 0.48115577889447236,
      "grad_norm": 0.0,
      "learning_rate": 2.5944626981059145e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9958
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.1875,
      "epoch": 0.48120409741012754,
      "grad_norm": 0.0,
      "learning_rate": 2.594221105527638e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9959
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.75,
      "epoch": 0.4812524159257828,
      "grad_norm": 0.0,
      "learning_rate": 2.5939795129493617e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9960
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.09375,
      "epoch": 0.48130073444143795,
      "grad_norm": 0.0,
      "learning_rate": 2.5937379203710864e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 9961
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.53125,
      "epoch": 0.4813490529570932,
      "grad_norm": 0.0,
      "learning_rate": 2.59349632779281e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9962
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.59375,
      "epoch": 0.48139737147274836,
      "grad_norm": 0.0,
      "learning_rate": 2.593254735214534e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9963
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.84375,
      "epoch": 0.48144568998840354,
      "grad_norm": 0.0,
      "learning_rate": 2.593013142636258e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9964
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.40625,
      "epoch": 0.48149400850405877,
      "grad_norm": 0.0,
      "learning_rate": 2.5927715500579824e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 9965
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.84375,
      "epoch": 0.48154232701971394,
      "grad_norm": 0.0,
      "learning_rate": 2.592529957479706e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9966
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.0,
      "epoch": 0.4815906455353692,
      "grad_norm": 0.0,
      "learning_rate": 2.59228836490143e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9967
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.6875,
      "epoch": 0.48163896405102435,
      "grad_norm": 0.0,
      "learning_rate": 2.592046772323154e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 9968
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.625,
      "epoch": 0.48168728256667953,
      "grad_norm": 0.0,
      "learning_rate": 2.591805179744878e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 9969
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.125,
      "epoch": 0.48173560108233476,
      "grad_norm": 0.0,
      "learning_rate": 2.591563587166602e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 9970
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.25,
      "epoch": 0.48178391959798994,
      "grad_norm": 0.0,
      "learning_rate": 2.591321994588326e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9971
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.5,
      "epoch": 0.4818322381136452,
      "grad_norm": 0.0,
      "learning_rate": 2.5910804020100503e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 9972
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.59375,
      "epoch": 0.48188055662930035,
      "grad_norm": 0.0,
      "learning_rate": 2.590838809431774e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9973
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.15625,
      "epoch": 0.4819288751449555,
      "grad_norm": 0.0,
      "learning_rate": 2.5905972168534985e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9974
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.375,
      "epoch": 0.48197719366061076,
      "grad_norm": 0.0,
      "learning_rate": 2.590355624275222e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9975
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.8125,
      "epoch": 0.48202551217626594,
      "grad_norm": 0.0,
      "learning_rate": 2.5901140316969463e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9976
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.625,
      "epoch": 0.48207383069192117,
      "grad_norm": 0.0,
      "learning_rate": 2.58987243911867e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 9977
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.71875,
      "epoch": 0.48212214920757634,
      "grad_norm": 0.0,
      "learning_rate": 2.5896308465403946e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 9978
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.78125,
      "epoch": 0.4821704677232315,
      "grad_norm": 0.0,
      "learning_rate": 2.589389253962118e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9979
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.84375,
      "epoch": 0.48221878623888675,
      "grad_norm": 0.0,
      "learning_rate": 2.589147661383842e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 9980
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.21875,
      "epoch": 0.48226710475454193,
      "grad_norm": 0.0,
      "learning_rate": 2.5889060688055664e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9981
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 77.71875,
      "epoch": 0.48231542327019716,
      "grad_norm": 0.0,
      "learning_rate": 2.58866447622729e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 9982
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.5,
      "epoch": 0.48236374178585234,
      "grad_norm": 0.0,
      "learning_rate": 2.588422883649014e-07,
      "loss": 0.0,
      "reward": 1.25,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 0.875,
      "step": 9983
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.1875,
      "epoch": 0.4824120603015075,
      "grad_norm": 0.0,
      "learning_rate": 2.5881812910707383e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 9984
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.4375,
      "epoch": 0.48246037881716275,
      "grad_norm": 0.0,
      "learning_rate": 2.5879396984924625e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 9985
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.9375,
      "epoch": 0.4825086973328179,
      "grad_norm": 0.0,
      "learning_rate": 2.587698105914186e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 9986
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.8125,
      "epoch": 0.48255701584847316,
      "grad_norm": 0.0,
      "learning_rate": 2.58745651333591e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9987
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.71875,
      "epoch": 0.48260533436412834,
      "grad_norm": 0.0,
      "learning_rate": 2.5872149207576343e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 9988
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.25,
      "epoch": 0.4826536528797835,
      "grad_norm": 0.0,
      "learning_rate": 2.586973328179358e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 9989
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.375,
      "epoch": 0.48270197139543874,
      "grad_norm": 0.0,
      "learning_rate": 2.586731735601082e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 9990
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.65625,
      "epoch": 0.4827502899110939,
      "grad_norm": 0.0,
      "learning_rate": 2.586490143022806e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9991
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.09375,
      "epoch": 0.48279860842674915,
      "grad_norm": 0.0,
      "learning_rate": 2.5862485504445304e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9992
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.1875,
      "epoch": 0.48284692694240433,
      "grad_norm": 0.0,
      "learning_rate": 2.586006957866254e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9993
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.46875,
      "epoch": 0.4828952454580595,
      "grad_norm": 0.0,
      "learning_rate": 2.5857653652879786e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 9994
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.46875,
      "epoch": 0.48294356397371474,
      "grad_norm": 0.0,
      "learning_rate": 2.585523772709702e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 9995
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.15625,
      "epoch": 0.4829918824893699,
      "grad_norm": 0.0,
      "learning_rate": 2.5852821801314264e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9996
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.625,
      "epoch": 0.48304020100502515,
      "grad_norm": 0.0,
      "learning_rate": 2.58504058755315e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 9997
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.5625,
      "epoch": 0.4830885195206803,
      "grad_norm": 0.0,
      "learning_rate": 2.5847989949748746e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 9998
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.90625,
      "epoch": 0.4831368380363355,
      "grad_norm": 0.0,
      "learning_rate": 2.584557402396598e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 9999
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.375,
      "epoch": 0.48318515655199074,
      "grad_norm": 0.0,
      "learning_rate": 2.584315809818322e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10000
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.71875,
      "epoch": 0.4832334750676459,
      "grad_norm": 0.0,
      "learning_rate": 2.5840742172400465e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10001
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.15625,
      "epoch": 0.48328179358330114,
      "grad_norm": 0.0,
      "learning_rate": 2.58383262466177e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10002
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.84375,
      "epoch": 0.4833301120989563,
      "grad_norm": 0.0,
      "learning_rate": 2.583591032083494e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10003
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.53125,
      "epoch": 0.4833784306146115,
      "grad_norm": 0.0,
      "learning_rate": 2.5833494395052184e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10004
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.90625,
      "epoch": 0.48342674913026673,
      "grad_norm": 0.0,
      "learning_rate": 2.5831078469269425e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10005
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.59375,
      "epoch": 0.4834750676459219,
      "grad_norm": 0.0,
      "learning_rate": 2.582866254348666e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10006
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.25,
      "epoch": 0.48352338616157714,
      "grad_norm": 0.0,
      "learning_rate": 2.5826246617703903e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10007
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.15625,
      "epoch": 0.4835717046772323,
      "grad_norm": 0.0,
      "learning_rate": 2.5823830691921144e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10008
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.875,
      "epoch": 0.4836200231928875,
      "grad_norm": 0.0,
      "learning_rate": 2.5821414766138385e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10009
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.46875,
      "epoch": 0.4836683417085427,
      "grad_norm": 0.0,
      "learning_rate": 2.581899884035562e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10010
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 76.4375,
      "epoch": 0.4837166602241979,
      "grad_norm": 12.896945362868331,
      "learning_rate": 2.5816582914572863e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.3445242941379547,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 0.96875,
      "step": 10011
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.375,
      "epoch": 0.48376497873985314,
      "grad_norm": 0.0,
      "learning_rate": 2.5814166988790104e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10012
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.375,
      "epoch": 0.4838132972555083,
      "grad_norm": 0.0,
      "learning_rate": 2.581175106300734e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10013
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.28125,
      "epoch": 0.4838616157711635,
      "grad_norm": 0.0,
      "learning_rate": 2.5809335137224587e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10014
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.1875,
      "epoch": 0.4839099342868187,
      "grad_norm": 0.0,
      "learning_rate": 2.5806919211441823e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10015
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.96875,
      "epoch": 0.4839582528024739,
      "grad_norm": 0.0,
      "learning_rate": 2.5804503285659064e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10016
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.0,
      "epoch": 0.48400657131812913,
      "grad_norm": 0.0,
      "learning_rate": 2.58020873598763e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10017
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 80.75,
      "epoch": 0.4840548898337843,
      "grad_norm": 165.22931507327633,
      "learning_rate": 2.5799671434093547e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10018
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.71875,
      "epoch": 0.4841032083494395,
      "grad_norm": 0.0,
      "learning_rate": 2.5797255508310783e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10019
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.59375,
      "epoch": 0.4841515268650947,
      "grad_norm": 0.0,
      "learning_rate": 2.579483958252802e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10020
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.21875,
      "epoch": 0.4841998453807499,
      "grad_norm": 0.0,
      "learning_rate": 2.5792423656745266e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10021
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.125,
      "epoch": 0.4842481638964051,
      "grad_norm": 0.0,
      "learning_rate": 2.57900077309625e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10022
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 81.28125,
      "epoch": 0.4842964824120603,
      "grad_norm": 28.120339650085686,
      "learning_rate": 2.5787591805179743e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10023
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.5625,
      "epoch": 0.4843448009277155,
      "grad_norm": 0.0,
      "learning_rate": 2.5785175879396985e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10024
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.1875,
      "epoch": 0.4843931194433707,
      "grad_norm": 0.0,
      "learning_rate": 2.5782759953614226e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10025
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.34375,
      "epoch": 0.4844414379590259,
      "grad_norm": 0.0,
      "learning_rate": 2.578034402783146e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10026
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.71875,
      "epoch": 0.4844897564746811,
      "grad_norm": 0.0,
      "learning_rate": 2.5777928102048704e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10027
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.40625,
      "epoch": 0.4845380749903363,
      "grad_norm": 0.0,
      "learning_rate": 2.5775512176265945e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10028
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.75,
      "epoch": 0.4845863935059915,
      "grad_norm": 0.0,
      "learning_rate": 2.5773096250483186e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10029
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.40625,
      "epoch": 0.4846347120216467,
      "grad_norm": 0.0,
      "learning_rate": 2.577068032470042e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10030
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.8125,
      "epoch": 0.4846830305373019,
      "grad_norm": 0.0,
      "learning_rate": 2.5768264398917664e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10031
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.53125,
      "epoch": 0.4847313490529571,
      "grad_norm": 0.0,
      "learning_rate": 2.5765848473134905e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10032
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.5,
      "epoch": 0.4847796675686123,
      "grad_norm": 0.0,
      "learning_rate": 2.576343254735214e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10033
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.90625,
      "epoch": 0.48482798608426747,
      "grad_norm": 55.61118332814823,
      "learning_rate": 2.576101662156939e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10034
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.53125,
      "epoch": 0.4848763045999227,
      "grad_norm": 0.0,
      "learning_rate": 2.5758600695786624e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 10035
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.4375,
      "epoch": 0.4849246231155779,
      "grad_norm": 0.0,
      "learning_rate": 2.5756184770003865e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10036
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.65625,
      "epoch": 0.4849729416312331,
      "grad_norm": 0.0,
      "learning_rate": 2.57537688442211e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10037
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.5,
      "epoch": 0.4850212601468883,
      "grad_norm": 0.0,
      "learning_rate": 2.575135291843835e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10038
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.75,
      "epoch": 0.48506957866254347,
      "grad_norm": 0.0,
      "learning_rate": 2.5748936992655584e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10039
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.28125,
      "epoch": 0.4851178971781987,
      "grad_norm": 0.0,
      "learning_rate": 2.5746521066872825e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10040
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.5,
      "epoch": 0.4851662156938539,
      "grad_norm": 0.0,
      "learning_rate": 2.5744105141090067e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10041
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.25,
      "epoch": 0.4852145342095091,
      "grad_norm": 0.0,
      "learning_rate": 2.5741689215307303e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10042
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.0625,
      "epoch": 0.4852628527251643,
      "grad_norm": 69.15947616525864,
      "learning_rate": 2.5739273289524544e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10043
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.59375,
      "epoch": 0.48531117124081946,
      "grad_norm": 0.0,
      "learning_rate": 2.5736857363741786e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10044
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.8125,
      "epoch": 0.4853594897564747,
      "grad_norm": 3.567632921798575,
      "learning_rate": 2.5734441437959027e-07,
      "loss": -0.0,
      "reward": 2.265625,
      "reward_std": 0.40822193026542664,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 0.96875,
      "step": 10045
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.3125,
      "epoch": 0.48540780827212987,
      "grad_norm": 0.0,
      "learning_rate": 2.5732025512176263e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10046
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.6875,
      "epoch": 0.4854561267877851,
      "grad_norm": 0.0,
      "learning_rate": 2.5729609586393504e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10047
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.59375,
      "epoch": 0.4855044453034403,
      "grad_norm": 0.0,
      "learning_rate": 2.5727193660610746e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10048
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.625,
      "epoch": 0.48555276381909546,
      "grad_norm": 0.0,
      "learning_rate": 2.5724777734827987e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10049
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.125,
      "epoch": 0.4856010823347507,
      "grad_norm": 0.0,
      "learning_rate": 2.5722361809045223e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10050
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.0,
      "epoch": 0.48564940085040587,
      "grad_norm": 0.0,
      "learning_rate": 2.571994588326247e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10051
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.59375,
      "epoch": 0.4856977193660611,
      "grad_norm": 0.0,
      "learning_rate": 2.5717529957479706e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10052
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.46875,
      "epoch": 0.4857460378817163,
      "grad_norm": 0.0,
      "learning_rate": 2.571511403169694e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10053
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.28125,
      "epoch": 0.48579435639737145,
      "grad_norm": 0.0,
      "learning_rate": 2.571269810591419e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10054
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.3125,
      "epoch": 0.4858426749130267,
      "grad_norm": 0.0,
      "learning_rate": 2.5710282180131425e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10055
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.53125,
      "epoch": 0.48589099342868186,
      "grad_norm": 0.0,
      "learning_rate": 2.5707866254348666e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10056
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.1875,
      "epoch": 0.4859393119443371,
      "grad_norm": 0.0,
      "learning_rate": 2.57054503285659e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10057
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.96875,
      "epoch": 0.48598763045999227,
      "grad_norm": 0.0,
      "learning_rate": 2.570303440278315e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10058
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.5625,
      "epoch": 0.48603594897564745,
      "grad_norm": 0.0,
      "learning_rate": 2.5700618477000385e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10059
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.15625,
      "epoch": 0.4860842674913027,
      "grad_norm": 0.0,
      "learning_rate": 2.5698202551217626e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10060
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.1875,
      "epoch": 0.48613258600695786,
      "grad_norm": 0.0,
      "learning_rate": 2.569578662543487e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10061
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.59375,
      "epoch": 0.4861809045226131,
      "grad_norm": 0.0,
      "learning_rate": 2.5693370699652104e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10062
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.34375,
      "epoch": 0.48622922303826827,
      "grad_norm": 0.0,
      "learning_rate": 2.5690954773869345e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10063
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.4375,
      "epoch": 0.48627754155392344,
      "grad_norm": 0.0,
      "learning_rate": 2.5688538848086586e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10064
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.625,
      "epoch": 0.4863258600695787,
      "grad_norm": 0.0,
      "learning_rate": 2.568612292230383e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10065
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.46875,
      "epoch": 0.48637417858523385,
      "grad_norm": 0.0,
      "learning_rate": 2.5683706996521064e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10066
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.625,
      "epoch": 0.4864224971008891,
      "grad_norm": 0.0,
      "learning_rate": 2.5681291070738305e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10067
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.5625,
      "epoch": 0.48647081561654426,
      "grad_norm": 0.0,
      "learning_rate": 2.5678875144955546e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10068
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.15625,
      "epoch": 0.48651913413219944,
      "grad_norm": 0.0,
      "learning_rate": 2.567645921917279e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10069
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.90625,
      "epoch": 0.48656745264785467,
      "grad_norm": 0.0,
      "learning_rate": 2.5674043293390024e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10070
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.03125,
      "epoch": 0.48661577116350985,
      "grad_norm": 0.0,
      "learning_rate": 2.567162736760727e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10071
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.09375,
      "epoch": 0.4866640896791651,
      "grad_norm": 0.0,
      "learning_rate": 2.5669211441824507e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10072
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.4375,
      "epoch": 0.48671240819482026,
      "grad_norm": 0.0,
      "learning_rate": 2.5666795516041743e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10073
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.53125,
      "epoch": 0.48676072671047543,
      "grad_norm": 0.0,
      "learning_rate": 2.566437959025899e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10074
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.3125,
      "epoch": 0.48680904522613067,
      "grad_norm": 0.0,
      "learning_rate": 2.5661963664476225e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10075
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.65625,
      "epoch": 0.48685736374178584,
      "grad_norm": 0.0,
      "learning_rate": 2.5659547738693467e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10076
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.125,
      "epoch": 0.4869056822574411,
      "grad_norm": 0.0,
      "learning_rate": 2.5657131812910703e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10077
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.0,
      "epoch": 0.48695400077309625,
      "grad_norm": 0.0,
      "learning_rate": 2.565471588712795e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10078
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.90625,
      "epoch": 0.48700231928875143,
      "grad_norm": 0.0,
      "learning_rate": 2.5652299961345186e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10079
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.875,
      "epoch": 0.48705063780440666,
      "grad_norm": 0.0,
      "learning_rate": 2.5649884035562427e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10080
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.5,
      "epoch": 0.48709895632006184,
      "grad_norm": 0.0,
      "learning_rate": 2.564746810977967e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10081
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 88.9375,
      "epoch": 0.48714727483571707,
      "grad_norm": 50.32088698914194,
      "learning_rate": 2.5645052183996904e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10082
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.625,
      "epoch": 0.48719559335137225,
      "grad_norm": 0.0,
      "learning_rate": 2.5642636258214146e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10083
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.28125,
      "epoch": 0.4872439118670274,
      "grad_norm": 0.0,
      "learning_rate": 2.5640220332431387e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10084
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 93.125,
      "epoch": 0.48729223038268266,
      "grad_norm": 14.506485727432434,
      "learning_rate": 2.563780440664863e-07,
      "loss": -0.0,
      "reward": 1.734375,
      "reward_std": 0.2573546767234802,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10085
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.4375,
      "epoch": 0.48734054889833783,
      "grad_norm": 0.0,
      "learning_rate": 2.5635388480865865e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 10086
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.96875,
      "epoch": 0.48738886741399307,
      "grad_norm": 0.0,
      "learning_rate": 2.5632972555083106e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10087
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.46875,
      "epoch": 0.48743718592964824,
      "grad_norm": 0.0,
      "learning_rate": 2.5630556629300347e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10088
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.375,
      "epoch": 0.4874855044453034,
      "grad_norm": 0.0,
      "learning_rate": 2.562814070351759e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10089
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.84375,
      "epoch": 0.48753382296095865,
      "grad_norm": 0.0,
      "learning_rate": 2.5625724777734825e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10090
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.1875,
      "epoch": 0.48758214147661383,
      "grad_norm": 0.0,
      "learning_rate": 2.562330885195207e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10091
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.5625,
      "epoch": 0.48763045999226906,
      "grad_norm": 0.0,
      "learning_rate": 2.562089292616931e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10092
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.71875,
      "epoch": 0.48767877850792424,
      "grad_norm": 0.0,
      "learning_rate": 2.5618477000386544e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10093
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.6875,
      "epoch": 0.4877270970235794,
      "grad_norm": 0.0,
      "learning_rate": 2.561606107460379e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10094
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.1875,
      "epoch": 0.48777541553923465,
      "grad_norm": 0.0,
      "learning_rate": 2.5613645148821026e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10095
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.65625,
      "epoch": 0.4878237340548898,
      "grad_norm": 0.0,
      "learning_rate": 2.561122922303827e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10096
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.5625,
      "epoch": 0.48787205257054506,
      "grad_norm": 0.0,
      "learning_rate": 2.5608813297255504e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10097
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.65625,
      "epoch": 0.48792037108620023,
      "grad_norm": 0.0,
      "learning_rate": 2.560639737147275e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10098
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.5625,
      "epoch": 0.4879686896018554,
      "grad_norm": 0.0,
      "learning_rate": 2.5603981445689986e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10099
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.4375,
      "epoch": 0.48801700811751064,
      "grad_norm": 0.0,
      "learning_rate": 2.560156551990723e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10100
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.0625,
      "epoch": 0.4880653266331658,
      "grad_norm": 0.0,
      "learning_rate": 2.559914959412447e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10101
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.96875,
      "epoch": 0.48811364514882105,
      "grad_norm": 0.0,
      "learning_rate": 2.559673366834171e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10102
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.21875,
      "epoch": 0.48816196366447623,
      "grad_norm": 26.598940273132815,
      "learning_rate": 2.5594317742558947e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10103
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.75,
      "epoch": 0.4882102821801314,
      "grad_norm": 0.0,
      "learning_rate": 2.559190181677619e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10104
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.9375,
      "epoch": 0.48825860069578664,
      "grad_norm": 0.0,
      "learning_rate": 2.558948589099343e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10105
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.125,
      "epoch": 0.4883069192114418,
      "grad_norm": 0.0,
      "learning_rate": 2.5587069965210665e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10106
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.28125,
      "epoch": 0.48835523772709705,
      "grad_norm": 0.0,
      "learning_rate": 2.5584654039427907e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10107
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.9375,
      "epoch": 0.4884035562427522,
      "grad_norm": 0.0,
      "learning_rate": 2.558223811364515e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10108
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.0625,
      "epoch": 0.4884518747584074,
      "grad_norm": 0.0,
      "learning_rate": 2.557982218786239e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10109
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.34375,
      "epoch": 0.48850019327406263,
      "grad_norm": 0.0,
      "learning_rate": 2.5577406262079625e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10110
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.4375,
      "epoch": 0.4885485117897178,
      "grad_norm": 0.0,
      "learning_rate": 2.557499033629687e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10111
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.09375,
      "epoch": 0.48859683030537304,
      "grad_norm": 0.0,
      "learning_rate": 2.557257441051411e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10112
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.71875,
      "epoch": 0.4886451488210282,
      "grad_norm": 0.0,
      "learning_rate": 2.5570158484731344e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10113
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.375,
      "epoch": 0.4886934673366834,
      "grad_norm": 0.0,
      "learning_rate": 2.556774255894859e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10114
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.75,
      "epoch": 0.48874178585233863,
      "grad_norm": 0.0,
      "learning_rate": 2.5565326633165827e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10115
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.125,
      "epoch": 0.4887901043679938,
      "grad_norm": 0.0,
      "learning_rate": 2.556291070738307e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10116
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.21875,
      "epoch": 0.48883842288364904,
      "grad_norm": 0.0,
      "learning_rate": 2.5560494781600304e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10117
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.5,
      "epoch": 0.4888867413993042,
      "grad_norm": 0.0,
      "learning_rate": 2.555807885581755e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10118
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.375,
      "epoch": 0.4889350599149594,
      "grad_norm": 0.0,
      "learning_rate": 2.5555662930034787e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10119
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.71875,
      "epoch": 0.4889833784306146,
      "grad_norm": 0.0,
      "learning_rate": 2.555324700425203e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10120
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.25,
      "epoch": 0.4890316969462698,
      "grad_norm": 0.0,
      "learning_rate": 2.555083107846927e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10121
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.6875,
      "epoch": 0.48908001546192503,
      "grad_norm": 0.0,
      "learning_rate": 2.554841515268651e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10122
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.5625,
      "epoch": 0.4891283339775802,
      "grad_norm": 0.0,
      "learning_rate": 2.5545999226903747e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10123
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 92.5,
      "epoch": 0.4891766524932354,
      "grad_norm": 20.522967137689268,
      "learning_rate": 2.554358330112099e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.37405285239219666,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10124
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 91.1875,
      "epoch": 0.4892249710088906,
      "grad_norm": 33.60773795277174,
      "learning_rate": 2.554116737533823e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.37450212240219116,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10125
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.65625,
      "epoch": 0.4892732895245458,
      "grad_norm": 0.0,
      "learning_rate": 2.5538751449555466e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10126
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.59375,
      "epoch": 0.48932160804020103,
      "grad_norm": 0.0,
      "learning_rate": 2.553633552377271e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10127
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.09375,
      "epoch": 0.4893699265558562,
      "grad_norm": 0.0,
      "learning_rate": 2.553391959798995e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10128
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.5625,
      "epoch": 0.4894182450715114,
      "grad_norm": 0.0,
      "learning_rate": 2.553150367220719e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10129
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.8125,
      "epoch": 0.4894665635871666,
      "grad_norm": 0.0,
      "learning_rate": 2.5529087746424426e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10130
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.3125,
      "epoch": 0.4895148821028218,
      "grad_norm": 0.0,
      "learning_rate": 2.5526671820641673e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10131
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.53125,
      "epoch": 0.489563200618477,
      "grad_norm": 0.0,
      "learning_rate": 2.552425589485891e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10132
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.28125,
      "epoch": 0.4896115191341322,
      "grad_norm": 45.5077866084042,
      "learning_rate": 2.5521839969076145e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10133
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.0,
      "epoch": 0.4896598376497874,
      "grad_norm": 0.0,
      "learning_rate": 2.551942404329339e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10134
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.3125,
      "epoch": 0.4897081561654426,
      "grad_norm": 0.0,
      "learning_rate": 2.551700811751063e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10135
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.125,
      "epoch": 0.4897564746810978,
      "grad_norm": 0.0,
      "learning_rate": 2.551459219172787e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10136
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.21875,
      "epoch": 0.489804793196753,
      "grad_norm": 0.0,
      "learning_rate": 2.5512176265945105e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10137
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.03125,
      "epoch": 0.4898531117124082,
      "grad_norm": 0.0,
      "learning_rate": 2.550976034016235e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10138
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 93.21875,
      "epoch": 0.4899014302280634,
      "grad_norm": 84.89885964060264,
      "learning_rate": 2.550734441437959e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.285190224647522,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10139
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 82.6875,
      "epoch": 0.4899497487437186,
      "grad_norm": 48.671277142807234,
      "learning_rate": 2.550492848859683e-07,
      "loss": -0.0,
      "reward": 2.21875,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10140
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.3125,
      "epoch": 0.4899980672593738,
      "grad_norm": 0.0,
      "learning_rate": 2.550251256281407e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10141
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.09375,
      "epoch": 0.490046385775029,
      "grad_norm": 0.0,
      "learning_rate": 2.550009663703131e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10142
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.625,
      "epoch": 0.4900947042906842,
      "grad_norm": 0.0,
      "learning_rate": 2.549768071124855e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10143
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.625,
      "epoch": 0.49014302280633937,
      "grad_norm": 0.0,
      "learning_rate": 2.549526478546579e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10144
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.59375,
      "epoch": 0.4901913413219946,
      "grad_norm": 0.0,
      "learning_rate": 2.549284885968303e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10145
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.78125,
      "epoch": 0.4902396598376498,
      "grad_norm": 0.0,
      "learning_rate": 2.5490432933900267e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10146
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.25,
      "epoch": 0.490287978353305,
      "grad_norm": 0.0,
      "learning_rate": 2.548801700811751e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10147
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.375,
      "epoch": 0.4903362968689602,
      "grad_norm": 0.0,
      "learning_rate": 2.548560108233475e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10148
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.125,
      "epoch": 0.49038461538461536,
      "grad_norm": 0.0,
      "learning_rate": 2.548318515655199e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10149
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.5,
      "epoch": 0.4904329339002706,
      "grad_norm": 0.0,
      "learning_rate": 2.5480769230769227e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10150
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.28125,
      "epoch": 0.4904812524159258,
      "grad_norm": 0.0,
      "learning_rate": 2.5478353304986474e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10151
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.4375,
      "epoch": 0.490529570931581,
      "grad_norm": 0.0,
      "learning_rate": 2.547593737920371e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10152
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.53125,
      "epoch": 0.4905778894472362,
      "grad_norm": 0.0,
      "learning_rate": 2.547352145342095e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10153
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.875,
      "epoch": 0.49062620796289136,
      "grad_norm": 0.0,
      "learning_rate": 2.547110552763819e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10154
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.90625,
      "epoch": 0.4906745264785466,
      "grad_norm": 0.0,
      "learning_rate": 2.546868960185543e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10155
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.84375,
      "epoch": 0.49072284499420177,
      "grad_norm": 0.0,
      "learning_rate": 2.546627367607267e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10156
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 109.25,
      "epoch": 0.490771163509857,
      "grad_norm": 0.0,
      "learning_rate": 2.5463857750289906e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10157
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.625,
      "epoch": 0.4908194820255122,
      "grad_norm": 0.0,
      "learning_rate": 2.5461441824507153e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10158
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.40625,
      "epoch": 0.49086780054116735,
      "grad_norm": 0.0,
      "learning_rate": 2.545902589872439e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10159
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.5,
      "epoch": 0.4909161190568226,
      "grad_norm": 0.0,
      "learning_rate": 2.545660997294163e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10160
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.03125,
      "epoch": 0.49096443757247776,
      "grad_norm": 0.0,
      "learning_rate": 2.545419404715887e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10161
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.15625,
      "epoch": 0.491012756088133,
      "grad_norm": 0.0,
      "learning_rate": 2.5451778121376113e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10162
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.8125,
      "epoch": 0.4910610746037882,
      "grad_norm": 0.0,
      "learning_rate": 2.544936219559335e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10163
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.46875,
      "epoch": 0.49110939311944335,
      "grad_norm": 0.0,
      "learning_rate": 2.5446946269810596e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 10164
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.40625,
      "epoch": 0.4911577116350986,
      "grad_norm": 0.0,
      "learning_rate": 2.544453034402783e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10165
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.375,
      "epoch": 0.49120603015075376,
      "grad_norm": 0.0,
      "learning_rate": 2.544211441824507e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10166
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.1875,
      "epoch": 0.491254348666409,
      "grad_norm": 0.0,
      "learning_rate": 2.543969849246231e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10167
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.375,
      "epoch": 0.49130266718206417,
      "grad_norm": 0.0,
      "learning_rate": 2.543728256667955e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10168
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.53125,
      "epoch": 0.49135098569771934,
      "grad_norm": 0.0,
      "learning_rate": 2.543486664089679e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10169
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.53125,
      "epoch": 0.4913993042133746,
      "grad_norm": 0.0,
      "learning_rate": 2.543245071511403e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10170
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.34375,
      "epoch": 0.49144762272902975,
      "grad_norm": 0.0,
      "learning_rate": 2.5430034789331275e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10171
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.96875,
      "epoch": 0.491495941244685,
      "grad_norm": 0.0,
      "learning_rate": 2.542761886354851e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10172
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.3125,
      "epoch": 0.49154425976034016,
      "grad_norm": 0.0,
      "learning_rate": 2.542520293776575e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10173
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.0,
      "epoch": 0.49159257827599534,
      "grad_norm": 0.0,
      "learning_rate": 2.5422787011982993e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10174
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.03125,
      "epoch": 0.4916408967916506,
      "grad_norm": 0.0,
      "learning_rate": 2.542037108620023e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10175
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.3125,
      "epoch": 0.49168921530730575,
      "grad_norm": 0.0,
      "learning_rate": 2.541795516041747e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10176
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.25,
      "epoch": 0.491737533822961,
      "grad_norm": 0.0,
      "learning_rate": 2.5415539234634707e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10177
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.75,
      "epoch": 0.49178585233861616,
      "grad_norm": 0.0,
      "learning_rate": 2.5413123308851953e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10178
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.21875,
      "epoch": 0.49183417085427134,
      "grad_norm": 0.0,
      "learning_rate": 2.541070738306919e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10179
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.875,
      "epoch": 0.49188248936992657,
      "grad_norm": 0.0,
      "learning_rate": 2.540829145728643e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10180
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.0625,
      "epoch": 0.49193080788558174,
      "grad_norm": 51.42826294992198,
      "learning_rate": 2.540587553150367e-07,
      "loss": -0.0,
      "reward": 2.109375,
      "reward_std": 0.27785414457321167,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10181
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.125,
      "epoch": 0.491979126401237,
      "grad_norm": 0.0,
      "learning_rate": 2.5403459605720914e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10182
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.8125,
      "epoch": 0.49202744491689215,
      "grad_norm": 0.0,
      "learning_rate": 2.540104367993815e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10183
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.15625,
      "epoch": 0.49207576343254733,
      "grad_norm": 0.0,
      "learning_rate": 2.5398627754155396e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10184
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.28125,
      "epoch": 0.49212408194820256,
      "grad_norm": 0.0,
      "learning_rate": 2.539621182837263e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10185
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.59375,
      "epoch": 0.49217240046385774,
      "grad_norm": 0.0,
      "learning_rate": 2.539379590258987e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10186
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.875,
      "epoch": 0.492220718979513,
      "grad_norm": 0.0,
      "learning_rate": 2.539137997680711e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.260896235704422,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10187
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.34375,
      "epoch": 0.49226903749516815,
      "grad_norm": 0.0,
      "learning_rate": 2.538896405102435e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10188
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.625,
      "epoch": 0.4923173560108233,
      "grad_norm": 0.0,
      "learning_rate": 2.538654812524159e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10189
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.6875,
      "epoch": 0.49236567452647856,
      "grad_norm": 0.0,
      "learning_rate": 2.538413219945883e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10190
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.09375,
      "epoch": 0.49241399304213374,
      "grad_norm": 0.0,
      "learning_rate": 2.5381716273676075e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10191
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.4375,
      "epoch": 0.49246231155778897,
      "grad_norm": 0.0,
      "learning_rate": 2.537930034789331e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10192
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.46875,
      "epoch": 0.49251063007344414,
      "grad_norm": 0.0,
      "learning_rate": 2.5376884422110553e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10193
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.09375,
      "epoch": 0.4925589485890993,
      "grad_norm": 0.0,
      "learning_rate": 2.5374468496327794e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10194
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.71875,
      "epoch": 0.49260726710475455,
      "grad_norm": 0.0,
      "learning_rate": 2.537205257054503e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10195
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 87.46875,
      "epoch": 0.49265558562040973,
      "grad_norm": 18.64347663166287,
      "learning_rate": 2.536963664476227e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10196
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.96875,
      "epoch": 0.49270390413606496,
      "grad_norm": 0.0,
      "learning_rate": 2.536722071897951e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10197
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.9375,
      "epoch": 0.49275222265172014,
      "grad_norm": 0.0,
      "learning_rate": 2.5364804793196754e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10198
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.28125,
      "epoch": 0.4928005411673753,
      "grad_norm": 0.0,
      "learning_rate": 2.536238886741399e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10199
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.75,
      "epoch": 0.49284885968303055,
      "grad_norm": 0.0,
      "learning_rate": 2.535997294163123e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10200
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.375,
      "epoch": 0.4928971781986857,
      "grad_norm": 0.0,
      "learning_rate": 2.5357557015848473e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10201
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.9375,
      "epoch": 0.49294549671434096,
      "grad_norm": 0.0,
      "learning_rate": 2.5355141090065714e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10202
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.46875,
      "epoch": 0.49299381522999614,
      "grad_norm": 0.0,
      "learning_rate": 2.535272516428295e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10203
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.5625,
      "epoch": 0.4930421337456513,
      "grad_norm": 0.0,
      "learning_rate": 2.5350309238500197e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10204
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.21875,
      "epoch": 0.49309045226130654,
      "grad_norm": 0.0,
      "learning_rate": 2.5347893312717433e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10205
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 107.8125,
      "epoch": 0.4931387707769617,
      "grad_norm": 0.0,
      "learning_rate": 2.534547738693467e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10206
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.4375,
      "epoch": 0.49318708929261695,
      "grad_norm": 0.0,
      "learning_rate": 2.534306146115191e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10207
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.96875,
      "epoch": 0.49323540780827213,
      "grad_norm": 0.0,
      "learning_rate": 2.534064553536915e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10208
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.84375,
      "epoch": 0.4932837263239273,
      "grad_norm": 0.0,
      "learning_rate": 2.5338229609586393e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10209
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.4375,
      "epoch": 0.49333204483958254,
      "grad_norm": 0.0,
      "learning_rate": 2.533581368380363e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10210
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.59375,
      "epoch": 0.4933803633552377,
      "grad_norm": 0.0,
      "learning_rate": 2.5333397758020876e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10211
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.78125,
      "epoch": 0.49342868187089295,
      "grad_norm": 0.0,
      "learning_rate": 2.533098183223811e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10212
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.1875,
      "epoch": 0.4934770003865481,
      "grad_norm": 0.0,
      "learning_rate": 2.5328565906455354e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10213
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.5,
      "epoch": 0.4935253189022033,
      "grad_norm": 0.0,
      "learning_rate": 2.5326149980672595e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10214
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.84375,
      "epoch": 0.49357363741785854,
      "grad_norm": 0.0,
      "learning_rate": 2.5323734054889836e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10215
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.53125,
      "epoch": 0.4936219559335137,
      "grad_norm": 0.0,
      "learning_rate": 2.532131812910707e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10216
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.96875,
      "epoch": 0.49367027444916894,
      "grad_norm": 0.0,
      "learning_rate": 2.531890220332431e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10217
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.1875,
      "epoch": 0.4937185929648241,
      "grad_norm": 0.0,
      "learning_rate": 2.5316486277541555e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10218
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.21875,
      "epoch": 0.4937669114804793,
      "grad_norm": 0.0,
      "learning_rate": 2.531407035175879e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10219
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 106.5625,
      "epoch": 0.49381522999613453,
      "grad_norm": 0.0,
      "learning_rate": 2.531165442597603e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 10220
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.78125,
      "epoch": 0.4938635485117897,
      "grad_norm": 0.0,
      "learning_rate": 2.5309238500193274e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10221
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.90625,
      "epoch": 0.49391186702744494,
      "grad_norm": 0.0,
      "learning_rate": 2.5306822574410515e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10222
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.03125,
      "epoch": 0.4939601855431001,
      "grad_norm": 0.0,
      "learning_rate": 2.530440664862775e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10223
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.15625,
      "epoch": 0.4940085040587553,
      "grad_norm": 0.0,
      "learning_rate": 2.5301990722845e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10224
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.375,
      "epoch": 0.4940568225744105,
      "grad_norm": 0.0,
      "learning_rate": 2.5299574797062234e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10225
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.65625,
      "epoch": 0.4941051410900657,
      "grad_norm": 0.0,
      "learning_rate": 2.529715887127947e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10226
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.90625,
      "epoch": 0.49415345960572093,
      "grad_norm": 0.0,
      "learning_rate": 2.529474294549671e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10227
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.65625,
      "epoch": 0.4942017781213761,
      "grad_norm": 0.0,
      "learning_rate": 2.5292327019713953e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10228
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.6875,
      "epoch": 0.4942500966370313,
      "grad_norm": 0.0,
      "learning_rate": 2.5289911093931194e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10229
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.0625,
      "epoch": 0.4942984151526865,
      "grad_norm": 0.0,
      "learning_rate": 2.528749516814843e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10230
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.09375,
      "epoch": 0.4943467336683417,
      "grad_norm": 0.0,
      "learning_rate": 2.5285079242365677e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10231
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.03125,
      "epoch": 0.49439505218399693,
      "grad_norm": 0.0,
      "learning_rate": 2.5282663316582913e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10232
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 93.875,
      "epoch": 0.4944433706996521,
      "grad_norm": 0.0,
      "learning_rate": 2.5280247390800154e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 10233
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.4375,
      "epoch": 0.4944916892153073,
      "grad_norm": 0.0,
      "learning_rate": 2.5277831465017396e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10234
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.875,
      "epoch": 0.4945400077309625,
      "grad_norm": 0.0,
      "learning_rate": 2.5275415539234637e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10235
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.78125,
      "epoch": 0.4945883262466177,
      "grad_norm": 0.0,
      "learning_rate": 2.5272999613451873e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10236
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.71875,
      "epoch": 0.4946366447622729,
      "grad_norm": 0.0,
      "learning_rate": 2.527058368766911e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10237
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.0625,
      "epoch": 0.4946849632779281,
      "grad_norm": 0.0,
      "learning_rate": 2.5268167761886356e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10238
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 103.625,
      "epoch": 0.4947332817935833,
      "grad_norm": 0.0,
      "learning_rate": 2.526575183610359e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10239
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 102.28125,
      "epoch": 0.4947816003092385,
      "grad_norm": 0.0,
      "learning_rate": 2.5263335910320833e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10240
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.53125,
      "epoch": 0.4948299188248937,
      "grad_norm": 0.0,
      "learning_rate": 2.5260919984538075e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10241
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.28125,
      "epoch": 0.4948782373405489,
      "grad_norm": 0.0,
      "learning_rate": 2.5258504058755316e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10242
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.125,
      "epoch": 0.4949265558562041,
      "grad_norm": 0.0,
      "learning_rate": 2.525608813297255e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10243
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.21875,
      "epoch": 0.4949748743718593,
      "grad_norm": 0.0,
      "learning_rate": 2.52536722071898e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10244
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 95.5625,
      "epoch": 0.4950231928875145,
      "grad_norm": 19.695443839655603,
      "learning_rate": 2.5251256281407035e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10245
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.375,
      "epoch": 0.4950715114031697,
      "grad_norm": 0.0,
      "learning_rate": 2.5248840355624276e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10246
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.625,
      "epoch": 0.4951198299188249,
      "grad_norm": 0.0,
      "learning_rate": 2.524642442984151e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10247
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.90625,
      "epoch": 0.4951681484344801,
      "grad_norm": 0.0,
      "learning_rate": 2.5244008504058754e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10248
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.4375,
      "epoch": 0.49521646695013527,
      "grad_norm": 0.0,
      "learning_rate": 2.5241592578275995e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10249
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.625,
      "epoch": 0.4952647854657905,
      "grad_norm": 0.0,
      "learning_rate": 2.523917665249323e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10250
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.25,
      "epoch": 0.4953131039814457,
      "grad_norm": 0.0,
      "learning_rate": 2.523676072671048e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10251
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.375,
      "epoch": 0.4953614224971009,
      "grad_norm": 0.0,
      "learning_rate": 2.5234344800927714e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10252
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.8125,
      "epoch": 0.4954097410127561,
      "grad_norm": 0.0,
      "learning_rate": 2.5231928875144955e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10253
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.96875,
      "epoch": 0.49545805952841127,
      "grad_norm": 0.0,
      "learning_rate": 2.5229512949362196e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10254
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.8125,
      "epoch": 0.4955063780440665,
      "grad_norm": 0.0,
      "learning_rate": 2.522709702357944e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10255
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.8125,
      "epoch": 0.4955546965597217,
      "grad_norm": 0.0,
      "learning_rate": 2.5224681097796674e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10256
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.4375,
      "epoch": 0.4956030150753769,
      "grad_norm": 0.0,
      "learning_rate": 2.522226517201391e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10257
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.875,
      "epoch": 0.4956513335910321,
      "grad_norm": 0.0,
      "learning_rate": 2.5219849246231157e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10258
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.625,
      "epoch": 0.49569965210668726,
      "grad_norm": 0.0,
      "learning_rate": 2.5217433320448393e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10259
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.84375,
      "epoch": 0.4957479706223425,
      "grad_norm": 0.0,
      "learning_rate": 2.5215017394665634e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10260
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.34375,
      "epoch": 0.49579628913799767,
      "grad_norm": 0.0,
      "learning_rate": 2.5212601468882875e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10261
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.9375,
      "epoch": 0.4958446076536529,
      "grad_norm": 0.0,
      "learning_rate": 2.5210185543100117e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10262
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.25,
      "epoch": 0.4958929261693081,
      "grad_norm": 0.0,
      "learning_rate": 2.5207769617317353e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10263
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.59375,
      "epoch": 0.49594124468496326,
      "grad_norm": 0.0,
      "learning_rate": 2.52053536915346e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10264
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.9375,
      "epoch": 0.4959895632006185,
      "grad_norm": 0.0,
      "learning_rate": 2.5202937765751836e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10265
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.96875,
      "epoch": 0.49603788171627367,
      "grad_norm": 0.0,
      "learning_rate": 2.5200521839969077e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10266
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.875,
      "epoch": 0.4960862002319289,
      "grad_norm": 0.0,
      "learning_rate": 2.5198105914186313e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10267
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.53125,
      "epoch": 0.4961345187475841,
      "grad_norm": 0.0,
      "learning_rate": 2.5195689988403554e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10268
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.09375,
      "epoch": 0.49618283726323925,
      "grad_norm": 0.0,
      "learning_rate": 2.5193274062620796e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10269
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.46875,
      "epoch": 0.4962311557788945,
      "grad_norm": 0.0,
      "learning_rate": 2.519085813683803e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10270
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.3125,
      "epoch": 0.49627947429454966,
      "grad_norm": 0.0,
      "learning_rate": 2.518844221105528e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10271
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.15625,
      "epoch": 0.4963277928102049,
      "grad_norm": 0.0,
      "learning_rate": 2.5186026285272515e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10272
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.25,
      "epoch": 0.49637611132586007,
      "grad_norm": 0.0,
      "learning_rate": 2.5183610359489756e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10273
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.375,
      "epoch": 0.49642442984151525,
      "grad_norm": 0.0,
      "learning_rate": 2.5181194433706997e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10274
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.6875,
      "epoch": 0.4964727483571705,
      "grad_norm": 0.0,
      "learning_rate": 2.517877850792424e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10275
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.25,
      "epoch": 0.49652106687282566,
      "grad_norm": 0.0,
      "learning_rate": 2.5176362582141475e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10276
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.9375,
      "epoch": 0.4965693853884809,
      "grad_norm": 0.0,
      "learning_rate": 2.517394665635871e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10277
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.03125,
      "epoch": 0.49661770390413607,
      "grad_norm": 0.0,
      "learning_rate": 2.517153073057596e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10278
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.46875,
      "epoch": 0.49666602241979124,
      "grad_norm": 0.0,
      "learning_rate": 2.5169114804793193e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880350112915,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10279
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.0,
      "epoch": 0.4967143409354465,
      "grad_norm": 0.0,
      "learning_rate": 2.5166698879010435e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10280
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.0625,
      "epoch": 0.49676265945110165,
      "grad_norm": 0.0,
      "learning_rate": 2.5164282953227676e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10281
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.34375,
      "epoch": 0.4968109779667569,
      "grad_norm": 0.0,
      "learning_rate": 2.516186702744492e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10282
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 96.375,
      "epoch": 0.49685929648241206,
      "grad_norm": 0.0,
      "learning_rate": 2.5159451101662154e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 10283
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.28125,
      "epoch": 0.49690761499806724,
      "grad_norm": 0.0,
      "learning_rate": 2.51570351758794e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10284
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.875,
      "epoch": 0.49695593351372247,
      "grad_norm": 0.0,
      "learning_rate": 2.5154619250096636e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10285
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.5,
      "epoch": 0.49700425202937765,
      "grad_norm": 0.0,
      "learning_rate": 2.515220332431388e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10286
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.40625,
      "epoch": 0.4970525705450329,
      "grad_norm": 0.0,
      "learning_rate": 2.5149787398531114e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10287
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.0625,
      "epoch": 0.49710088906068806,
      "grad_norm": 0.0,
      "learning_rate": 2.5147371472748355e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10288
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.0625,
      "epoch": 0.49714920757634323,
      "grad_norm": 0.0,
      "learning_rate": 2.5144955546965597e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10289
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.9375,
      "epoch": 0.49719752609199847,
      "grad_norm": 0.0,
      "learning_rate": 2.514253962118283e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10290
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.15625,
      "epoch": 0.49724584460765364,
      "grad_norm": 0.0,
      "learning_rate": 2.514012369540008e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10291
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.34375,
      "epoch": 0.4972941631233089,
      "grad_norm": 0.0,
      "learning_rate": 2.5137707769617315e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10292
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.9375,
      "epoch": 0.49734248163896405,
      "grad_norm": 0.0,
      "learning_rate": 2.5135291843834557e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10293
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.96875,
      "epoch": 0.49739080015461923,
      "grad_norm": 0.0,
      "learning_rate": 2.51328759180518e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10294
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.0625,
      "epoch": 0.49743911867027446,
      "grad_norm": 0.0,
      "learning_rate": 2.513045999226904e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10295
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.8125,
      "epoch": 0.49748743718592964,
      "grad_norm": 0.0,
      "learning_rate": 2.5128044066486275e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 10296
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.34375,
      "epoch": 0.49753575570158487,
      "grad_norm": 0.0,
      "learning_rate": 2.5125628140703517e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10297
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.4375,
      "epoch": 0.49758407421724005,
      "grad_norm": 0.0,
      "learning_rate": 2.512321221492076e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10298
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.9375,
      "epoch": 0.4976323927328952,
      "grad_norm": 0.0,
      "learning_rate": 2.5120796289137994e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10299
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.8125,
      "epoch": 0.49768071124855046,
      "grad_norm": 0.0,
      "learning_rate": 2.5118380363355236e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10300
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.03125,
      "epoch": 0.49772902976420563,
      "grad_norm": 0.0,
      "learning_rate": 2.5115964437572477e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10301
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.5625,
      "epoch": 0.49777734827986087,
      "grad_norm": 0.0,
      "learning_rate": 2.511354851178972e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10302
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.9375,
      "epoch": 0.49782566679551604,
      "grad_norm": 0.0,
      "learning_rate": 2.5111132586006954e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10303
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.90625,
      "epoch": 0.4978739853111712,
      "grad_norm": 0.0,
      "learning_rate": 2.51087166602242e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10304
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.375,
      "epoch": 0.49792230382682645,
      "grad_norm": 0.0,
      "learning_rate": 2.5106300734441437e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10305
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.8125,
      "epoch": 0.49797062234248163,
      "grad_norm": 0.0,
      "learning_rate": 2.510388480865868e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10306
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.65625,
      "epoch": 0.49801894085813686,
      "grad_norm": 0.0,
      "learning_rate": 2.5101468882875915e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10307
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.03125,
      "epoch": 0.49806725937379204,
      "grad_norm": 0.0,
      "learning_rate": 2.509905295709316e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10308
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.46875,
      "epoch": 0.4981155778894472,
      "grad_norm": 0.0,
      "learning_rate": 2.5096637031310397e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 10309
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 100.40625,
      "epoch": 0.49816389640510245,
      "grad_norm": 42.763926486975635,
      "learning_rate": 2.5094221105527633e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10310
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.8125,
      "epoch": 0.4982122149207576,
      "grad_norm": 0.0,
      "learning_rate": 2.509180517974488e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10311
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.875,
      "epoch": 0.49826053343641286,
      "grad_norm": 0.0,
      "learning_rate": 2.5089389253962116e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10312
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.625,
      "epoch": 0.49830885195206803,
      "grad_norm": 0.0,
      "learning_rate": 2.508697332817936e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10313
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.1875,
      "epoch": 0.4983571704677232,
      "grad_norm": 0.0,
      "learning_rate": 2.50845574023966e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10314
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.53125,
      "epoch": 0.49840548898337844,
      "grad_norm": 0.0,
      "learning_rate": 2.508214147661384e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10315
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.21875,
      "epoch": 0.4984538074990336,
      "grad_norm": 0.0,
      "learning_rate": 2.5079725550831076e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10316
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.84375,
      "epoch": 0.49850212601468885,
      "grad_norm": 0.0,
      "learning_rate": 2.507730962504832e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10317
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.0,
      "epoch": 0.49855044453034403,
      "grad_norm": 0.0,
      "learning_rate": 2.507489369926556e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10318
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 100.75,
      "epoch": 0.4985987630459992,
      "grad_norm": 0.0,
      "learning_rate": 2.5072477773482795e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10319
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.15625,
      "epoch": 0.49864708156165444,
      "grad_norm": 0.0,
      "learning_rate": 2.5070061847700036e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10320
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.75,
      "epoch": 0.4986954000773096,
      "grad_norm": 0.0,
      "learning_rate": 2.506764592191728e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10321
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.96875,
      "epoch": 0.49874371859296485,
      "grad_norm": 0.0,
      "learning_rate": 2.506522999613452e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10322
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.0625,
      "epoch": 0.49879203710862,
      "grad_norm": 0.0,
      "learning_rate": 2.5062814070351755e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10323
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.25,
      "epoch": 0.4988403556242752,
      "grad_norm": 0.0,
      "learning_rate": 2.5060398144569e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10324
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.96875,
      "epoch": 0.49888867413993043,
      "grad_norm": 0.0,
      "learning_rate": 2.505798221878624e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10325
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.5,
      "epoch": 0.4989369926555856,
      "grad_norm": 0.0,
      "learning_rate": 2.505556629300348e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10326
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.15625,
      "epoch": 0.49898531117124084,
      "grad_norm": 0.0,
      "learning_rate": 2.5053150367220715e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10327
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.71875,
      "epoch": 0.499033629686896,
      "grad_norm": 0.0,
      "learning_rate": 2.505073444143796e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10328
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.375,
      "epoch": 0.4990819482025512,
      "grad_norm": 0.0,
      "learning_rate": 2.50483185156552e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10329
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.53125,
      "epoch": 0.49913026671820643,
      "grad_norm": 0.0,
      "learning_rate": 2.5045902589872434e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10330
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.53125,
      "epoch": 0.4991785852338616,
      "grad_norm": 0.0,
      "learning_rate": 2.504348666408968e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10331
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.3125,
      "epoch": 0.49922690374951684,
      "grad_norm": 0.0,
      "learning_rate": 2.5041070738306917e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10332
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.40625,
      "epoch": 0.499275222265172,
      "grad_norm": 0.0,
      "learning_rate": 2.503865481252416e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10333
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.25,
      "epoch": 0.4993235407808272,
      "grad_norm": 0.0,
      "learning_rate": 2.50362388867414e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10334
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.1875,
      "epoch": 0.4993718592964824,
      "grad_norm": 0.0,
      "learning_rate": 2.503382296095864e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10335
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.03125,
      "epoch": 0.4994201778121376,
      "grad_norm": 0.0,
      "learning_rate": 2.5031407035175877e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10336
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.5,
      "epoch": 0.49946849632779283,
      "grad_norm": 0.0,
      "learning_rate": 2.502899110939312e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10337
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.09375,
      "epoch": 0.499516814843448,
      "grad_norm": 0.0,
      "learning_rate": 2.502657518361036e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10338
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.84375,
      "epoch": 0.4995651333591032,
      "grad_norm": 0.0,
      "learning_rate": 2.5024159257827596e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10339
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.53125,
      "epoch": 0.4996134518747584,
      "grad_norm": 0.0,
      "learning_rate": 2.5021743332044837e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10340
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.78125,
      "epoch": 0.4996617703904136,
      "grad_norm": 0.0,
      "learning_rate": 2.501932740626208e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 10341
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.875,
      "epoch": 0.49971008890606883,
      "grad_norm": 0.0,
      "learning_rate": 2.501691148047932e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10342
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.53125,
      "epoch": 0.499758407421724,
      "grad_norm": 0.0,
      "learning_rate": 2.5014495554696556e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10343
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.90625,
      "epoch": 0.4998067259373792,
      "grad_norm": 0.0,
      "learning_rate": 2.50120796289138e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10344
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.3125,
      "epoch": 0.4998550444530344,
      "grad_norm": 0.0,
      "learning_rate": 2.500966370313104e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10345
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.34375,
      "epoch": 0.4999033629686896,
      "grad_norm": 0.0,
      "learning_rate": 2.500724777734828e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10346
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.1875,
      "epoch": 0.4999516814843448,
      "grad_norm": 0.0,
      "learning_rate": 2.5004831851565516e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10347
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.59375,
      "epoch": 0.5,
      "grad_norm": 0.0,
      "learning_rate": 2.5002415925782763e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10348
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.84375,
      "epoch": 0.5000483185156552,
      "grad_norm": 0.0,
      "learning_rate": 2.5e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10349
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.1875,
      "epoch": 0.5000966370313104,
      "grad_norm": 0.0,
      "learning_rate": 2.499758407421724e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10350
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.78125,
      "epoch": 0.5001449555469656,
      "grad_norm": 0.0,
      "learning_rate": 2.4995168148434476e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10351
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.09375,
      "epoch": 0.5001932740626208,
      "grad_norm": 0.0,
      "learning_rate": 2.499275222265172e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10352
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.5625,
      "epoch": 0.500241592578276,
      "grad_norm": 0.0,
      "learning_rate": 2.499033629686896e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10353
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.625,
      "epoch": 0.5002899110939312,
      "grad_norm": 0.0,
      "learning_rate": 2.49879203710862e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10354
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.875,
      "epoch": 0.5003382296095864,
      "grad_norm": 0.0,
      "learning_rate": 2.498550444530344e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10355
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.84375,
      "epoch": 0.5003865481252416,
      "grad_norm": 0.0,
      "learning_rate": 2.498308851952068e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10356
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.1875,
      "epoch": 0.5004348666408968,
      "grad_norm": 0.0,
      "learning_rate": 2.498067259373792e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10357
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.75,
      "epoch": 0.500483185156552,
      "grad_norm": 0.0,
      "learning_rate": 2.497825666795516e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10358
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.875,
      "epoch": 0.5005315036722072,
      "grad_norm": 0.0,
      "learning_rate": 2.49758407421724e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10359
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.0,
      "epoch": 0.5005798221878623,
      "grad_norm": 0.0,
      "learning_rate": 2.497342481638964e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10360
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.8125,
      "epoch": 0.5006281407035176,
      "grad_norm": 0.0,
      "learning_rate": 2.497100889060688e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10361
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.84375,
      "epoch": 0.5006764592191728,
      "grad_norm": 0.0,
      "learning_rate": 2.496859296482412e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10362
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.53125,
      "epoch": 0.500724777734828,
      "grad_norm": 0.0,
      "learning_rate": 2.4966177039041357e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10363
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.21875,
      "epoch": 0.5007730962504832,
      "grad_norm": 0.0,
      "learning_rate": 2.49637611132586e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10364
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.0625,
      "epoch": 0.5008214147661384,
      "grad_norm": 0.0,
      "learning_rate": 2.496134518747584e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10365
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.78125,
      "epoch": 0.5008697332817936,
      "grad_norm": 0.0,
      "learning_rate": 2.495892926169308e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10366
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.03125,
      "epoch": 0.5009180517974487,
      "grad_norm": 0.0,
      "learning_rate": 2.495651333591032e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10367
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.4375,
      "epoch": 0.500966370313104,
      "grad_norm": 0.0,
      "learning_rate": 2.495409741012756e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10368
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.59375,
      "epoch": 0.5010146888287592,
      "grad_norm": 0.0,
      "learning_rate": 2.49516814843448e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10369
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.78125,
      "epoch": 0.5010630073444143,
      "grad_norm": 0.0,
      "learning_rate": 2.494926555856204e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10370
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.90625,
      "epoch": 0.5011113258600696,
      "grad_norm": 38.107311595903376,
      "learning_rate": 2.4946849632779277e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.21316051483154297,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10371
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.53125,
      "epoch": 0.5011596443757248,
      "grad_norm": 0.0,
      "learning_rate": 2.494443370699652e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10372
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 109.21875,
      "epoch": 0.50120796289138,
      "grad_norm": 0.0,
      "learning_rate": 2.494201778121376e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 10373
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.5,
      "epoch": 0.5012562814070352,
      "grad_norm": 0.0,
      "learning_rate": 2.4939601855431e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10374
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.96875,
      "epoch": 0.5013045999226904,
      "grad_norm": 0.0,
      "learning_rate": 2.493718592964824e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10375
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.3125,
      "epoch": 0.5013529184383456,
      "grad_norm": 0.0,
      "learning_rate": 2.493477000386548e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10376
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.5,
      "epoch": 0.5014012369540007,
      "grad_norm": 0.0,
      "learning_rate": 2.493235407808272e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10377
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.0625,
      "epoch": 0.501449555469656,
      "grad_norm": 0.0,
      "learning_rate": 2.492993815229996e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10378
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.96875,
      "epoch": 0.5014978739853112,
      "grad_norm": 0.0,
      "learning_rate": 2.4927522226517203e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10379
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.21875,
      "epoch": 0.5015461925009663,
      "grad_norm": 0.0,
      "learning_rate": 2.4925106300734444e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10380
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.21875,
      "epoch": 0.5015945110166216,
      "grad_norm": 0.0,
      "learning_rate": 2.492269037495168e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10381
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.34375,
      "epoch": 0.5016428295322768,
      "grad_norm": 0.0,
      "learning_rate": 2.492027444916892e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10382
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.3125,
      "epoch": 0.501691148047932,
      "grad_norm": 0.0,
      "learning_rate": 2.491785852338616e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10383
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.65625,
      "epoch": 0.5017394665635871,
      "grad_norm": 0.0,
      "learning_rate": 2.49154425976034e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10384
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.25,
      "epoch": 0.5017877850792424,
      "grad_norm": 0.0,
      "learning_rate": 2.491302667182064e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10385
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.53125,
      "epoch": 0.5018361035948976,
      "grad_norm": 0.0,
      "learning_rate": 2.491061074603788e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10386
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.3125,
      "epoch": 0.5018844221105527,
      "grad_norm": 0.0,
      "learning_rate": 2.4908194820255123e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10387
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.875,
      "epoch": 0.501932740626208,
      "grad_norm": 0.0,
      "learning_rate": 2.490577889447236e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10388
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.65625,
      "epoch": 0.5019810591418632,
      "grad_norm": 0.0,
      "learning_rate": 2.49033629686896e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10389
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.75,
      "epoch": 0.5020293776575183,
      "grad_norm": 0.0,
      "learning_rate": 2.490094704290684e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10390
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.78125,
      "epoch": 0.5020776961731735,
      "grad_norm": 0.0,
      "learning_rate": 2.489853111712408e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10391
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.46875,
      "epoch": 0.5021260146888288,
      "grad_norm": 0.0,
      "learning_rate": 2.489611519134132e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10392
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.4375,
      "epoch": 0.502174333204484,
      "grad_norm": 0.0,
      "learning_rate": 2.489369926555856e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10393
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.125,
      "epoch": 0.5022226517201391,
      "grad_norm": 0.0,
      "learning_rate": 2.48912833397758e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10394
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.625,
      "epoch": 0.5022709702357944,
      "grad_norm": 0.0,
      "learning_rate": 2.4888867413993043e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10395
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.6875,
      "epoch": 0.5023192887514496,
      "grad_norm": 0.0,
      "learning_rate": 2.488645148821028e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10396
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.875,
      "epoch": 0.5023676072671047,
      "grad_norm": 0.0,
      "learning_rate": 2.488403556242752e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10397
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.5625,
      "epoch": 0.50241592578276,
      "grad_norm": 0.0,
      "learning_rate": 2.488161963664476e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10398
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.8125,
      "epoch": 0.5024642442984152,
      "grad_norm": 0.0,
      "learning_rate": 2.4879203710862003e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.260896235704422,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10399
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.21875,
      "epoch": 0.5025125628140703,
      "grad_norm": 0.0,
      "learning_rate": 2.4876787785079245e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10400
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.5625,
      "epoch": 0.5025608813297255,
      "grad_norm": 0.0,
      "learning_rate": 2.487437185929648e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.249358132481575,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10401
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.125,
      "epoch": 0.5026091998453808,
      "grad_norm": 0.0,
      "learning_rate": 2.487195593351372e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10402
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.1875,
      "epoch": 0.502657518361036,
      "grad_norm": 0.0,
      "learning_rate": 2.486954000773096e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10403
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 85.75,
      "epoch": 0.5027058368766911,
      "grad_norm": 7.77955253582964,
      "learning_rate": 2.48671240819482e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10404
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 98.5625,
      "epoch": 0.5027541553923464,
      "grad_norm": 0.0,
      "learning_rate": 2.486470815616544e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 10405
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.78125,
      "epoch": 0.5028024739080016,
      "grad_norm": 0.0,
      "learning_rate": 2.486229223038268e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10406
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.75,
      "epoch": 0.5028507924236567,
      "grad_norm": 0.0,
      "learning_rate": 2.4859876304599924e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10407
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.6875,
      "epoch": 0.5028991109393119,
      "grad_norm": 0.0,
      "learning_rate": 2.485746037881716e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10408
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.65625,
      "epoch": 0.5029474294549672,
      "grad_norm": 0.0,
      "learning_rate": 2.48550444530344e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10409
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.3125,
      "epoch": 0.5029957479706223,
      "grad_norm": 0.0,
      "learning_rate": 2.485262852725164e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10410
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 97.5,
      "epoch": 0.5030440664862775,
      "grad_norm": 0.0,
      "learning_rate": 2.485021260146888e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 10411
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.09375,
      "epoch": 0.5030923850019328,
      "grad_norm": 0.0,
      "learning_rate": 2.484779667568612e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10412
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.09375,
      "epoch": 0.503140703517588,
      "grad_norm": 0.0,
      "learning_rate": 2.484538074990336e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10413
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.71875,
      "epoch": 0.5031890220332431,
      "grad_norm": 0.0,
      "learning_rate": 2.4842964824120603e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10414
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.53125,
      "epoch": 0.5032373405488983,
      "grad_norm": 0.0,
      "learning_rate": 2.4840548898337844e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10415
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.40625,
      "epoch": 0.5032856590645536,
      "grad_norm": 0.0,
      "learning_rate": 2.483813297255508e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10416
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.84375,
      "epoch": 0.5033339775802087,
      "grad_norm": 0.0,
      "learning_rate": 2.483571704677232e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10417
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 105.9375,
      "epoch": 0.5033822960958639,
      "grad_norm": 0.0,
      "learning_rate": 2.4833301120989563e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10418
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.90625,
      "epoch": 0.5034306146115192,
      "grad_norm": 0.0,
      "learning_rate": 2.4830885195206804e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10419
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.125,
      "epoch": 0.5034789331271743,
      "grad_norm": 0.0,
      "learning_rate": 2.4828469269424046e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 10420
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.625,
      "epoch": 0.5035272516428295,
      "grad_norm": 0.0,
      "learning_rate": 2.482605334364128e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10421
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 106.28125,
      "epoch": 0.5035755701584848,
      "grad_norm": 0.0,
      "learning_rate": 2.4823637417858523e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 10422
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.59375,
      "epoch": 0.50362388867414,
      "grad_norm": 0.0,
      "learning_rate": 2.482122149207576e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10423
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 104.90625,
      "epoch": 0.5036722071897951,
      "grad_norm": 12.93276622982168,
      "learning_rate": 2.4818805566293e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.23685522377490997,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10424
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.46875,
      "epoch": 0.5037205257054503,
      "grad_norm": 0.0,
      "learning_rate": 2.481638964051024e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10425
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.1875,
      "epoch": 0.5037688442211056,
      "grad_norm": 0.0,
      "learning_rate": 2.4813973714727483e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10426
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.25,
      "epoch": 0.5038171627367607,
      "grad_norm": 0.0,
      "learning_rate": 2.4811557788944725e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10427
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.75,
      "epoch": 0.5038654812524159,
      "grad_norm": 0.0,
      "learning_rate": 2.480914186316196e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10428
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.375,
      "epoch": 0.5039137997680712,
      "grad_norm": 0.0,
      "learning_rate": 2.48067259373792e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10429
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.84375,
      "epoch": 0.5039621182837263,
      "grad_norm": 0.0,
      "learning_rate": 2.4804310011596443e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10430
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.09375,
      "epoch": 0.5040104367993815,
      "grad_norm": 0.0,
      "learning_rate": 2.4801894085813685e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10431
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.65625,
      "epoch": 0.5040587553150367,
      "grad_norm": 0.0,
      "learning_rate": 2.479947816003092e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10432
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.5,
      "epoch": 0.504107073830692,
      "grad_norm": 0.0,
      "learning_rate": 2.479706223424816e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10433
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 97.375,
      "epoch": 0.5041553923463471,
      "grad_norm": 17.139216629603524,
      "learning_rate": 2.4794646308465404e-07,
      "loss": -0.0,
      "reward": 2.40625,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10434
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 101.0,
      "epoch": 0.5042037108620023,
      "grad_norm": 0.0,
      "learning_rate": 2.4792230382682645e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10435
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.125,
      "epoch": 0.5042520293776576,
      "grad_norm": 0.0,
      "learning_rate": 2.478981445689988e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10436
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.625,
      "epoch": 0.5043003478933127,
      "grad_norm": 0.0,
      "learning_rate": 2.478739853111712e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10437
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.5625,
      "epoch": 0.5043486664089679,
      "grad_norm": 0.0,
      "learning_rate": 2.4784982605334364e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10438
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.03125,
      "epoch": 0.5043969849246231,
      "grad_norm": 0.0,
      "learning_rate": 2.4782566679551605e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10439
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.6875,
      "epoch": 0.5044453034402783,
      "grad_norm": 0.0,
      "learning_rate": 2.4780150753768846e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10440
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.25,
      "epoch": 0.5044936219559335,
      "grad_norm": 0.0,
      "learning_rate": 2.477773482798608e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10441
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.9375,
      "epoch": 0.5045419404715887,
      "grad_norm": 0.0,
      "learning_rate": 2.4775318902203324e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10442
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.40625,
      "epoch": 0.504590258987244,
      "grad_norm": 0.0,
      "learning_rate": 2.477290297642056e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10443
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.9375,
      "epoch": 0.5046385775028991,
      "grad_norm": 0.0,
      "learning_rate": 2.47704870506378e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10444
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.0,
      "epoch": 0.5046868960185543,
      "grad_norm": 0.0,
      "learning_rate": 2.476807112485504e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10445
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.46875,
      "epoch": 0.5047352145342096,
      "grad_norm": 0.0,
      "learning_rate": 2.4765655199072284e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10446
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.21875,
      "epoch": 0.5047835330498647,
      "grad_norm": 0.0,
      "learning_rate": 2.4763239273289525e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10447
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.71875,
      "epoch": 0.5048318515655199,
      "grad_norm": 0.0,
      "learning_rate": 2.476082334750676e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 10448
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.78125,
      "epoch": 0.5048801700811751,
      "grad_norm": 0.0,
      "learning_rate": 2.4758407421724003e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10449
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.59375,
      "epoch": 0.5049284885968303,
      "grad_norm": 0.0,
      "learning_rate": 2.4755991495941244e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10450
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.875,
      "epoch": 0.5049768071124855,
      "grad_norm": 0.0,
      "learning_rate": 2.4753575570158486e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10451
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 101.71875,
      "epoch": 0.5050251256281407,
      "grad_norm": 49.947683328742784,
      "learning_rate": 2.475115964437572e-07,
      "loss": 0.0,
      "reward": 1.5625,
      "reward_std": 0.22049658000469208,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10452
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 118.71875,
      "epoch": 0.505073444143796,
      "grad_norm": 0.0,
      "learning_rate": 2.4748743718592963e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10453
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.3125,
      "epoch": 0.5051217626594511,
      "grad_norm": 0.0,
      "learning_rate": 2.4746327792810204e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10454
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.25,
      "epoch": 0.5051700811751063,
      "grad_norm": 0.0,
      "learning_rate": 2.4743911867027446e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10455
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.71875,
      "epoch": 0.5052183996907615,
      "grad_norm": 0.0,
      "learning_rate": 2.474149594124468e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10456
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.65625,
      "epoch": 0.5052667182064167,
      "grad_norm": 0.0,
      "learning_rate": 2.4739080015461923e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10457
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.71875,
      "epoch": 0.5053150367220719,
      "grad_norm": 0.0,
      "learning_rate": 2.4736664089679164e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10458
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.96875,
      "epoch": 0.5053633552377271,
      "grad_norm": 0.0,
      "learning_rate": 2.4734248163896406e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10459
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.375,
      "epoch": 0.5054116737533823,
      "grad_norm": 0.0,
      "learning_rate": 2.4731832238113647e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 10460
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.90625,
      "epoch": 0.5054599922690375,
      "grad_norm": 0.0,
      "learning_rate": 2.4729416312330883e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10461
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.875,
      "epoch": 0.5055083107846927,
      "grad_norm": 0.0,
      "learning_rate": 2.4727000386548125e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10462
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.125,
      "epoch": 0.505556629300348,
      "grad_norm": 0.0,
      "learning_rate": 2.472458446076536e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10463
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.125,
      "epoch": 0.5056049478160031,
      "grad_norm": 0.0,
      "learning_rate": 2.47221685349826e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10464
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.75,
      "epoch": 0.5056532663316583,
      "grad_norm": 0.0,
      "learning_rate": 2.4719752609199843e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10465
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.25,
      "epoch": 0.5057015848473135,
      "grad_norm": 0.0,
      "learning_rate": 2.4717336683417085e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10466
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.25,
      "epoch": 0.5057499033629687,
      "grad_norm": 0.0,
      "learning_rate": 2.4714920757634326e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10467
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.0,
      "epoch": 0.5057982218786239,
      "grad_norm": 0.0,
      "learning_rate": 2.471250483185156e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10468
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.6875,
      "epoch": 0.5058465403942791,
      "grad_norm": 0.0,
      "learning_rate": 2.4710088906068804e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10469
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.71875,
      "epoch": 0.5058948589099342,
      "grad_norm": 0.0,
      "learning_rate": 2.4707672980286045e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10470
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.0,
      "epoch": 0.5059431774255895,
      "grad_norm": 0.0,
      "learning_rate": 2.4705257054503286e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10471
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.4375,
      "epoch": 0.5059914959412447,
      "grad_norm": 0.0,
      "learning_rate": 2.470284112872053e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10472
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.25,
      "epoch": 0.5060398144568999,
      "grad_norm": 0.0,
      "learning_rate": 2.4700425202937764e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10473
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.90625,
      "epoch": 0.5060881329725551,
      "grad_norm": 0.0,
      "learning_rate": 2.4698009277155005e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10474
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.59375,
      "epoch": 0.5061364514882103,
      "grad_norm": 0.0,
      "learning_rate": 2.4695593351372246e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10475
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.625,
      "epoch": 0.5061847700038655,
      "grad_norm": 0.0,
      "learning_rate": 2.469317742558948e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10476
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.46875,
      "epoch": 0.5062330885195206,
      "grad_norm": 0.0,
      "learning_rate": 2.4690761499806724e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10477
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.71875,
      "epoch": 0.5062814070351759,
      "grad_norm": 0.0,
      "learning_rate": 2.4688345574023965e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10478
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.71875,
      "epoch": 0.5063297255508311,
      "grad_norm": 0.0,
      "learning_rate": 2.4685929648241207e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10479
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.03125,
      "epoch": 0.5063780440664862,
      "grad_norm": 0.0,
      "learning_rate": 2.468351372245845e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10480
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.78125,
      "epoch": 0.5064263625821415,
      "grad_norm": 0.0,
      "learning_rate": 2.4681097796675684e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10481
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.625,
      "epoch": 0.5064746810977967,
      "grad_norm": 0.0,
      "learning_rate": 2.4678681870892925e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10482
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.15625,
      "epoch": 0.5065229996134519,
      "grad_norm": 0.0,
      "learning_rate": 2.467626594511016e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10483
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.25,
      "epoch": 0.506571318129107,
      "grad_norm": 0.0,
      "learning_rate": 2.4673850019327403e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 10484
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.3125,
      "epoch": 0.5066196366447623,
      "grad_norm": 0.0,
      "learning_rate": 2.4671434093544644e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10485
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 111.625,
      "epoch": 0.5066679551604175,
      "grad_norm": 13.077882969371275,
      "learning_rate": 2.4669018167761886e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10486
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.125,
      "epoch": 0.5067162736760726,
      "grad_norm": 0.0,
      "learning_rate": 2.4666602241979127e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10487
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.34375,
      "epoch": 0.5067645921917279,
      "grad_norm": 0.0,
      "learning_rate": 2.4664186316196363e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10488
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.96875,
      "epoch": 0.5068129107073831,
      "grad_norm": 0.0,
      "learning_rate": 2.4661770390413604e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10489
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.78125,
      "epoch": 0.5068612292230382,
      "grad_norm": 0.0,
      "learning_rate": 2.4659354464630846e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10490
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.25,
      "epoch": 0.5069095477386935,
      "grad_norm": 0.0,
      "learning_rate": 2.4656938538848087e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10491
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.78125,
      "epoch": 0.5069578662543487,
      "grad_norm": 0.0,
      "learning_rate": 2.465452261306533e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10492
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.21875,
      "epoch": 0.5070061847700039,
      "grad_norm": 0.0,
      "learning_rate": 2.4652106687282565e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10493
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.59375,
      "epoch": 0.507054503285659,
      "grad_norm": 0.0,
      "learning_rate": 2.4649690761499806e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10494
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.78125,
      "epoch": 0.5071028218013143,
      "grad_norm": 0.0,
      "learning_rate": 2.4647274835717047e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10495
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.875,
      "epoch": 0.5071511403169695,
      "grad_norm": 0.0,
      "learning_rate": 2.4644858909934283e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10496
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.75,
      "epoch": 0.5071994588326246,
      "grad_norm": 0.0,
      "learning_rate": 2.4642442984151525e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10497
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.65625,
      "epoch": 0.5072477773482799,
      "grad_norm": 0.0,
      "learning_rate": 2.4640027058368766e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10498
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.21875,
      "epoch": 0.5072960958639351,
      "grad_norm": 0.0,
      "learning_rate": 2.463761113258601e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10499
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.4375,
      "epoch": 0.5073444143795902,
      "grad_norm": 0.0,
      "learning_rate": 2.463519520680325e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10500
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.5625,
      "epoch": 0.5073927328952454,
      "grad_norm": 0.0,
      "learning_rate": 2.4632779281020485e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10501
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.75,
      "epoch": 0.5074410514109007,
      "grad_norm": 0.0,
      "learning_rate": 2.4630363355237726e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10502
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 103.5,
      "epoch": 0.5074893699265559,
      "grad_norm": 0.0,
      "learning_rate": 2.462794742945497e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10503
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.5625,
      "epoch": 0.507537688442211,
      "grad_norm": 0.0,
      "learning_rate": 2.4625531503672204e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10504
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.59375,
      "epoch": 0.5075860069578663,
      "grad_norm": 0.0,
      "learning_rate": 2.4623115577889445e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10505
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 93.6875,
      "epoch": 0.5076343254735215,
      "grad_norm": 63.765290185788466,
      "learning_rate": 2.4620699652106686e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.3629639744758606,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10506
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.3125,
      "epoch": 0.5076826439891766,
      "grad_norm": 0.0,
      "learning_rate": 2.461828372632393e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10507
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.46875,
      "epoch": 0.5077309625048319,
      "grad_norm": 0.0,
      "learning_rate": 2.4615867800541164e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10508
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.40625,
      "epoch": 0.5077792810204871,
      "grad_norm": 0.0,
      "learning_rate": 2.4613451874758405e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10509
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.78125,
      "epoch": 0.5078275995361422,
      "grad_norm": 0.0,
      "learning_rate": 2.4611035948975647e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10510
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.8125,
      "epoch": 0.5078759180517974,
      "grad_norm": 0.0,
      "learning_rate": 2.460862002319289e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10511
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.90625,
      "epoch": 0.5079242365674527,
      "grad_norm": 0.0,
      "learning_rate": 2.460620409741013e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10512
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.0,
      "epoch": 0.5079725550831079,
      "grad_norm": 0.0,
      "learning_rate": 2.4603788171627365e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10513
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.5625,
      "epoch": 0.508020873598763,
      "grad_norm": 0.0,
      "learning_rate": 2.4601372245844607e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10514
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.4375,
      "epoch": 0.5080691921144183,
      "grad_norm": 0.0,
      "learning_rate": 2.459895632006185e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10515
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.84375,
      "epoch": 0.5081175106300735,
      "grad_norm": 0.0,
      "learning_rate": 2.4596540394279084e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10516
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.4375,
      "epoch": 0.5081658291457286,
      "grad_norm": 0.0,
      "learning_rate": 2.4594124468496325e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10517
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.3125,
      "epoch": 0.5082141476613838,
      "grad_norm": 0.0,
      "learning_rate": 2.4591708542713567e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10518
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.09375,
      "epoch": 0.5082624661770391,
      "grad_norm": 136.16675370030669,
      "learning_rate": 2.458929261693081e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10519
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.53125,
      "epoch": 0.5083107846926942,
      "grad_norm": 0.0,
      "learning_rate": 2.458687669114805e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10520
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.53125,
      "epoch": 0.5083591032083494,
      "grad_norm": 0.0,
      "learning_rate": 2.4584460765365286e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10521
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.65625,
      "epoch": 0.5084074217240047,
      "grad_norm": 0.0,
      "learning_rate": 2.4582044839582527e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10522
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.40625,
      "epoch": 0.5084557402396599,
      "grad_norm": 0.0,
      "learning_rate": 2.457962891379977e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10523
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.09375,
      "epoch": 0.508504058755315,
      "grad_norm": 0.0,
      "learning_rate": 2.4577212988017004e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10524
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.5625,
      "epoch": 0.5085523772709702,
      "grad_norm": 0.0,
      "learning_rate": 2.4574797062234246e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10525
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.25,
      "epoch": 0.5086006957866255,
      "grad_norm": 0.0,
      "learning_rate": 2.4572381136451487e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10526
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.15625,
      "epoch": 0.5086490143022806,
      "grad_norm": 0.0,
      "learning_rate": 2.456996521066873e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10527
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.53125,
      "epoch": 0.5086973328179358,
      "grad_norm": 0.0,
      "learning_rate": 2.4567549284885965e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10528
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.96875,
      "epoch": 0.5087456513335911,
      "grad_norm": 0.0,
      "learning_rate": 2.4565133359103206e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10529
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.40625,
      "epoch": 0.5087939698492462,
      "grad_norm": 0.0,
      "learning_rate": 2.4562717433320447e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10530
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.0,
      "epoch": 0.5088422883649014,
      "grad_norm": 0.0,
      "learning_rate": 2.456030150753769e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10531
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.0625,
      "epoch": 0.5088906068805567,
      "grad_norm": 0.0,
      "learning_rate": 2.455788558175493e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 10532
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.875,
      "epoch": 0.5089389253962119,
      "grad_norm": 0.0,
      "learning_rate": 2.4555469655972166e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10533
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.6875,
      "epoch": 0.508987243911867,
      "grad_norm": 0.0,
      "learning_rate": 2.455305373018941e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10534
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.3125,
      "epoch": 0.5090355624275222,
      "grad_norm": 0.0,
      "learning_rate": 2.455063780440665e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10535
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.59375,
      "epoch": 0.5090838809431775,
      "grad_norm": 0.0,
      "learning_rate": 2.4548221878623885e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10536
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.40625,
      "epoch": 0.5091321994588326,
      "grad_norm": 0.0,
      "learning_rate": 2.4545805952841126e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10537
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.71875,
      "epoch": 0.5091805179744878,
      "grad_norm": 0.0,
      "learning_rate": 2.454339002705837e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10538
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.84375,
      "epoch": 0.509228836490143,
      "grad_norm": 0.0,
      "learning_rate": 2.454097410127561e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10539
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.46875,
      "epoch": 0.5092771550057982,
      "grad_norm": 0.0,
      "learning_rate": 2.453855817549285e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10540
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.9375,
      "epoch": 0.5093254735214534,
      "grad_norm": 0.0,
      "learning_rate": 2.4536142249710086e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10541
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.3125,
      "epoch": 0.5093737920371086,
      "grad_norm": 0.0,
      "learning_rate": 2.453372632392733e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10542
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 102.78125,
      "epoch": 0.5094221105527639,
      "grad_norm": 0.0,
      "learning_rate": 2.453131039814457e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 10543
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.0625,
      "epoch": 0.509470429068419,
      "grad_norm": 0.0,
      "learning_rate": 2.452889447236181e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10544
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.78125,
      "epoch": 0.5095187475840742,
      "grad_norm": 0.0,
      "learning_rate": 2.4526478546579047e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 10545
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.5625,
      "epoch": 0.5095670660997295,
      "grad_norm": 0.0,
      "learning_rate": 2.452406262079629e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10546
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.71875,
      "epoch": 0.5096153846153846,
      "grad_norm": 0.0,
      "learning_rate": 2.452164669501353e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10547
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.40625,
      "epoch": 0.5096637031310398,
      "grad_norm": 0.0,
      "learning_rate": 2.4519230769230765e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10548
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.21875,
      "epoch": 0.509712021646695,
      "grad_norm": 0.0,
      "learning_rate": 2.4516814843448007e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10549
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.3125,
      "epoch": 0.5097603401623502,
      "grad_norm": 0.0,
      "learning_rate": 2.451439891766525e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10550
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.46875,
      "epoch": 0.5098086586780054,
      "grad_norm": 0.0,
      "learning_rate": 2.451198299188249e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10551
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.84375,
      "epoch": 0.5098569771936606,
      "grad_norm": 0.0,
      "learning_rate": 2.450956706609973e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10552
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.75,
      "epoch": 0.5099052957093159,
      "grad_norm": 0.0,
      "learning_rate": 2.4507151140316967e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10553
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 108.6875,
      "epoch": 0.509953614224971,
      "grad_norm": 0.0,
      "learning_rate": 2.450473521453421e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10554
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.75,
      "epoch": 0.5100019327406262,
      "grad_norm": 0.0,
      "learning_rate": 2.450231928875145e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10555
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.78125,
      "epoch": 0.5100502512562815,
      "grad_norm": 0.0,
      "learning_rate": 2.4499903362968686e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10556
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.375,
      "epoch": 0.5100985697719366,
      "grad_norm": 0.0,
      "learning_rate": 2.4497487437185927e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10557
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.5,
      "epoch": 0.5101468882875918,
      "grad_norm": 0.0,
      "learning_rate": 2.449507151140317e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10558
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.9375,
      "epoch": 0.510195206803247,
      "grad_norm": 0.0,
      "learning_rate": 2.449265558562041e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10559
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.875,
      "epoch": 0.5102435253189022,
      "grad_norm": 0.0,
      "learning_rate": 2.449023965983765e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10560
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.84375,
      "epoch": 0.5102918438345574,
      "grad_norm": 0.0,
      "learning_rate": 2.4487823734054887e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880052089691,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10561
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.9375,
      "epoch": 0.5103401623502126,
      "grad_norm": 0.0,
      "learning_rate": 2.448540780827213e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10562
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.5625,
      "epoch": 0.5103884808658679,
      "grad_norm": 0.0,
      "learning_rate": 2.448299188248937e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10563
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.1875,
      "epoch": 0.510436799381523,
      "grad_norm": 0.0,
      "learning_rate": 2.448057595670661e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10564
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.8125,
      "epoch": 0.5104851178971782,
      "grad_norm": 0.0,
      "learning_rate": 2.4478160030923853e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10565
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.5625,
      "epoch": 0.5105334364128334,
      "grad_norm": 0.0,
      "learning_rate": 2.447574410514109e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10566
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.125,
      "epoch": 0.5105817549284886,
      "grad_norm": 0.0,
      "learning_rate": 2.447332817935833e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10567
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.5625,
      "epoch": 0.5106300734441438,
      "grad_norm": 0.0,
      "learning_rate": 2.4470912253575566e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10568
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.46875,
      "epoch": 0.510678391959799,
      "grad_norm": 0.0,
      "learning_rate": 2.446849632779281e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10569
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.71875,
      "epoch": 0.5107267104754541,
      "grad_norm": 0.0,
      "learning_rate": 2.446608040201005e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10570
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.25,
      "epoch": 0.5107750289911094,
      "grad_norm": 0.0,
      "learning_rate": 2.446366447622729e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10571
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.40625,
      "epoch": 0.5108233475067646,
      "grad_norm": 0.0,
      "learning_rate": 2.446124855044453e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10572
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.90625,
      "epoch": 0.5108716660224198,
      "grad_norm": 0.0,
      "learning_rate": 2.445883262466177e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10573
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 103.5,
      "epoch": 0.510919984538075,
      "grad_norm": 0.0,
      "learning_rate": 2.445641669887901e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10574
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.96875,
      "epoch": 0.5109683030537302,
      "grad_norm": 0.0,
      "learning_rate": 2.445400077309625e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10575
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.34375,
      "epoch": 0.5110166215693854,
      "grad_norm": 0.0,
      "learning_rate": 2.4451584847313486e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10576
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.4375,
      "epoch": 0.5110649400850406,
      "grad_norm": 0.0,
      "learning_rate": 2.444916892153073e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10577
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.84375,
      "epoch": 0.5111132586006958,
      "grad_norm": 42.43490112526713,
      "learning_rate": 2.444675299574797e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10578
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.78125,
      "epoch": 0.511161577116351,
      "grad_norm": 0.0,
      "learning_rate": 2.444433706996521e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10579
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.0625,
      "epoch": 0.5112098956320061,
      "grad_norm": 0.0,
      "learning_rate": 2.444192114418245e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10580
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.09375,
      "epoch": 0.5112582141476614,
      "grad_norm": 0.0,
      "learning_rate": 2.443950521839969e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 10581
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.53125,
      "epoch": 0.5113065326633166,
      "grad_norm": 0.0,
      "learning_rate": 2.443708929261693e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10582
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.09375,
      "epoch": 0.5113548511789718,
      "grad_norm": 0.0,
      "learning_rate": 2.443467336683417e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10583
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.15625,
      "epoch": 0.511403169694627,
      "grad_norm": 0.0,
      "learning_rate": 2.443225744105141e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10584
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.875,
      "epoch": 0.5114514882102822,
      "grad_norm": 0.0,
      "learning_rate": 2.4429841515268653e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10585
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.5625,
      "epoch": 0.5114998067259374,
      "grad_norm": 0.0,
      "learning_rate": 2.442742558948589e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10586
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.15625,
      "epoch": 0.5115481252415925,
      "grad_norm": 0.0,
      "learning_rate": 2.442500966370313e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10587
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.375,
      "epoch": 0.5115964437572478,
      "grad_norm": 0.0,
      "learning_rate": 2.4422593737920367e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10588
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.65625,
      "epoch": 0.511644762272903,
      "grad_norm": 0.0,
      "learning_rate": 2.442017781213761e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10589
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.0625,
      "epoch": 0.5116930807885581,
      "grad_norm": 0.0,
      "learning_rate": 2.441776188635485e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10590
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.375,
      "epoch": 0.5117413993042134,
      "grad_norm": 0.0,
      "learning_rate": 2.441534596057209e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10591
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 100.59375,
      "epoch": 0.5117897178198686,
      "grad_norm": 0.0,
      "learning_rate": 2.441293003478933e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10592
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.15625,
      "epoch": 0.5118380363355238,
      "grad_norm": 0.0,
      "learning_rate": 2.441051410900657e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10593
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.34375,
      "epoch": 0.511886354851179,
      "grad_norm": 0.0,
      "learning_rate": 2.440809818322381e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 10594
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.5625,
      "epoch": 0.5119346733668342,
      "grad_norm": 0.0,
      "learning_rate": 2.440568225744105e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 10595
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.6875,
      "epoch": 0.5119829918824894,
      "grad_norm": 0.0,
      "learning_rate": 2.4403266331658287e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10596
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.90625,
      "epoch": 0.5120313103981445,
      "grad_norm": 0.0,
      "learning_rate": 2.440085040587553e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10597
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.1875,
      "epoch": 0.5120796289137998,
      "grad_norm": 0.0,
      "learning_rate": 2.439843448009277e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10598
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.25,
      "epoch": 0.512127947429455,
      "grad_norm": 0.0,
      "learning_rate": 2.439601855431001e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10599
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.46875,
      "epoch": 0.5121762659451101,
      "grad_norm": 0.0,
      "learning_rate": 2.4393602628527253e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10600
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.375,
      "epoch": 0.5122245844607654,
      "grad_norm": 0.0,
      "learning_rate": 2.439118670274449e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 10601
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.78125,
      "epoch": 0.5122729029764206,
      "grad_norm": 0.0,
      "learning_rate": 2.438877077696173e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10602
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 83.375,
      "epoch": 0.5123212214920758,
      "grad_norm": 16.37954583927189,
      "learning_rate": 2.438635485117897e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10603
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.8125,
      "epoch": 0.5123695400077309,
      "grad_norm": 0.0,
      "learning_rate": 2.4383938925396213e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10604
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 93.03125,
      "epoch": 0.5124178585233862,
      "grad_norm": 54.97454256548343,
      "learning_rate": 2.4381522999613454e-07,
      "loss": -0.0,
      "reward": 2.234375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10605
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.90625,
      "epoch": 0.5124661770390414,
      "grad_norm": 0.0,
      "learning_rate": 2.437910707383069e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10606
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.8125,
      "epoch": 0.5125144955546965,
      "grad_norm": 0.0,
      "learning_rate": 2.437669114804793e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10607
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.5,
      "epoch": 0.5125628140703518,
      "grad_norm": 0.0,
      "learning_rate": 2.437427522226517e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10608
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.71875,
      "epoch": 0.512611132586007,
      "grad_norm": 0.0,
      "learning_rate": 2.437185929648241e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10609
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.375,
      "epoch": 0.5126594511016621,
      "grad_norm": 0.0,
      "learning_rate": 2.436944337069965e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10610
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.625,
      "epoch": 0.5127077696173173,
      "grad_norm": 0.0,
      "learning_rate": 2.436702744491689e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10611
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.5625,
      "epoch": 0.5127560881329726,
      "grad_norm": 0.0,
      "learning_rate": 2.4364611519134133e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10612
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.84375,
      "epoch": 0.5128044066486278,
      "grad_norm": 0.0,
      "learning_rate": 2.436219559335137e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10613
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.9375,
      "epoch": 0.5128527251642829,
      "grad_norm": 0.0,
      "learning_rate": 2.435977966756861e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10614
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.0,
      "epoch": 0.5129010436799382,
      "grad_norm": 0.0,
      "learning_rate": 2.435736374178585e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10615
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.15625,
      "epoch": 0.5129493621955934,
      "grad_norm": 0.0,
      "learning_rate": 2.4354947816003093e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10616
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.53125,
      "epoch": 0.5129976807112485,
      "grad_norm": 0.0,
      "learning_rate": 2.435253189022033e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10617
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.59375,
      "epoch": 0.5130459992269037,
      "grad_norm": 0.0,
      "learning_rate": 2.435011596443757e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10618
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.09375,
      "epoch": 0.513094317742559,
      "grad_norm": 0.0,
      "learning_rate": 2.434770003865481e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10619
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.875,
      "epoch": 0.5131426362582141,
      "grad_norm": 0.0,
      "learning_rate": 2.4345284112872053e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10620
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 97.9375,
      "epoch": 0.5131909547738693,
      "grad_norm": 0.0,
      "learning_rate": 2.434286818708929e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10621
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.5625,
      "epoch": 0.5132392732895246,
      "grad_norm": 0.0,
      "learning_rate": 2.434045226130653e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10622
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.78125,
      "epoch": 0.5132875918051798,
      "grad_norm": 0.0,
      "learning_rate": 2.433803633552377e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10623
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.3125,
      "epoch": 0.5133359103208349,
      "grad_norm": 0.0,
      "learning_rate": 2.4335620409741014e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10624
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.65625,
      "epoch": 0.5133842288364902,
      "grad_norm": 0.0,
      "learning_rate": 2.4333204483958255e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10625
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.8125,
      "epoch": 0.5134325473521454,
      "grad_norm": 0.0,
      "learning_rate": 2.433078855817549e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10626
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.5625,
      "epoch": 0.5134808658678005,
      "grad_norm": 0.0,
      "learning_rate": 2.432837263239273e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10627
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.4375,
      "epoch": 0.5135291843834557,
      "grad_norm": 0.0,
      "learning_rate": 2.432595670660997e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10628
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.84375,
      "epoch": 0.513577502899111,
      "grad_norm": 0.0,
      "learning_rate": 2.432354078082721e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10629
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.0625,
      "epoch": 0.5136258214147661,
      "grad_norm": 0.0,
      "learning_rate": 2.432112485504445e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10630
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.53125,
      "epoch": 0.5136741399304213,
      "grad_norm": 0.0,
      "learning_rate": 2.431870892926169e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10631
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.78125,
      "epoch": 0.5137224584460766,
      "grad_norm": 0.0,
      "learning_rate": 2.4316293003478934e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10632
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.4375,
      "epoch": 0.5137707769617318,
      "grad_norm": 0.0,
      "learning_rate": 2.431387707769617e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10633
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.375,
      "epoch": 0.5138190954773869,
      "grad_norm": 0.0,
      "learning_rate": 2.431146115191341e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10634
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.15625,
      "epoch": 0.5138674139930421,
      "grad_norm": 0.0,
      "learning_rate": 2.4309045226130653e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10635
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.25,
      "epoch": 0.5139157325086974,
      "grad_norm": 0.0,
      "learning_rate": 2.4306629300347894e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10636
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.6875,
      "epoch": 0.5139640510243525,
      "grad_norm": 0.0,
      "learning_rate": 2.4304213374565135e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10637
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.5625,
      "epoch": 0.5140123695400077,
      "grad_norm": 0.0,
      "learning_rate": 2.430179744878237e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10638
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.28125,
      "epoch": 0.514060688055663,
      "grad_norm": 0.0,
      "learning_rate": 2.4299381522999613e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10639
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.875,
      "epoch": 0.5141090065713181,
      "grad_norm": 0.0,
      "learning_rate": 2.4296965597216854e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10640
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.09375,
      "epoch": 0.5141573250869733,
      "grad_norm": 0.0,
      "learning_rate": 2.429454967143409e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10641
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.84375,
      "epoch": 0.5142056436026285,
      "grad_norm": 0.0,
      "learning_rate": 2.429213374565133e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10642
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.65625,
      "epoch": 0.5142539621182838,
      "grad_norm": 0.0,
      "learning_rate": 2.4289717819868573e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10643
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.46875,
      "epoch": 0.5143022806339389,
      "grad_norm": 0.0,
      "learning_rate": 2.4287301894085814e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.19727617502212524,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10644
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.28125,
      "epoch": 0.5143505991495941,
      "grad_norm": 0.0,
      "learning_rate": 2.4284885968303056e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10645
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.5,
      "epoch": 0.5143989176652494,
      "grad_norm": 0.0,
      "learning_rate": 2.428247004252029e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10646
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.03125,
      "epoch": 0.5144472361809045,
      "grad_norm": 0.0,
      "learning_rate": 2.4280054116737533e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10647
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.625,
      "epoch": 0.5144955546965597,
      "grad_norm": 0.0,
      "learning_rate": 2.427763819095477e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10648
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.6875,
      "epoch": 0.514543873212215,
      "grad_norm": 0.0,
      "learning_rate": 2.427522226517201e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10649
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.71875,
      "epoch": 0.5145921917278701,
      "grad_norm": 0.0,
      "learning_rate": 2.427280633938925e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10650
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.0,
      "epoch": 0.5146405102435253,
      "grad_norm": 0.0,
      "learning_rate": 2.4270390413606493e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10651
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.0,
      "epoch": 0.5146888287591805,
      "grad_norm": 0.0,
      "learning_rate": 2.4267974487823735e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10652
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.90625,
      "epoch": 0.5147371472748358,
      "grad_norm": 0.0,
      "learning_rate": 2.426555856204097e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10653
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.84375,
      "epoch": 0.5147854657904909,
      "grad_norm": 0.0,
      "learning_rate": 2.426314263625821e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10654
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.46875,
      "epoch": 0.5148337843061461,
      "grad_norm": 0.0,
      "learning_rate": 2.4260726710475454e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10655
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.1875,
      "epoch": 0.5148821028218014,
      "grad_norm": 0.0,
      "learning_rate": 2.4258310784692695e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10656
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.28125,
      "epoch": 0.5149304213374565,
      "grad_norm": 0.0,
      "learning_rate": 2.4255894858909936e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10657
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.6875,
      "epoch": 0.5149787398531117,
      "grad_norm": 0.0,
      "learning_rate": 2.425347893312717e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10658
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.28125,
      "epoch": 0.5150270583687669,
      "grad_norm": 0.0,
      "learning_rate": 2.4251063007344414e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10659
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.375,
      "epoch": 0.5150753768844221,
      "grad_norm": 0.0,
      "learning_rate": 2.4248647081561655e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10660
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.875,
      "epoch": 0.5151236954000773,
      "grad_norm": 0.0,
      "learning_rate": 2.424623115577889e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10661
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.375,
      "epoch": 0.5151720139157325,
      "grad_norm": 0.0,
      "learning_rate": 2.424381522999613e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10662
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.4375,
      "epoch": 0.5152203324313878,
      "grad_norm": 0.0,
      "learning_rate": 2.4241399304213374e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 10663
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.40625,
      "epoch": 0.5152686509470429,
      "grad_norm": 0.0,
      "learning_rate": 2.4238983378430615e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10664
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.53125,
      "epoch": 0.5153169694626981,
      "grad_norm": 0.0,
      "learning_rate": 2.4236567452647857e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10665
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 99.59375,
      "epoch": 0.5153652879783533,
      "grad_norm": 77.99403732004852,
      "learning_rate": 2.4234151526865093e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.31663835048675537,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10666
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.59375,
      "epoch": 0.5154136064940085,
      "grad_norm": 0.0,
      "learning_rate": 2.4231735601082334e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10667
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.78125,
      "epoch": 0.5154619250096637,
      "grad_norm": 0.0,
      "learning_rate": 2.422931967529957e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10668
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.53125,
      "epoch": 0.5155102435253189,
      "grad_norm": 0.0,
      "learning_rate": 2.422690374951681e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10669
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.71875,
      "epoch": 0.515558562040974,
      "grad_norm": 0.0,
      "learning_rate": 2.4224487823734053e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10670
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.21875,
      "epoch": 0.5156068805566293,
      "grad_norm": 0.0,
      "learning_rate": 2.4222071897951294e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10671
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.6875,
      "epoch": 0.5156551990722845,
      "grad_norm": 0.0,
      "learning_rate": 2.4219655972168536e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10672
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 101.46875,
      "epoch": 0.5157035175879398,
      "grad_norm": 15.654975380419389,
      "learning_rate": 2.421724004638577e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.2920827269554138,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10673
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.65625,
      "epoch": 0.5157518361035949,
      "grad_norm": 0.0,
      "learning_rate": 2.4214824120603013e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10674
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.125,
      "epoch": 0.5158001546192501,
      "grad_norm": 0.0,
      "learning_rate": 2.4212408194820254e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10675
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.21875,
      "epoch": 0.5158484731349053,
      "grad_norm": 0.0,
      "learning_rate": 2.4209992269037496e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10676
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.125,
      "epoch": 0.5158967916505605,
      "grad_norm": 0.0,
      "learning_rate": 2.4207576343254737e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10677
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 98.875,
      "epoch": 0.5159451101662157,
      "grad_norm": 0.0,
      "learning_rate": 2.4205160417471973e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 10678
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.6875,
      "epoch": 0.5159934286818709,
      "grad_norm": 0.0,
      "learning_rate": 2.4202744491689214e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10679
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.40625,
      "epoch": 0.516041747197526,
      "grad_norm": 0.0,
      "learning_rate": 2.4200328565906456e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10680
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.03125,
      "epoch": 0.5160900657131813,
      "grad_norm": 0.0,
      "learning_rate": 2.419791264012369e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10681
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.75,
      "epoch": 0.5161383842288365,
      "grad_norm": 0.0,
      "learning_rate": 2.4195496714340933e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10682
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.9375,
      "epoch": 0.5161867027444917,
      "grad_norm": 0.0,
      "learning_rate": 2.4193080788558175e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10683
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.0,
      "epoch": 0.5162350212601469,
      "grad_norm": 0.0,
      "learning_rate": 2.4190664862775416e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10684
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.6875,
      "epoch": 0.5162833397758021,
      "grad_norm": 0.0,
      "learning_rate": 2.418824893699266e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 10685
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.40625,
      "epoch": 0.5163316582914573,
      "grad_norm": 0.0,
      "learning_rate": 2.4185833011209893e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10686
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.625,
      "epoch": 0.5163799768071125,
      "grad_norm": 0.0,
      "learning_rate": 2.4183417085427135e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10687
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.875,
      "epoch": 0.5164282953227677,
      "grad_norm": 0.0,
      "learning_rate": 2.4181001159644376e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10688
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.59375,
      "epoch": 0.5164766138384229,
      "grad_norm": 0.0,
      "learning_rate": 2.417858523386161e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10689
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.21875,
      "epoch": 0.516524932354078,
      "grad_norm": 0.0,
      "learning_rate": 2.4176169308078854e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10690
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.5625,
      "epoch": 0.5165732508697333,
      "grad_norm": 0.0,
      "learning_rate": 2.4173753382296095e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10691
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.5625,
      "epoch": 0.5166215693853885,
      "grad_norm": 0.0,
      "learning_rate": 2.4171337456513336e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10692
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.5,
      "epoch": 0.5166698879010437,
      "grad_norm": 0.0,
      "learning_rate": 2.416892153073057e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10693
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.5,
      "epoch": 0.5167182064166989,
      "grad_norm": 11.062791655377774,
      "learning_rate": 2.4166505604947814e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10694
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.75,
      "epoch": 0.5167665249323541,
      "grad_norm": 0.0,
      "learning_rate": 2.4164089679165055e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10695
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.6875,
      "epoch": 0.5168148434480093,
      "grad_norm": 0.0,
      "learning_rate": 2.4161673753382296e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10696
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.0,
      "epoch": 0.5168631619636644,
      "grad_norm": 0.0,
      "learning_rate": 2.415925782759954e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10697
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.625,
      "epoch": 0.5169114804793197,
      "grad_norm": 0.0,
      "learning_rate": 2.4156841901816774e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10698
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.5625,
      "epoch": 0.5169597989949749,
      "grad_norm": 0.0,
      "learning_rate": 2.4154425976034015e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10699
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.15625,
      "epoch": 0.51700811751063,
      "grad_norm": 0.0,
      "learning_rate": 2.4152010050251257e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.19727617502212524,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10700
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.78125,
      "epoch": 0.5170564360262853,
      "grad_norm": 0.0,
      "learning_rate": 2.4149594124468493e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10701
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.59375,
      "epoch": 0.5171047545419405,
      "grad_norm": 0.0,
      "learning_rate": 2.4147178198685734e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10702
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.53125,
      "epoch": 0.5171530730575957,
      "grad_norm": 0.0,
      "learning_rate": 2.4144762272902975e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10703
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.71875,
      "epoch": 0.5172013915732508,
      "grad_norm": 0.0,
      "learning_rate": 2.4142346347120217e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10704
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.75,
      "epoch": 0.5172497100889061,
      "grad_norm": 0.0,
      "learning_rate": 2.413993042133746e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10705
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 113.75,
      "epoch": 0.5172980286045613,
      "grad_norm": 0.0,
      "learning_rate": 2.4137514495554694e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 10706
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.96875,
      "epoch": 0.5173463471202164,
      "grad_norm": 0.0,
      "learning_rate": 2.4135098569771936e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10707
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.125,
      "epoch": 0.5173946656358717,
      "grad_norm": 0.0,
      "learning_rate": 2.4132682643989177e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10708
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.90625,
      "epoch": 0.5174429841515269,
      "grad_norm": 0.0,
      "learning_rate": 2.413026671820642e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10709
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.46875,
      "epoch": 0.517491302667182,
      "grad_norm": 0.0,
      "learning_rate": 2.4127850792423654e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10710
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.1875,
      "epoch": 0.5175396211828373,
      "grad_norm": 0.0,
      "learning_rate": 2.4125434866640896e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10711
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.03125,
      "epoch": 0.5175879396984925,
      "grad_norm": 0.0,
      "learning_rate": 2.4123018940858137e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10712
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.4375,
      "epoch": 0.5176362582141477,
      "grad_norm": 0.0,
      "learning_rate": 2.4120603015075373e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10713
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.8125,
      "epoch": 0.5176845767298028,
      "grad_norm": 0.0,
      "learning_rate": 2.4118187089292615e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10714
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.8125,
      "epoch": 0.5177328952454581,
      "grad_norm": 0.0,
      "learning_rate": 2.4115771163509856e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10715
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.5625,
      "epoch": 0.5177812137611133,
      "grad_norm": 0.0,
      "learning_rate": 2.4113355237727097e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10716
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.65625,
      "epoch": 0.5178295322767684,
      "grad_norm": 0.0,
      "learning_rate": 2.411093931194434e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10717
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.625,
      "epoch": 0.5178778507924237,
      "grad_norm": 0.0,
      "learning_rate": 2.4108523386161575e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10718
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.0,
      "epoch": 0.5179261693080789,
      "grad_norm": 0.0,
      "learning_rate": 2.4106107460378816e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10719
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 99.96875,
      "epoch": 0.517974487823734,
      "grad_norm": 23.799129464149093,
      "learning_rate": 2.410369153459606e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.28247910737991333,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10720
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.28125,
      "epoch": 0.5180228063393892,
      "grad_norm": 0.0,
      "learning_rate": 2.4101275608813294e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10721
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.03125,
      "epoch": 0.5180711248550445,
      "grad_norm": 0.0,
      "learning_rate": 2.4098859683030535e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10722
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.5,
      "epoch": 0.5181194433706997,
      "grad_norm": 0.0,
      "learning_rate": 2.4096443757247776e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10723
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.28125,
      "epoch": 0.5181677618863548,
      "grad_norm": 0.0,
      "learning_rate": 2.409402783146502e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10724
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.9375,
      "epoch": 0.5182160804020101,
      "grad_norm": 0.0,
      "learning_rate": 2.409161190568226e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10725
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.40625,
      "epoch": 0.5182643989176653,
      "grad_norm": 0.0,
      "learning_rate": 2.4089195979899495e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10726
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.03125,
      "epoch": 0.5183127174333204,
      "grad_norm": 0.0,
      "learning_rate": 2.4086780054116736e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10727
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.03125,
      "epoch": 0.5183610359489756,
      "grad_norm": 0.0,
      "learning_rate": 2.408436412833398e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10728
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.28125,
      "epoch": 0.5184093544646309,
      "grad_norm": 0.0,
      "learning_rate": 2.408194820255122e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10729
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.25,
      "epoch": 0.518457672980286,
      "grad_norm": 0.0,
      "learning_rate": 2.4079532276768455e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10730
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.25,
      "epoch": 0.5185059914959412,
      "grad_norm": 0.0,
      "learning_rate": 2.4077116350985697e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10731
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 98.09375,
      "epoch": 0.5185543100115965,
      "grad_norm": 0.0,
      "learning_rate": 2.407470042520294e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10732
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.09375,
      "epoch": 0.5186026285272517,
      "grad_norm": 0.0,
      "learning_rate": 2.4072284499420174e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10733
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.59375,
      "epoch": 0.5186509470429068,
      "grad_norm": 0.0,
      "learning_rate": 2.4069868573637415e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10734
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.09375,
      "epoch": 0.518699265558562,
      "grad_norm": 0.0,
      "learning_rate": 2.4067452647854657e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10735
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.21875,
      "epoch": 0.5187475840742173,
      "grad_norm": 0.0,
      "learning_rate": 2.40650367220719e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10736
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.96875,
      "epoch": 0.5187959025898724,
      "grad_norm": 0.0,
      "learning_rate": 2.406262079628914e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10737
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.03125,
      "epoch": 0.5188442211055276,
      "grad_norm": 0.0,
      "learning_rate": 2.4060204870506375e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10738
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.84375,
      "epoch": 0.5188925396211829,
      "grad_norm": 0.0,
      "learning_rate": 2.4057788944723617e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10739
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.90625,
      "epoch": 0.518940858136838,
      "grad_norm": 0.0,
      "learning_rate": 2.405537301894086e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10740
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.9375,
      "epoch": 0.5189891766524932,
      "grad_norm": 0.0,
      "learning_rate": 2.4052957093158094e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10741
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.125,
      "epoch": 0.5190374951681485,
      "grad_norm": 0.0,
      "learning_rate": 2.4050541167375336e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10742
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 100.09375,
      "epoch": 0.5190858136838037,
      "grad_norm": 32.13920817094833,
      "learning_rate": 2.4048125241592577e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.20785759389400482,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10743
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.4375,
      "epoch": 0.5191341321994588,
      "grad_norm": 0.0,
      "learning_rate": 2.404570931580982e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10744
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.21875,
      "epoch": 0.519182450715114,
      "grad_norm": 0.0,
      "learning_rate": 2.404329339002706e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10745
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.1875,
      "epoch": 0.5192307692307693,
      "grad_norm": 0.0,
      "learning_rate": 2.4040877464244296e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10746
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.90625,
      "epoch": 0.5192790877464244,
      "grad_norm": 0.0,
      "learning_rate": 2.4038461538461537e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10747
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.0625,
      "epoch": 0.5193274062620796,
      "grad_norm": 0.0,
      "learning_rate": 2.403604561267878e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10748
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.46875,
      "epoch": 0.5193757247777349,
      "grad_norm": 0.0,
      "learning_rate": 2.403362968689602e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10749
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.6875,
      "epoch": 0.51942404329339,
      "grad_norm": 0.0,
      "learning_rate": 2.403121376111326e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10750
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.96875,
      "epoch": 0.5194723618090452,
      "grad_norm": 0.0,
      "learning_rate": 2.4028797835330497e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10751
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.40625,
      "epoch": 0.5195206803247004,
      "grad_norm": 0.0,
      "learning_rate": 2.402638190954774e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10752
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 96.46875,
      "epoch": 0.5195689988403557,
      "grad_norm": 56.19390940552968,
      "learning_rate": 2.4023965983764975e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10753
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.59375,
      "epoch": 0.5196173173560108,
      "grad_norm": 0.0,
      "learning_rate": 2.4021550057982216e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10754
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.46875,
      "epoch": 0.519665635871666,
      "grad_norm": 0.0,
      "learning_rate": 2.401913413219946e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10755
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.59375,
      "epoch": 0.5197139543873213,
      "grad_norm": 0.0,
      "learning_rate": 2.40167182064167e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10756
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.96875,
      "epoch": 0.5197622729029764,
      "grad_norm": 0.0,
      "learning_rate": 2.401430228063394e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10757
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.6875,
      "epoch": 0.5198105914186316,
      "grad_norm": 0.0,
      "learning_rate": 2.4011886354851176e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10758
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.3125,
      "epoch": 0.5198589099342869,
      "grad_norm": 0.0,
      "learning_rate": 2.400947042906842e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10759
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.21875,
      "epoch": 0.519907228449942,
      "grad_norm": 0.0,
      "learning_rate": 2.400705450328566e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10760
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.25,
      "epoch": 0.5199555469655972,
      "grad_norm": 0.0,
      "learning_rate": 2.4004638577502895e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10761
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.78125,
      "epoch": 0.5200038654812524,
      "grad_norm": 0.0,
      "learning_rate": 2.4002222651720136e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10762
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.03125,
      "epoch": 0.5200521839969077,
      "grad_norm": 0.0,
      "learning_rate": 2.399980672593738e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10763
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.375,
      "epoch": 0.5201005025125628,
      "grad_norm": 0.0,
      "learning_rate": 2.399739080015462e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10764
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.6875,
      "epoch": 0.520148821028218,
      "grad_norm": 0.0,
      "learning_rate": 2.399497487437186e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10765
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.25,
      "epoch": 0.5201971395438733,
      "grad_norm": 0.0,
      "learning_rate": 2.3992558948589097e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10766
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.53125,
      "epoch": 0.5202454580595284,
      "grad_norm": 0.0,
      "learning_rate": 2.399014302280634e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10767
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.0625,
      "epoch": 0.5202937765751836,
      "grad_norm": 0.0,
      "learning_rate": 2.398772709702358e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10768
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.15625,
      "epoch": 0.5203420950908388,
      "grad_norm": 0.0,
      "learning_rate": 2.398531117124082e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10769
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.875,
      "epoch": 0.520390413606494,
      "grad_norm": 0.0,
      "learning_rate": 2.398289524545806e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10770
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.625,
      "epoch": 0.5204387321221492,
      "grad_norm": 0.0,
      "learning_rate": 2.39804793196753e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10771
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.125,
      "epoch": 0.5204870506378044,
      "grad_norm": 0.0,
      "learning_rate": 2.397806339389254e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10772
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.03125,
      "epoch": 0.5205353691534597,
      "grad_norm": 0.0,
      "learning_rate": 2.3975647468109776e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10773
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.46875,
      "epoch": 0.5205836876691148,
      "grad_norm": 0.0,
      "learning_rate": 2.3973231542327017e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 10774
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.3125,
      "epoch": 0.52063200618477,
      "grad_norm": 0.0,
      "learning_rate": 2.397081561654426e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10775
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 98.09375,
      "epoch": 0.5206803247004252,
      "grad_norm": 0.0,
      "learning_rate": 2.39683996907615e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 10776
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.375,
      "epoch": 0.5207286432160804,
      "grad_norm": 0.0,
      "learning_rate": 2.396598376497874e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 10777
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.0625,
      "epoch": 0.5207769617317356,
      "grad_norm": 0.0,
      "learning_rate": 2.3963567839195977e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880052089691,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10778
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.0625,
      "epoch": 0.5208252802473908,
      "grad_norm": 0.0,
      "learning_rate": 2.396115191341322e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10779
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.0625,
      "epoch": 0.520873598763046,
      "grad_norm": 0.0,
      "learning_rate": 2.395873598763046e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10780
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.8125,
      "epoch": 0.5209219172787012,
      "grad_norm": 0.0,
      "learning_rate": 2.3956320061847696e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10781
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 83.0,
      "epoch": 0.5209702357943564,
      "grad_norm": 67.25466944531293,
      "learning_rate": 2.3953904136064937e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.30617380142211914,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10782
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.9375,
      "epoch": 0.5210185543100117,
      "grad_norm": 0.0,
      "learning_rate": 2.395148821028218e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10783
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.28125,
      "epoch": 0.5210668728256668,
      "grad_norm": 0.0,
      "learning_rate": 2.394907228449942e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10784
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.0625,
      "epoch": 0.521115191341322,
      "grad_norm": 0.0,
      "learning_rate": 2.394665635871666e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10785
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.375,
      "epoch": 0.5211635098569772,
      "grad_norm": 0.0,
      "learning_rate": 2.39442404329339e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10786
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.1875,
      "epoch": 0.5212118283726324,
      "grad_norm": 0.0,
      "learning_rate": 2.394182450715114e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10787
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.875,
      "epoch": 0.5212601468882876,
      "grad_norm": 0.0,
      "learning_rate": 2.393940858136838e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10788
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.34375,
      "epoch": 0.5213084654039428,
      "grad_norm": 0.0,
      "learning_rate": 2.393699265558562e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10789
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.65625,
      "epoch": 0.5213567839195979,
      "grad_norm": 0.0,
      "learning_rate": 2.3934576729802863e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10790
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.40625,
      "epoch": 0.5214051024352532,
      "grad_norm": 0.0,
      "learning_rate": 2.39321608040201e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10791
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.90625,
      "epoch": 0.5214534209509084,
      "grad_norm": 0.0,
      "learning_rate": 2.392974487823734e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10792
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.28125,
      "epoch": 0.5215017394665636,
      "grad_norm": 0.0,
      "learning_rate": 2.3927328952454576e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10793
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.5625,
      "epoch": 0.5215500579822188,
      "grad_norm": 0.0,
      "learning_rate": 2.392491302667182e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10794
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.71875,
      "epoch": 0.521598376497874,
      "grad_norm": 0.0,
      "learning_rate": 2.392249710088906e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10795
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.46875,
      "epoch": 0.5216466950135292,
      "grad_norm": 0.0,
      "learning_rate": 2.39200811751063e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10796
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.1875,
      "epoch": 0.5216950135291843,
      "grad_norm": 0.0,
      "learning_rate": 2.391766524932354e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10797
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.78125,
      "epoch": 0.5217433320448396,
      "grad_norm": 0.0,
      "learning_rate": 2.391524932354078e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10798
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.15625,
      "epoch": 0.5217916505604948,
      "grad_norm": 0.0,
      "learning_rate": 2.391283339775802e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10799
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.0625,
      "epoch": 0.5218399690761499,
      "grad_norm": 0.0,
      "learning_rate": 2.391041747197526e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10800
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 89.4375,
      "epoch": 0.5218882875918052,
      "grad_norm": 34.204185507677806,
      "learning_rate": 2.39080015461925e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.20785759389400482,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10801
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.375,
      "epoch": 0.5219366061074604,
      "grad_norm": 0.0,
      "learning_rate": 2.390558562040974e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10802
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.6875,
      "epoch": 0.5219849246231156,
      "grad_norm": 0.0,
      "learning_rate": 2.390316969462698e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10803
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.9375,
      "epoch": 0.5220332431387708,
      "grad_norm": 0.0,
      "learning_rate": 2.390075376884422e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10804
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.9375,
      "epoch": 0.522081561654426,
      "grad_norm": 0.0,
      "learning_rate": 2.389833784306146e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10805
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.21875,
      "epoch": 0.5221298801700812,
      "grad_norm": 0.0,
      "learning_rate": 2.38959219172787e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10806
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.03125,
      "epoch": 0.5221781986857363,
      "grad_norm": 0.0,
      "learning_rate": 2.389350599149594e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10807
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.78125,
      "epoch": 0.5222265172013916,
      "grad_norm": 0.0,
      "learning_rate": 2.389109006571318e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10808
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.28125,
      "epoch": 0.5222748357170468,
      "grad_norm": 0.0,
      "learning_rate": 2.388867413993042e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10809
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.875,
      "epoch": 0.5223231542327019,
      "grad_norm": 0.0,
      "learning_rate": 2.3886258214147664e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10810
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.5625,
      "epoch": 0.5223714727483572,
      "grad_norm": 0.0,
      "learning_rate": 2.38838422883649e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 10811
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.5,
      "epoch": 0.5224197912640124,
      "grad_norm": 0.0,
      "learning_rate": 2.388142636258214e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10812
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.71875,
      "epoch": 0.5224681097796676,
      "grad_norm": 0.0,
      "learning_rate": 2.3879010436799377e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10813
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.625,
      "epoch": 0.5225164282953227,
      "grad_norm": 0.0,
      "learning_rate": 2.387659451101662e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10814
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.9375,
      "epoch": 0.522564746810978,
      "grad_norm": 0.0,
      "learning_rate": 2.387417858523386e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10815
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.8125,
      "epoch": 0.5226130653266332,
      "grad_norm": 0.0,
      "learning_rate": 2.38717626594511e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10816
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.5,
      "epoch": 0.5226613838422883,
      "grad_norm": 0.0,
      "learning_rate": 2.386934673366834e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10817
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.78125,
      "epoch": 0.5227097023579436,
      "grad_norm": 0.0,
      "learning_rate": 2.386693080788558e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10818
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.03125,
      "epoch": 0.5227580208735988,
      "grad_norm": 0.0,
      "learning_rate": 2.386451488210282e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10819
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.28125,
      "epoch": 0.5228063393892539,
      "grad_norm": 0.0,
      "learning_rate": 2.386209895632006e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10820
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 102.5625,
      "epoch": 0.5228546579049091,
      "grad_norm": 51.87943843262186,
      "learning_rate": 2.3859683030537303e-07,
      "loss": -0.0,
      "reward": 1.421875,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10821
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.8125,
      "epoch": 0.5229029764205644,
      "grad_norm": 0.0,
      "learning_rate": 2.3857267104754544e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.24251842498779297,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10822
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.4375,
      "epoch": 0.5229512949362196,
      "grad_norm": 0.0,
      "learning_rate": 2.385485117897178e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10823
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.4375,
      "epoch": 0.5229996134518747,
      "grad_norm": 0.0,
      "learning_rate": 2.385243525318902e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10824
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.59375,
      "epoch": 0.52304793196753,
      "grad_norm": 0.0,
      "learning_rate": 2.3850019327406263e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10825
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.03125,
      "epoch": 0.5230962504831852,
      "grad_norm": 0.0,
      "learning_rate": 2.38476034016235e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10826
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.75,
      "epoch": 0.5231445689988403,
      "grad_norm": 0.0,
      "learning_rate": 2.384518747584074e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10827
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.1875,
      "epoch": 0.5231928875144956,
      "grad_norm": 0.0,
      "learning_rate": 2.384277155005798e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10828
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.34375,
      "epoch": 0.5232412060301508,
      "grad_norm": 0.0,
      "learning_rate": 2.384035562427522e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10829
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.9375,
      "epoch": 0.5232895245458059,
      "grad_norm": 0.0,
      "learning_rate": 2.3837939698492462e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10830
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.15625,
      "epoch": 0.5233378430614611,
      "grad_norm": 0.0,
      "learning_rate": 2.38355237727097e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10831
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.28125,
      "epoch": 0.5233861615771164,
      "grad_norm": 0.0,
      "learning_rate": 2.3833107846926942e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10832
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.875,
      "epoch": 0.5234344800927716,
      "grad_norm": 0.0,
      "learning_rate": 2.383069192114418e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10833
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.78125,
      "epoch": 0.5234827986084267,
      "grad_norm": 0.0,
      "learning_rate": 2.3828275995361422e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10834
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.1875,
      "epoch": 0.523531117124082,
      "grad_norm": 0.0,
      "learning_rate": 2.3825860069578663e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10835
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.1875,
      "epoch": 0.5235794356397372,
      "grad_norm": 0.0,
      "learning_rate": 2.3823444143795902e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10836
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.625,
      "epoch": 0.5236277541553923,
      "grad_norm": 0.0,
      "learning_rate": 2.3821028218013143e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10837
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.34375,
      "epoch": 0.5236760726710475,
      "grad_norm": 0.0,
      "learning_rate": 2.381861229223038e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10838
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.75,
      "epoch": 0.5237243911867028,
      "grad_norm": 0.0,
      "learning_rate": 2.381619636644762e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10839
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.59375,
      "epoch": 0.5237727097023579,
      "grad_norm": 0.0,
      "learning_rate": 2.3813780440664862e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10840
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 83.21875,
      "epoch": 0.5238210282180131,
      "grad_norm": 0.0,
      "learning_rate": 2.38113645148821e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 10841
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.09375,
      "epoch": 0.5238693467336684,
      "grad_norm": 0.0,
      "learning_rate": 2.3808948589099342e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 10842
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.28125,
      "epoch": 0.5239176652493236,
      "grad_norm": 0.0,
      "learning_rate": 2.380653266331658e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10843
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.65625,
      "epoch": 0.5239659837649787,
      "grad_norm": 0.0,
      "learning_rate": 2.3804116737533822e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10844
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.53125,
      "epoch": 0.524014302280634,
      "grad_norm": 0.0,
      "learning_rate": 2.3801700811751064e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10845
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.375,
      "epoch": 0.5240626207962892,
      "grad_norm": 0.0,
      "learning_rate": 2.3799284885968302e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10846
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.875,
      "epoch": 0.5241109393119443,
      "grad_norm": 0.0,
      "learning_rate": 2.3796868960185544e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 10847
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.4375,
      "epoch": 0.5241592578275995,
      "grad_norm": 0.0,
      "learning_rate": 2.379445303440278e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10848
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 88.71875,
      "epoch": 0.5242075763432548,
      "grad_norm": 58.692612010619044,
      "learning_rate": 2.379203710862002e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10849
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.53125,
      "epoch": 0.5242558948589099,
      "grad_norm": 0.0,
      "learning_rate": 2.3789621182837263e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10850
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.59375,
      "epoch": 0.5243042133745651,
      "grad_norm": 0.0,
      "learning_rate": 2.37872052570545e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10851
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.5,
      "epoch": 0.5243525318902204,
      "grad_norm": 0.0,
      "learning_rate": 2.3784789331271743e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10852
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.375,
      "epoch": 0.5244008504058756,
      "grad_norm": 0.0,
      "learning_rate": 2.3782373405488981e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10853
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.5625,
      "epoch": 0.5244491689215307,
      "grad_norm": 0.0,
      "learning_rate": 2.3779957479706223e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 10854
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.78125,
      "epoch": 0.5244974874371859,
      "grad_norm": 0.0,
      "learning_rate": 2.3777541553923464e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10855
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.78125,
      "epoch": 0.5245458059528412,
      "grad_norm": 0.0,
      "learning_rate": 2.3775125628140703e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10856
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.5625,
      "epoch": 0.5245941244684963,
      "grad_norm": 0.0,
      "learning_rate": 2.3772709702357944e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10857
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.25,
      "epoch": 0.5246424429841515,
      "grad_norm": 0.0,
      "learning_rate": 2.377029377657518e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24251842498779297,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10858
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.96875,
      "epoch": 0.5246907614998068,
      "grad_norm": 0.0,
      "learning_rate": 2.3767877850792422e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10859
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.75,
      "epoch": 0.5247390800154619,
      "grad_norm": 0.0,
      "learning_rate": 2.3765461925009663e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10860
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.75,
      "epoch": 0.5247873985311171,
      "grad_norm": 0.0,
      "learning_rate": 2.3763045999226902e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10861
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.78125,
      "epoch": 0.5248357170467723,
      "grad_norm": 0.0,
      "learning_rate": 2.3760630073444143e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10862
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.4375,
      "epoch": 0.5248840355624276,
      "grad_norm": 0.0,
      "learning_rate": 2.3758214147661382e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10863
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.28125,
      "epoch": 0.5249323540780827,
      "grad_norm": 0.0,
      "learning_rate": 2.3755798221878623e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10864
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.9375,
      "epoch": 0.5249806725937379,
      "grad_norm": 0.0,
      "learning_rate": 2.3753382296095864e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10865
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.53125,
      "epoch": 0.5250289911093932,
      "grad_norm": 0.0,
      "learning_rate": 2.3750966370313103e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10866
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.3125,
      "epoch": 0.5250773096250483,
      "grad_norm": 0.0,
      "learning_rate": 2.3748550444530345e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 10867
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.15625,
      "epoch": 0.5251256281407035,
      "grad_norm": 22.743366982344497,
      "learning_rate": 2.3746134518747583e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.2641943693161011,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10868
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.1875,
      "epoch": 0.5251739466563587,
      "grad_norm": 0.0,
      "learning_rate": 2.3743718592964822e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10869
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.5,
      "epoch": 0.5252222651720139,
      "grad_norm": 0.0,
      "learning_rate": 2.3741302667182063e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10870
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.1875,
      "epoch": 0.5252705836876691,
      "grad_norm": 0.0,
      "learning_rate": 2.3738886741399302e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10871
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.6875,
      "epoch": 0.5253189022033243,
      "grad_norm": 0.0,
      "learning_rate": 2.3736470815616543e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10872
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.375,
      "epoch": 0.5253672207189796,
      "grad_norm": 0.0,
      "learning_rate": 2.3734054889833782e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10873
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.21875,
      "epoch": 0.5254155392346347,
      "grad_norm": 0.0,
      "learning_rate": 2.3731638964051023e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10874
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.96875,
      "epoch": 0.5254638577502899,
      "grad_norm": 0.0,
      "learning_rate": 2.3729223038268265e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10875
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.0625,
      "epoch": 0.5255121762659452,
      "grad_norm": 0.0,
      "learning_rate": 2.3726807112485504e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10876
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.125,
      "epoch": 0.5255604947816003,
      "grad_norm": 96.86510616324287,
      "learning_rate": 2.3724391186702745e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10877
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.90625,
      "epoch": 0.5256088132972555,
      "grad_norm": 0.0,
      "learning_rate": 2.3721975260919984e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10878
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.8125,
      "epoch": 0.5256571318129107,
      "grad_norm": 0.0,
      "learning_rate": 2.3719559335137222e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10879
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.6875,
      "epoch": 0.5257054503285659,
      "grad_norm": 0.0,
      "learning_rate": 2.3717143409354464e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10880
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.40625,
      "epoch": 0.5257537688442211,
      "grad_norm": 0.0,
      "learning_rate": 2.3714727483571702e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10881
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 102.09375,
      "epoch": 0.5258020873598763,
      "grad_norm": 0.0,
      "learning_rate": 2.3712311557788944e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10882
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.59375,
      "epoch": 0.5258504058755316,
      "grad_norm": 0.0,
      "learning_rate": 2.3709895632006183e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 10883
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.0,
      "epoch": 0.5258987243911867,
      "grad_norm": 0.0,
      "learning_rate": 2.3707479706223424e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10884
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.625,
      "epoch": 0.5259470429068419,
      "grad_norm": 0.0,
      "learning_rate": 2.3705063780440665e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10885
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.90625,
      "epoch": 0.5259953614224971,
      "grad_norm": 0.0,
      "learning_rate": 2.3702647854657904e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10886
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.53125,
      "epoch": 0.5260436799381523,
      "grad_norm": 0.0,
      "learning_rate": 2.3700231928875145e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10887
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.34375,
      "epoch": 0.5260919984538075,
      "grad_norm": 0.0,
      "learning_rate": 2.3697816003092384e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10888
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.34375,
      "epoch": 0.5261403169694627,
      "grad_norm": 0.0,
      "learning_rate": 2.3695400077309625e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10889
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.0,
      "epoch": 0.5261886354851179,
      "grad_norm": 0.0,
      "learning_rate": 2.3692984151526864e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10890
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.59375,
      "epoch": 0.5262369540007731,
      "grad_norm": 0.0,
      "learning_rate": 2.3690568225744103e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10891
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.90625,
      "epoch": 0.5262852725164283,
      "grad_norm": 0.0,
      "learning_rate": 2.3688152299961344e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10892
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.90625,
      "epoch": 0.5263335910320835,
      "grad_norm": 0.0,
      "learning_rate": 2.3685736374178583e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10893
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.28125,
      "epoch": 0.5263819095477387,
      "grad_norm": 0.0,
      "learning_rate": 2.3683320448395824e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10894
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.5,
      "epoch": 0.5264302280633939,
      "grad_norm": 0.0,
      "learning_rate": 2.3680904522613066e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10895
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 94.28125,
      "epoch": 0.5264785465790491,
      "grad_norm": 0.0,
      "learning_rate": 2.3678488596830304e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 10896
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.96875,
      "epoch": 0.5265268650947043,
      "grad_norm": 0.0,
      "learning_rate": 2.3676072671047546e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10897
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.3125,
      "epoch": 0.5265751836103595,
      "grad_norm": 0.0,
      "learning_rate": 2.3673656745264784e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10898
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.09375,
      "epoch": 0.5266235021260147,
      "grad_norm": 0.0,
      "learning_rate": 2.3671240819482026e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10899
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.71875,
      "epoch": 0.5266718206416698,
      "grad_norm": 0.0,
      "learning_rate": 2.3668824893699265e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10900
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.78125,
      "epoch": 0.5267201391573251,
      "grad_norm": 0.0,
      "learning_rate": 2.3666408967916503e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10901
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.96875,
      "epoch": 0.5267684576729803,
      "grad_norm": 0.0,
      "learning_rate": 2.3663993042133745e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10902
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.03125,
      "epoch": 0.5268167761886355,
      "grad_norm": 0.0,
      "learning_rate": 2.3661577116350983e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10903
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.09375,
      "epoch": 0.5268650947042907,
      "grad_norm": 0.0,
      "learning_rate": 2.3659161190568225e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10904
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.1875,
      "epoch": 0.5269134132199459,
      "grad_norm": 0.0,
      "learning_rate": 2.3656745264785466e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10905
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.15625,
      "epoch": 0.5269617317356011,
      "grad_norm": 0.0,
      "learning_rate": 2.3654329339002705e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10906
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.09375,
      "epoch": 0.5270100502512562,
      "grad_norm": 0.0,
      "learning_rate": 2.3651913413219946e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10907
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 95.5,
      "epoch": 0.5270583687669115,
      "grad_norm": 69.26075061738975,
      "learning_rate": 2.3649497487437185e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.24831002950668335,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10908
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.3125,
      "epoch": 0.5271066872825667,
      "grad_norm": 0.0,
      "learning_rate": 2.3647081561654426e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10909
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.09375,
      "epoch": 0.5271550057982218,
      "grad_norm": 0.0,
      "learning_rate": 2.3644665635871665e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10910
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.625,
      "epoch": 0.5272033243138771,
      "grad_norm": 0.0,
      "learning_rate": 2.3642249710088904e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10911
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.4375,
      "epoch": 0.5272516428295323,
      "grad_norm": 0.0,
      "learning_rate": 2.3639833784306145e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10912
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.625,
      "epoch": 0.5272999613451875,
      "grad_norm": 0.0,
      "learning_rate": 2.3637417858523384e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10913
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.1875,
      "epoch": 0.5273482798608427,
      "grad_norm": 0.0,
      "learning_rate": 2.3635001932740625e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10914
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.59375,
      "epoch": 0.5273965983764979,
      "grad_norm": 40.757457483903984,
      "learning_rate": 2.3632586006957866e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.24386970698833466,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10915
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.625,
      "epoch": 0.5274449168921531,
      "grad_norm": 0.0,
      "learning_rate": 2.3630170081175105e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10916
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.90625,
      "epoch": 0.5274932354078082,
      "grad_norm": 0.0,
      "learning_rate": 2.3627754155392347e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10917
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.625,
      "epoch": 0.5275415539234635,
      "grad_norm": 0.0,
      "learning_rate": 2.3625338229609585e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10918
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.5,
      "epoch": 0.5275898724391187,
      "grad_norm": 0.0,
      "learning_rate": 2.3622922303826827e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10919
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.46875,
      "epoch": 0.5276381909547738,
      "grad_norm": 0.0,
      "learning_rate": 2.3620506378044068e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10920
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.125,
      "epoch": 0.527686509470429,
      "grad_norm": 0.0,
      "learning_rate": 2.3618090452261304e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10921
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.03125,
      "epoch": 0.5277348279860843,
      "grad_norm": 0.0,
      "learning_rate": 2.3615674526478545e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10922
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.90625,
      "epoch": 0.5277831465017395,
      "grad_norm": 0.0,
      "learning_rate": 2.3613258600695784e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10923
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.90625,
      "epoch": 0.5278314650173946,
      "grad_norm": 0.0,
      "learning_rate": 2.3610842674913025e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10924
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.5,
      "epoch": 0.5278797835330499,
      "grad_norm": 0.0,
      "learning_rate": 2.3608426749130267e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10925
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.0,
      "epoch": 0.5279281020487051,
      "grad_norm": 0.0,
      "learning_rate": 2.3606010823347506e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 10926
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.84375,
      "epoch": 0.5279764205643602,
      "grad_norm": 0.0,
      "learning_rate": 2.3603594897564747e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10927
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.8125,
      "epoch": 0.5280247390800155,
      "grad_norm": 0.0,
      "learning_rate": 2.3601178971781986e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10928
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.375,
      "epoch": 0.5280730575956707,
      "grad_norm": 0.0,
      "learning_rate": 2.3598763045999227e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10929
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.125,
      "epoch": 0.5281213761113258,
      "grad_norm": 0.0,
      "learning_rate": 2.3596347120216468e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10930
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.65625,
      "epoch": 0.528169694626981,
      "grad_norm": 0.0,
      "learning_rate": 2.3593931194433704e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10931
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.75,
      "epoch": 0.5282180131426363,
      "grad_norm": 0.0,
      "learning_rate": 2.3591515268650946e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10932
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.3125,
      "epoch": 0.5282663316582915,
      "grad_norm": 0.0,
      "learning_rate": 2.3589099342868184e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10933
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.65625,
      "epoch": 0.5283146501739466,
      "grad_norm": 0.0,
      "learning_rate": 2.3586683417085426e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10934
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.75,
      "epoch": 0.5283629686896019,
      "grad_norm": 0.0,
      "learning_rate": 2.3584267491302667e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10935
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.8125,
      "epoch": 0.5284112872052571,
      "grad_norm": 0.0,
      "learning_rate": 2.3581851565519906e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10936
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.0,
      "epoch": 0.5284596057209122,
      "grad_norm": 0.0,
      "learning_rate": 2.3579435639737147e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10937
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.71875,
      "epoch": 0.5285079242365675,
      "grad_norm": 0.0,
      "learning_rate": 2.3577019713954386e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10938
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.53125,
      "epoch": 0.5285562427522227,
      "grad_norm": 0.0,
      "learning_rate": 2.3574603788171627e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 10939
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.875,
      "epoch": 0.5286045612678778,
      "grad_norm": 0.0,
      "learning_rate": 2.357218786238887e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 10940
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.3125,
      "epoch": 0.528652879783533,
      "grad_norm": 0.0,
      "learning_rate": 2.3569771936606105e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10941
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.90625,
      "epoch": 0.5287011982991883,
      "grad_norm": 0.0,
      "learning_rate": 2.3567356010823346e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10942
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.21875,
      "epoch": 0.5287495168148435,
      "grad_norm": 0.0,
      "learning_rate": 2.3564940085040585e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 10943
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.4375,
      "epoch": 0.5287978353304986,
      "grad_norm": 0.0,
      "learning_rate": 2.3562524159257826e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10944
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.375,
      "epoch": 0.5288461538461539,
      "grad_norm": 0.0,
      "learning_rate": 2.3560108233475068e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10945
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.375,
      "epoch": 0.5288944723618091,
      "grad_norm": 0.0,
      "learning_rate": 2.3557692307692306e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10946
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.84375,
      "epoch": 0.5289427908774642,
      "grad_norm": 0.0,
      "learning_rate": 2.3555276381909548e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10947
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.46875,
      "epoch": 0.5289911093931194,
      "grad_norm": 0.0,
      "learning_rate": 2.3552860456126786e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10948
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.6875,
      "epoch": 0.5290394279087747,
      "grad_norm": 0.0,
      "learning_rate": 2.3550444530344028e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10949
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.875,
      "epoch": 0.5290877464244298,
      "grad_norm": 0.0,
      "learning_rate": 2.354802860456127e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10950
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.03125,
      "epoch": 0.529136064940085,
      "grad_norm": 0.0,
      "learning_rate": 2.3545612678778505e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10951
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.1875,
      "epoch": 0.5291843834557403,
      "grad_norm": 0.0,
      "learning_rate": 2.3543196752995747e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10952
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.9375,
      "epoch": 0.5292327019713955,
      "grad_norm": 0.0,
      "learning_rate": 2.3540780827212985e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10953
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.59375,
      "epoch": 0.5292810204870506,
      "grad_norm": 0.0,
      "learning_rate": 2.3538364901430227e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10954
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.65625,
      "epoch": 0.5293293390027058,
      "grad_norm": 0.0,
      "learning_rate": 2.3535948975647468e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 10955
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.28125,
      "epoch": 0.5293776575183611,
      "grad_norm": 0.0,
      "learning_rate": 2.3533533049864707e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 10956
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.65625,
      "epoch": 0.5294259760340162,
      "grad_norm": 0.0,
      "learning_rate": 2.3531117124081948e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10957
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.9375,
      "epoch": 0.5294742945496714,
      "grad_norm": 0.0,
      "learning_rate": 2.3528701198299187e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10958
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.25,
      "epoch": 0.5295226130653267,
      "grad_norm": 0.0,
      "learning_rate": 2.3526285272516428e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 10959
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.1875,
      "epoch": 0.5295709315809818,
      "grad_norm": 0.0,
      "learning_rate": 2.352386934673367e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 10960
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.4375,
      "epoch": 0.529619250096637,
      "grad_norm": 0.0,
      "learning_rate": 2.3521453420950906e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10961
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.25,
      "epoch": 0.5296675686122922,
      "grad_norm": 0.0,
      "learning_rate": 2.3519037495168147e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10962
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.125,
      "epoch": 0.5297158871279475,
      "grad_norm": 0.0,
      "learning_rate": 2.3516621569385386e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10963
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.96875,
      "epoch": 0.5297642056436026,
      "grad_norm": 0.0,
      "learning_rate": 2.3514205643602627e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10964
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.125,
      "epoch": 0.5298125241592578,
      "grad_norm": 0.0,
      "learning_rate": 2.3511789717819868e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 10965
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.03125,
      "epoch": 0.5298608426749131,
      "grad_norm": 0.0,
      "learning_rate": 2.3509373792037107e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10966
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.21875,
      "epoch": 0.5299091611905682,
      "grad_norm": 0.0,
      "learning_rate": 2.3506957866254348e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10967
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.0625,
      "epoch": 0.5299574797062234,
      "grad_norm": 0.0,
      "learning_rate": 2.3504541940471587e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10968
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.28125,
      "epoch": 0.5300057982218787,
      "grad_norm": 0.0,
      "learning_rate": 2.3502126014688829e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10969
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.65625,
      "epoch": 0.5300541167375338,
      "grad_norm": 0.0,
      "learning_rate": 2.349971008890607e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10970
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.03125,
      "epoch": 0.530102435253189,
      "grad_norm": 0.0,
      "learning_rate": 2.3497294163123309e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10971
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.0,
      "epoch": 0.5301507537688442,
      "grad_norm": 0.0,
      "learning_rate": 2.3494878237340547e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 10972
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.875,
      "epoch": 0.5301990722844995,
      "grad_norm": 0.0,
      "learning_rate": 2.3492462311557786e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10973
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.65625,
      "epoch": 0.5302473908001546,
      "grad_norm": 0.0,
      "learning_rate": 2.3490046385775027e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10974
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.65625,
      "epoch": 0.5302957093158098,
      "grad_norm": 0.0,
      "learning_rate": 2.348763045999227e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10975
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.46875,
      "epoch": 0.5303440278314651,
      "grad_norm": 0.0,
      "learning_rate": 2.3485214534209508e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10976
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.5625,
      "epoch": 0.5303923463471202,
      "grad_norm": 0.0,
      "learning_rate": 2.348279860842675e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10977
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.6875,
      "epoch": 0.5304406648627754,
      "grad_norm": 0.0,
      "learning_rate": 2.3480382682643988e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10978
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.03125,
      "epoch": 0.5304889833784306,
      "grad_norm": 0.0,
      "learning_rate": 2.347796675686123e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 10979
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.34375,
      "epoch": 0.5305373018940858,
      "grad_norm": 0.0,
      "learning_rate": 2.347555083107847e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10980
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.1875,
      "epoch": 0.530585620409741,
      "grad_norm": 0.0,
      "learning_rate": 2.347313490529571e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 10981
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.53125,
      "epoch": 0.5306339389253962,
      "grad_norm": 0.0,
      "learning_rate": 2.3470718979512948e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 10982
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.25,
      "epoch": 0.5306822574410515,
      "grad_norm": 0.0,
      "learning_rate": 2.3468303053730186e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 10983
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.625,
      "epoch": 0.5307305759567066,
      "grad_norm": 0.0,
      "learning_rate": 2.3465887127947428e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10984
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.5,
      "epoch": 0.5307788944723618,
      "grad_norm": 0.0,
      "learning_rate": 2.346347120216467e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10985
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.34375,
      "epoch": 0.530827212988017,
      "grad_norm": 0.0,
      "learning_rate": 2.3461055276381908e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10986
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.5625,
      "epoch": 0.5308755315036722,
      "grad_norm": 0.0,
      "learning_rate": 2.345863935059915e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10987
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.75,
      "epoch": 0.5309238500193274,
      "grad_norm": 0.0,
      "learning_rate": 2.3456223424816388e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 10988
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.21875,
      "epoch": 0.5309721685349826,
      "grad_norm": 0.0,
      "learning_rate": 2.345380749903363e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10989
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.875,
      "epoch": 0.5310204870506378,
      "grad_norm": 0.0,
      "learning_rate": 2.345139157325087e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 10990
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.09375,
      "epoch": 0.531068805566293,
      "grad_norm": 0.0,
      "learning_rate": 2.344897564746811e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 10991
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.1875,
      "epoch": 0.5311171240819482,
      "grad_norm": 0.0,
      "learning_rate": 2.344655972168535e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 10992
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.5625,
      "epoch": 0.5311654425976035,
      "grad_norm": 0.0,
      "learning_rate": 2.3444143795902587e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 10993
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.53125,
      "epoch": 0.5312137611132586,
      "grad_norm": 0.0,
      "learning_rate": 2.3441727870119828e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10994
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.84375,
      "epoch": 0.5312620796289138,
      "grad_norm": 0.0,
      "learning_rate": 2.343931194433707e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10995
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.8125,
      "epoch": 0.531310398144569,
      "grad_norm": 0.0,
      "learning_rate": 2.3436896018554308e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 10996
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.03125,
      "epoch": 0.5313587166602242,
      "grad_norm": 0.0,
      "learning_rate": 2.343448009277155e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 10997
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.71875,
      "epoch": 0.5314070351758794,
      "grad_norm": 0.0,
      "learning_rate": 2.3432064166988788e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 10998
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.3125,
      "epoch": 0.5314553536915346,
      "grad_norm": 0.0,
      "learning_rate": 2.342964824120603e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 10999
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.34375,
      "epoch": 0.5315036722071897,
      "grad_norm": 0.0,
      "learning_rate": 2.342723231542327e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11000
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 84.28125,
      "epoch": 0.531551990722845,
      "grad_norm": 0.0,
      "learning_rate": 2.342481638964051e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 11001
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.65625,
      "epoch": 0.5316003092385002,
      "grad_norm": 0.0,
      "learning_rate": 2.342240046385775e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11002
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.0,
      "epoch": 0.5316486277541554,
      "grad_norm": 0.0,
      "learning_rate": 2.3419984538074987e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11003
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.625,
      "epoch": 0.5316969462698106,
      "grad_norm": 0.0,
      "learning_rate": 2.3417568612292229e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11004
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.9375,
      "epoch": 0.5317452647854658,
      "grad_norm": 0.0,
      "learning_rate": 2.341515268650947e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11005
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 92.90625,
      "epoch": 0.531793583301121,
      "grad_norm": 0.0,
      "learning_rate": 2.341273676072671e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11006
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.6875,
      "epoch": 0.5318419018167762,
      "grad_norm": 0.0,
      "learning_rate": 2.341032083494395e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11007
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.25,
      "epoch": 0.5318902203324314,
      "grad_norm": 0.0,
      "learning_rate": 2.340790490916119e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11008
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.6875,
      "epoch": 0.5319385388480866,
      "grad_norm": 0.0,
      "learning_rate": 2.340548898337843e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11009
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.96875,
      "epoch": 0.5319868573637417,
      "grad_norm": 0.0,
      "learning_rate": 2.3403073057595671e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11010
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.3125,
      "epoch": 0.532035175879397,
      "grad_norm": 0.0,
      "learning_rate": 2.340065713181291e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11011
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.78125,
      "epoch": 0.5320834943950522,
      "grad_norm": 0.0,
      "learning_rate": 2.3398241206030152e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11012
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.78125,
      "epoch": 0.5321318129107074,
      "grad_norm": 0.0,
      "learning_rate": 2.3395825280247388e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 11013
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.46875,
      "epoch": 0.5321801314263626,
      "grad_norm": 0.0,
      "learning_rate": 2.339340935446463e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11014
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.875,
      "epoch": 0.5322284499420178,
      "grad_norm": 0.0,
      "learning_rate": 2.339099342868187e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11015
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.25,
      "epoch": 0.532276768457673,
      "grad_norm": 0.0,
      "learning_rate": 2.338857750289911e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11016
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.8125,
      "epoch": 0.5323250869733281,
      "grad_norm": 0.0,
      "learning_rate": 2.338616157711635e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11017
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.28125,
      "epoch": 0.5323734054889834,
      "grad_norm": 0.0,
      "learning_rate": 2.338374565133359e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 11018
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.5625,
      "epoch": 0.5324217240046386,
      "grad_norm": 0.0,
      "learning_rate": 2.338132972555083e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 11019
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.4375,
      "epoch": 0.5324700425202937,
      "grad_norm": 0.0,
      "learning_rate": 2.3378913799768072e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11020
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.625,
      "epoch": 0.532518361035949,
      "grad_norm": 0.0,
      "learning_rate": 2.337649787398531e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11021
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.21875,
      "epoch": 0.5325666795516042,
      "grad_norm": 0.0,
      "learning_rate": 2.3374081948202552e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11022
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.28125,
      "epoch": 0.5326149980672594,
      "grad_norm": 0.0,
      "learning_rate": 2.3371666022419788e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11023
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.09375,
      "epoch": 0.5326633165829145,
      "grad_norm": 0.0,
      "learning_rate": 2.336925009663703e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11024
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.125,
      "epoch": 0.5327116350985698,
      "grad_norm": 0.0,
      "learning_rate": 2.336683417085427e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11025
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.625,
      "epoch": 0.532759953614225,
      "grad_norm": 0.0,
      "learning_rate": 2.336441824507151e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11026
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.75,
      "epoch": 0.5328082721298801,
      "grad_norm": 0.0,
      "learning_rate": 2.336200231928875e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11027
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.0,
      "epoch": 0.5328565906455354,
      "grad_norm": 0.0,
      "learning_rate": 2.335958639350599e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11028
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.34375,
      "epoch": 0.5329049091611906,
      "grad_norm": 0.0,
      "learning_rate": 2.335717046772323e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11029
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.875,
      "epoch": 0.5329532276768457,
      "grad_norm": 0.0,
      "learning_rate": 2.3354754541940472e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11030
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.21875,
      "epoch": 0.533001546192501,
      "grad_norm": 0.0,
      "learning_rate": 2.335233861615771e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11031
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.0625,
      "epoch": 0.5330498647081562,
      "grad_norm": 0.0,
      "learning_rate": 2.3349922690374952e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11032
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.09375,
      "epoch": 0.5330981832238114,
      "grad_norm": 0.0,
      "learning_rate": 2.3347506764592188e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11033
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.78125,
      "epoch": 0.5331465017394665,
      "grad_norm": 0.0,
      "learning_rate": 2.334509083880943e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11034
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.53125,
      "epoch": 0.5331948202551218,
      "grad_norm": 0.0,
      "learning_rate": 2.334267491302667e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11035
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.96875,
      "epoch": 0.533243138770777,
      "grad_norm": 0.0,
      "learning_rate": 2.334025898724391e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11036
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.75,
      "epoch": 0.5332914572864321,
      "grad_norm": 0.0,
      "learning_rate": 2.333784306146115e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11037
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.0625,
      "epoch": 0.5333397758020874,
      "grad_norm": 0.0,
      "learning_rate": 2.333542713567839e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11038
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.3125,
      "epoch": 0.5333880943177426,
      "grad_norm": 0.0,
      "learning_rate": 2.333301120989563e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11039
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.25,
      "epoch": 0.5334364128333977,
      "grad_norm": 0.0,
      "learning_rate": 2.3330595284112873e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11040
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.0625,
      "epoch": 0.5334847313490529,
      "grad_norm": 0.0,
      "learning_rate": 2.3328179358330111e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11041
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.96875,
      "epoch": 0.5335330498647082,
      "grad_norm": 0.0,
      "learning_rate": 2.3325763432547353e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11042
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.03125,
      "epoch": 0.5335813683803634,
      "grad_norm": 0.0,
      "learning_rate": 2.3323347506764591e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11043
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.15625,
      "epoch": 0.5336296868960185,
      "grad_norm": 0.0,
      "learning_rate": 2.332093158098183e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11044
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.21875,
      "epoch": 0.5336780054116738,
      "grad_norm": 0.0,
      "learning_rate": 2.3318515655199072e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11045
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.09375,
      "epoch": 0.533726323927329,
      "grad_norm": 0.0,
      "learning_rate": 2.331609972941631e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11046
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 97.125,
      "epoch": 0.5337746424429841,
      "grad_norm": 0.0,
      "learning_rate": 2.3313683803633552e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11047
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.53125,
      "epoch": 0.5338229609586393,
      "grad_norm": 0.0,
      "learning_rate": 2.331126787785079e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11048
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.6875,
      "epoch": 0.5338712794742946,
      "grad_norm": 0.0,
      "learning_rate": 2.3308851952068032e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11049
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.0625,
      "epoch": 0.5339195979899497,
      "grad_norm": 0.0,
      "learning_rate": 2.3306436026285273e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 11050
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.5625,
      "epoch": 0.5339679165056049,
      "grad_norm": 0.0,
      "learning_rate": 2.3304020100502512e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11051
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.09375,
      "epoch": 0.5340162350212602,
      "grad_norm": 0.0,
      "learning_rate": 2.3301604174719753e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11052
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.6875,
      "epoch": 0.5340645535369154,
      "grad_norm": 0.0,
      "learning_rate": 2.3299188248936992e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11053
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.75,
      "epoch": 0.5341128720525705,
      "grad_norm": 0.0,
      "learning_rate": 2.329677232315423e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11054
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.6875,
      "epoch": 0.5341611905682258,
      "grad_norm": 0.0,
      "learning_rate": 2.3294356397371472e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11055
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.625,
      "epoch": 0.534209509083881,
      "grad_norm": 0.0,
      "learning_rate": 2.329194047158871e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11056
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.46875,
      "epoch": 0.5342578275995361,
      "grad_norm": 0.0,
      "learning_rate": 2.3289524545805952e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11057
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.71875,
      "epoch": 0.5343061461151913,
      "grad_norm": 0.0,
      "learning_rate": 2.328710862002319e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11058
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.1875,
      "epoch": 0.5343544646308466,
      "grad_norm": 0.0,
      "learning_rate": 2.3284692694240432e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11059
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.5,
      "epoch": 0.5344027831465017,
      "grad_norm": 0.0,
      "learning_rate": 2.3282276768457673e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11060
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.875,
      "epoch": 0.5344511016621569,
      "grad_norm": 0.0,
      "learning_rate": 2.3279860842674912e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11061
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.28125,
      "epoch": 0.5344994201778122,
      "grad_norm": 0.0,
      "learning_rate": 2.3277444916892154e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11062
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.15625,
      "epoch": 0.5345477386934674,
      "grad_norm": 0.0,
      "learning_rate": 2.3275028991109392e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11063
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.375,
      "epoch": 0.5345960572091225,
      "grad_norm": 43.36872950456265,
      "learning_rate": 2.327261306532663e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11064
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.8125,
      "epoch": 0.5346443757247777,
      "grad_norm": 0.0,
      "learning_rate": 2.3270197139543872e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11065
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.3125,
      "epoch": 0.534692694240433,
      "grad_norm": 0.0,
      "learning_rate": 2.326778121376111e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11066
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.59375,
      "epoch": 0.5347410127560881,
      "grad_norm": 0.0,
      "learning_rate": 2.3265365287978352e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11067
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.59375,
      "epoch": 0.5347893312717433,
      "grad_norm": 0.0,
      "learning_rate": 2.326294936219559e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11068
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.875,
      "epoch": 0.5348376497873986,
      "grad_norm": 0.0,
      "learning_rate": 2.3260533436412832e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11069
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.625,
      "epoch": 0.5348859683030537,
      "grad_norm": 0.0,
      "learning_rate": 2.3258117510630074e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11070
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.78125,
      "epoch": 0.5349342868187089,
      "grad_norm": 0.0,
      "learning_rate": 2.3255701584847313e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11071
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.28125,
      "epoch": 0.5349826053343641,
      "grad_norm": 0.0,
      "learning_rate": 2.3253285659064554e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11072
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.9375,
      "epoch": 0.5350309238500194,
      "grad_norm": 0.0,
      "learning_rate": 2.3250869733281793e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11073
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.84375,
      "epoch": 0.5350792423656745,
      "grad_norm": 0.0,
      "learning_rate": 2.3248453807499034e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11074
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.75,
      "epoch": 0.5351275608813297,
      "grad_norm": 0.0,
      "learning_rate": 2.3246037881716273e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11075
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.21875,
      "epoch": 0.535175879396985,
      "grad_norm": 0.0,
      "learning_rate": 2.3243621955933511e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11076
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.875,
      "epoch": 0.5352241979126401,
      "grad_norm": 0.0,
      "learning_rate": 2.3241206030150753e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11077
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.59375,
      "epoch": 0.5352725164282953,
      "grad_norm": 0.0,
      "learning_rate": 2.3238790104367992e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11078
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.125,
      "epoch": 0.5353208349439506,
      "grad_norm": 0.0,
      "learning_rate": 2.3236374178585233e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11079
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.125,
      "epoch": 0.5353691534596057,
      "grad_norm": 0.0,
      "learning_rate": 2.3233958252802474e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11080
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 96.9375,
      "epoch": 0.5354174719752609,
      "grad_norm": 110.99419977587448,
      "learning_rate": 2.3231542327019713e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.22738909721374512,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11081
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.15625,
      "epoch": 0.5354657904909161,
      "grad_norm": 0.0,
      "learning_rate": 2.3229126401236954e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11082
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.28125,
      "epoch": 0.5355141090065714,
      "grad_norm": 0.0,
      "learning_rate": 2.3226710475454193e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11083
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.09375,
      "epoch": 0.5355624275222265,
      "grad_norm": 0.0,
      "learning_rate": 2.3224294549671434e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11084
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.59375,
      "epoch": 0.5356107460378817,
      "grad_norm": 0.0,
      "learning_rate": 2.3221878623888673e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11085
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.4375,
      "epoch": 0.535659064553537,
      "grad_norm": 0.0,
      "learning_rate": 2.3219462698105912e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11086
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.0,
      "epoch": 0.5357073830691921,
      "grad_norm": 0.0,
      "learning_rate": 2.3217046772323153e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11087
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.34375,
      "epoch": 0.5357557015848473,
      "grad_norm": 0.0,
      "learning_rate": 2.3214630846540392e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11088
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.78125,
      "epoch": 0.5358040201005025,
      "grad_norm": 0.0,
      "learning_rate": 2.3212214920757633e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11089
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.28125,
      "epoch": 0.5358523386161577,
      "grad_norm": 0.0,
      "learning_rate": 2.3209798994974875e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11090
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.4375,
      "epoch": 0.5359006571318129,
      "grad_norm": 0.0,
      "learning_rate": 2.3207383069192113e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11091
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.875,
      "epoch": 0.5359489756474681,
      "grad_norm": 0.0,
      "learning_rate": 2.3204967143409355e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11092
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.0,
      "epoch": 0.5359972941631234,
      "grad_norm": 0.0,
      "learning_rate": 2.3202551217626593e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11093
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.90625,
      "epoch": 0.5360456126787785,
      "grad_norm": 0.0,
      "learning_rate": 2.3200135291843835e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11094
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.0,
      "epoch": 0.5360939311944337,
      "grad_norm": 0.0,
      "learning_rate": 2.3197719366061074e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11095
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.84375,
      "epoch": 0.536142249710089,
      "grad_norm": 0.0,
      "learning_rate": 2.3195303440278312e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11096
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.03125,
      "epoch": 0.5361905682257441,
      "grad_norm": 0.0,
      "learning_rate": 2.3192887514495554e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11097
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.625,
      "epoch": 0.5362388867413993,
      "grad_norm": 0.0,
      "learning_rate": 2.3190471588712792e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11098
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.9375,
      "epoch": 0.5362872052570545,
      "grad_norm": 0.0,
      "learning_rate": 2.3188055662930034e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11099
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.15625,
      "epoch": 0.5363355237727097,
      "grad_norm": 0.0,
      "learning_rate": 2.3185639737147275e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11100
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.46875,
      "epoch": 0.5363838422883649,
      "grad_norm": 0.0,
      "learning_rate": 2.3183223811364514e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11101
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.40625,
      "epoch": 0.5364321608040201,
      "grad_norm": 0.0,
      "learning_rate": 2.3180807885581755e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11102
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.28125,
      "epoch": 0.5364804793196754,
      "grad_norm": 0.0,
      "learning_rate": 2.3178391959798994e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11103
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.65625,
      "epoch": 0.5365287978353305,
      "grad_norm": 0.0,
      "learning_rate": 2.3175976034016235e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880350112915,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11104
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.875,
      "epoch": 0.5365771163509857,
      "grad_norm": 0.0,
      "learning_rate": 2.3173560108233477e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11105
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.375,
      "epoch": 0.5366254348666409,
      "grad_norm": 0.0,
      "learning_rate": 2.3171144182450713e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11106
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.90625,
      "epoch": 0.5366737533822961,
      "grad_norm": 0.0,
      "learning_rate": 2.3168728256667954e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11107
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.96875,
      "epoch": 0.5367220718979513,
      "grad_norm": 0.0,
      "learning_rate": 2.3166312330885193e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11108
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.78125,
      "epoch": 0.5367703904136065,
      "grad_norm": 0.0,
      "learning_rate": 2.3163896405102434e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11109
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.375,
      "epoch": 0.5368187089292616,
      "grad_norm": 0.0,
      "learning_rate": 2.3161480479319675e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11110
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.59375,
      "epoch": 0.5368670274449169,
      "grad_norm": 0.0,
      "learning_rate": 2.3159064553536914e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11111
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.21875,
      "epoch": 0.5369153459605721,
      "grad_norm": 0.0,
      "learning_rate": 2.3156648627754155e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11112
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.71875,
      "epoch": 0.5369636644762273,
      "grad_norm": 0.0,
      "learning_rate": 2.3154232701971394e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11113
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.96875,
      "epoch": 0.5370119829918825,
      "grad_norm": 0.0,
      "learning_rate": 2.3151816776188636e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11114
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.9375,
      "epoch": 0.5370603015075377,
      "grad_norm": 0.0,
      "learning_rate": 2.3149400850405877e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11115
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 77.9375,
      "epoch": 0.5371086200231929,
      "grad_norm": 6.267747918219851,
      "learning_rate": 2.3146984924623113e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11116
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.3125,
      "epoch": 0.537156938538848,
      "grad_norm": 0.0,
      "learning_rate": 2.3144568998840354e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11117
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.84375,
      "epoch": 0.5372052570545033,
      "grad_norm": 0.0,
      "learning_rate": 2.3142153073057593e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11118
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.21875,
      "epoch": 0.5372535755701585,
      "grad_norm": 0.0,
      "learning_rate": 2.3139737147274834e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11119
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.125,
      "epoch": 0.5373018940858136,
      "grad_norm": 0.0,
      "learning_rate": 2.3137321221492076e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11120
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.8125,
      "epoch": 0.5373502126014689,
      "grad_norm": 0.0,
      "learning_rate": 2.3134905295709315e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11121
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.875,
      "epoch": 0.5373985311171241,
      "grad_norm": 0.0,
      "learning_rate": 2.3132489369926556e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11122
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.03125,
      "epoch": 0.5374468496327793,
      "grad_norm": 0.0,
      "learning_rate": 2.3130073444143795e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11123
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.90625,
      "epoch": 0.5374951681484345,
      "grad_norm": 0.0,
      "learning_rate": 2.3127657518361036e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11124
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.59375,
      "epoch": 0.5375434866640897,
      "grad_norm": 0.0,
      "learning_rate": 2.3125241592578277e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11125
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.59375,
      "epoch": 0.5375918051797449,
      "grad_norm": 0.0,
      "learning_rate": 2.3122825666795513e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11126
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.375,
      "epoch": 0.5376401236954,
      "grad_norm": 0.0,
      "learning_rate": 2.3120409741012755e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11127
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.78125,
      "epoch": 0.5376884422110553,
      "grad_norm": 0.0,
      "learning_rate": 2.3117993815229993e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 11128
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.5625,
      "epoch": 0.5377367607267105,
      "grad_norm": 0.0,
      "learning_rate": 2.3115577889447235e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11129
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.0,
      "epoch": 0.5377850792423656,
      "grad_norm": 0.0,
      "learning_rate": 2.3113161963664476e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11130
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.6875,
      "epoch": 0.5378333977580209,
      "grad_norm": 0.0,
      "learning_rate": 2.3110746037881715e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11131
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.03125,
      "epoch": 0.5378817162736761,
      "grad_norm": 0.0,
      "learning_rate": 2.3108330112098956e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11132
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.875,
      "epoch": 0.5379300347893313,
      "grad_norm": 0.0,
      "learning_rate": 2.3105914186316195e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11133
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.59375,
      "epoch": 0.5379783533049864,
      "grad_norm": 0.0,
      "learning_rate": 2.3103498260533436e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11134
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.6875,
      "epoch": 0.5380266718206417,
      "grad_norm": 0.0,
      "learning_rate": 2.3101082334750678e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11135
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.25,
      "epoch": 0.5380749903362969,
      "grad_norm": 0.0,
      "learning_rate": 2.3098666408967914e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11136
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.53125,
      "epoch": 0.538123308851952,
      "grad_norm": 0.0,
      "learning_rate": 2.3096250483185155e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11137
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.75,
      "epoch": 0.5381716273676073,
      "grad_norm": 0.0,
      "learning_rate": 2.3093834557402394e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11138
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.84375,
      "epoch": 0.5382199458832625,
      "grad_norm": 0.0,
      "learning_rate": 2.3091418631619635e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11139
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.875,
      "epoch": 0.5382682643989176,
      "grad_norm": 0.0,
      "learning_rate": 2.3089002705836877e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11140
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.46875,
      "epoch": 0.5383165829145728,
      "grad_norm": 0.0,
      "learning_rate": 2.3086586780054115e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11141
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.9375,
      "epoch": 0.5383649014302281,
      "grad_norm": 0.0,
      "learning_rate": 2.3084170854271357e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11142
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.34375,
      "epoch": 0.5384132199458833,
      "grad_norm": 0.0,
      "learning_rate": 2.3081754928488595e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11143
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.21875,
      "epoch": 0.5384615384615384,
      "grad_norm": 0.0,
      "learning_rate": 2.3079339002705837e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11144
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.125,
      "epoch": 0.5385098569771937,
      "grad_norm": 0.0,
      "learning_rate": 2.3076923076923078e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11145
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.78125,
      "epoch": 0.5385581754928489,
      "grad_norm": 0.0,
      "learning_rate": 2.3074507151140317e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11146
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.65625,
      "epoch": 0.538606494008504,
      "grad_norm": 0.0,
      "learning_rate": 2.3072091225357556e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11147
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.3125,
      "epoch": 0.5386548125241593,
      "grad_norm": 0.0,
      "learning_rate": 2.3069675299574794e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11148
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.0,
      "epoch": 0.5387031310398145,
      "grad_norm": 0.0,
      "learning_rate": 2.3067259373792036e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11149
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.125,
      "epoch": 0.5387514495554696,
      "grad_norm": 0.0,
      "learning_rate": 2.3064843448009277e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11150
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.46875,
      "epoch": 0.5387997680711248,
      "grad_norm": 0.0,
      "learning_rate": 2.3062427522226516e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11151
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.46875,
      "epoch": 0.5388480865867801,
      "grad_norm": 0.0,
      "learning_rate": 2.3060011596443757e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 11152
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.46875,
      "epoch": 0.5388964051024353,
      "grad_norm": 0.0,
      "learning_rate": 2.3057595670660996e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11153
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.59375,
      "epoch": 0.5389447236180904,
      "grad_norm": 0.0,
      "learning_rate": 2.3055179744878237e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11154
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.0,
      "epoch": 0.5389930421337457,
      "grad_norm": 0.0,
      "learning_rate": 2.3052763819095479e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11155
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.90625,
      "epoch": 0.5390413606494009,
      "grad_norm": 0.0,
      "learning_rate": 2.3050347893312717e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11156
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.71875,
      "epoch": 0.539089679165056,
      "grad_norm": 0.0,
      "learning_rate": 2.3047931967529956e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11157
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.46875,
      "epoch": 0.5391379976807112,
      "grad_norm": 0.0,
      "learning_rate": 2.3045516041747195e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11158
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.8125,
      "epoch": 0.5391863161963665,
      "grad_norm": 0.0,
      "learning_rate": 2.3043100115964436e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11159
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.75,
      "epoch": 0.5392346347120216,
      "grad_norm": 0.0,
      "learning_rate": 2.3040684190181677e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11160
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.375,
      "epoch": 0.5392829532276768,
      "grad_norm": 0.0,
      "learning_rate": 2.3038268264398916e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11161
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.03125,
      "epoch": 0.5393312717433321,
      "grad_norm": 0.0,
      "learning_rate": 2.3035852338616157e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11162
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.25,
      "epoch": 0.5393795902589873,
      "grad_norm": 0.0,
      "learning_rate": 2.3033436412833396e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11163
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.25,
      "epoch": 0.5394279087746424,
      "grad_norm": 0.0,
      "learning_rate": 2.3031020487050638e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11164
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.28125,
      "epoch": 0.5394762272902976,
      "grad_norm": 0.0,
      "learning_rate": 2.302860456126788e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11165
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.5625,
      "epoch": 0.5395245458059529,
      "grad_norm": 0.0,
      "learning_rate": 2.3026188635485118e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11166
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.15625,
      "epoch": 0.539572864321608,
      "grad_norm": 0.0,
      "learning_rate": 2.3023772709702356e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11167
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.375,
      "epoch": 0.5396211828372632,
      "grad_norm": 0.0,
      "learning_rate": 2.3021356783919595e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11168
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.5625,
      "epoch": 0.5396695013529185,
      "grad_norm": 0.0,
      "learning_rate": 2.3018940858136836e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11169
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.1875,
      "epoch": 0.5397178198685736,
      "grad_norm": 0.0,
      "learning_rate": 2.3016524932354078e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11170
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.09375,
      "epoch": 0.5397661383842288,
      "grad_norm": 0.0,
      "learning_rate": 2.3014109006571316e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11171
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.03125,
      "epoch": 0.539814456899884,
      "grad_norm": 0.0,
      "learning_rate": 2.3011693080788558e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11172
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.25,
      "epoch": 0.5398627754155393,
      "grad_norm": 0.0,
      "learning_rate": 2.3009277155005797e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11173
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.1875,
      "epoch": 0.5399110939311944,
      "grad_norm": 0.0,
      "learning_rate": 2.3006861229223038e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11174
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.90625,
      "epoch": 0.5399594124468496,
      "grad_norm": 0.0,
      "learning_rate": 2.300444530344028e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11175
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.28125,
      "epoch": 0.5400077309625049,
      "grad_norm": 0.0,
      "learning_rate": 2.3002029377657518e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11176
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.09375,
      "epoch": 0.54005604947816,
      "grad_norm": 0.0,
      "learning_rate": 2.299961345187476e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11177
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.875,
      "epoch": 0.5401043679938152,
      "grad_norm": 0.0,
      "learning_rate": 2.2997197526091995e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11178
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.78125,
      "epoch": 0.5401526865094705,
      "grad_norm": 0.0,
      "learning_rate": 2.2994781600309237e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11179
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.25,
      "epoch": 0.5402010050251256,
      "grad_norm": 0.0,
      "learning_rate": 2.2992365674526478e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11180
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.71875,
      "epoch": 0.5402493235407808,
      "grad_norm": 0.0,
      "learning_rate": 2.2989949748743717e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11181
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.03125,
      "epoch": 0.540297642056436,
      "grad_norm": 0.0,
      "learning_rate": 2.2987533822960958e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11182
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.375,
      "epoch": 0.5403459605720913,
      "grad_norm": 0.0,
      "learning_rate": 2.2985117897178197e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11183
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.15625,
      "epoch": 0.5403942790877464,
      "grad_norm": 0.0,
      "learning_rate": 2.2982701971395438e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11184
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.03125,
      "epoch": 0.5404425976034016,
      "grad_norm": 0.0,
      "learning_rate": 2.298028604561268e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11185
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.78125,
      "epoch": 0.5404909161190569,
      "grad_norm": 0.0,
      "learning_rate": 2.2977870119829918e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11186
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.21875,
      "epoch": 0.540539234634712,
      "grad_norm": 0.0,
      "learning_rate": 2.297545419404716e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11187
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.125,
      "epoch": 0.5405875531503672,
      "grad_norm": 0.0,
      "learning_rate": 2.2973038268264396e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11188
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.3125,
      "epoch": 0.5406358716660224,
      "grad_norm": 0.0,
      "learning_rate": 2.2970622342481637e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11189
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.84375,
      "epoch": 0.5406841901816776,
      "grad_norm": 0.0,
      "learning_rate": 2.2968206416698879e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11190
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.65625,
      "epoch": 0.5407325086973328,
      "grad_norm": 0.0,
      "learning_rate": 2.2965790490916117e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11191
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.65625,
      "epoch": 0.540780827212988,
      "grad_norm": 0.0,
      "learning_rate": 2.2963374565133359e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11192
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.90625,
      "epoch": 0.5408291457286433,
      "grad_norm": 0.0,
      "learning_rate": 2.2960958639350597e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11193
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.375,
      "epoch": 0.5408774642442984,
      "grad_norm": 0.0,
      "learning_rate": 2.295854271356784e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11194
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.40625,
      "epoch": 0.5409257827599536,
      "grad_norm": 0.0,
      "learning_rate": 2.295612678778508e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11195
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.84375,
      "epoch": 0.5409741012756089,
      "grad_norm": 0.0,
      "learning_rate": 2.295371086200232e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11196
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.5625,
      "epoch": 0.541022419791264,
      "grad_norm": 0.0,
      "learning_rate": 2.295129493621956e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11197
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.0625,
      "epoch": 0.5410707383069192,
      "grad_norm": 0.0,
      "learning_rate": 2.2948879010436796e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11198
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.21875,
      "epoch": 0.5411190568225744,
      "grad_norm": 0.0,
      "learning_rate": 2.2946463084654038e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11199
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.125,
      "epoch": 0.5411673753382296,
      "grad_norm": 0.0,
      "learning_rate": 2.294404715887128e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11200
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.8125,
      "epoch": 0.5412156938538848,
      "grad_norm": 0.0,
      "learning_rate": 2.2941631233088518e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11201
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.0625,
      "epoch": 0.54126401236954,
      "grad_norm": 0.0,
      "learning_rate": 2.293921530730576e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11202
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 87.5625,
      "epoch": 0.5413123308851953,
      "grad_norm": 38.03575114533472,
      "learning_rate": 2.2936799381522998e-07,
      "loss": 0.0,
      "reward": 1.484375,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 11203
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.40625,
      "epoch": 0.5413606494008504,
      "grad_norm": 0.0,
      "learning_rate": 2.293438345574024e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11204
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.40625,
      "epoch": 0.5414089679165056,
      "grad_norm": 0.0,
      "learning_rate": 2.293196752995748e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11205
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.4375,
      "epoch": 0.5414572864321608,
      "grad_norm": 21.245413592948395,
      "learning_rate": 2.292955160417472e-07,
      "loss": -0.0,
      "reward": 2.25,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11206
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.34375,
      "epoch": 0.541505604947816,
      "grad_norm": 0.0,
      "learning_rate": 2.292713567839196e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11207
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.625,
      "epoch": 0.5415539234634712,
      "grad_norm": 0.0,
      "learning_rate": 2.2924719752609197e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11208
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.5,
      "epoch": 0.5416022419791264,
      "grad_norm": 0.0,
      "learning_rate": 2.2922303826826438e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11209
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.78125,
      "epoch": 0.5416505604947816,
      "grad_norm": 0.0,
      "learning_rate": 2.291988790104368e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11210
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.9375,
      "epoch": 0.5416988790104368,
      "grad_norm": 0.0,
      "learning_rate": 2.2917471975260918e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11211
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.4375,
      "epoch": 0.541747197526092,
      "grad_norm": 0.0,
      "learning_rate": 2.291505604947816e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11212
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.03125,
      "epoch": 0.5417955160417472,
      "grad_norm": 0.0,
      "learning_rate": 2.2912640123695398e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11213
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.90625,
      "epoch": 0.5418438345574024,
      "grad_norm": 0.0,
      "learning_rate": 2.291022419791264e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11214
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.90625,
      "epoch": 0.5418921530730576,
      "grad_norm": 0.0,
      "learning_rate": 2.290780827212988e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11215
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.40625,
      "epoch": 0.5419404715887128,
      "grad_norm": 0.0,
      "learning_rate": 2.290539234634712e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11216
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.5,
      "epoch": 0.541988790104368,
      "grad_norm": 0.0,
      "learning_rate": 2.290297642056436e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11217
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.1875,
      "epoch": 0.5420371086200232,
      "grad_norm": 0.0,
      "learning_rate": 2.29005604947816e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11218
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.40625,
      "epoch": 0.5420854271356784,
      "grad_norm": 0.0,
      "learning_rate": 2.2898144568998838e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11219
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.25,
      "epoch": 0.5421337456513335,
      "grad_norm": 0.0,
      "learning_rate": 2.289572864321608e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11220
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.3125,
      "epoch": 0.5421820641669888,
      "grad_norm": 0.0,
      "learning_rate": 2.2893312717433318e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11221
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.59375,
      "epoch": 0.542230382682644,
      "grad_norm": 0.0,
      "learning_rate": 2.289089679165056e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11222
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.0,
      "epoch": 0.5422787011982992,
      "grad_norm": 0.0,
      "learning_rate": 2.2888480865867799e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11223
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.84375,
      "epoch": 0.5423270197139544,
      "grad_norm": 0.0,
      "learning_rate": 2.288606494008504e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11224
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.4375,
      "epoch": 0.5423753382296096,
      "grad_norm": 0.0,
      "learning_rate": 2.288364901430228e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11225
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.3125,
      "epoch": 0.5424236567452648,
      "grad_norm": 0.0,
      "learning_rate": 2.288123308851952e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11226
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.53125,
      "epoch": 0.54247197526092,
      "grad_norm": 0.0,
      "learning_rate": 2.2878817162736761e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11227
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.25,
      "epoch": 0.5425202937765752,
      "grad_norm": 0.0,
      "learning_rate": 2.2876401236954e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11228
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.3125,
      "epoch": 0.5425686122922304,
      "grad_norm": 0.0,
      "learning_rate": 2.287398531117124e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11229
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.3125,
      "epoch": 0.5426169308078855,
      "grad_norm": 0.0,
      "learning_rate": 2.287156938538848e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11230
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.96875,
      "epoch": 0.5426652493235408,
      "grad_norm": 0.0,
      "learning_rate": 2.286915345960572e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11231
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.21875,
      "epoch": 0.542713567839196,
      "grad_norm": 0.0,
      "learning_rate": 2.286673753382296e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11232
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.875,
      "epoch": 0.5427618863548512,
      "grad_norm": 0.0,
      "learning_rate": 2.28643216080402e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11233
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.59375,
      "epoch": 0.5428102048705064,
      "grad_norm": 0.0,
      "learning_rate": 2.286190568225744e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11234
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.375,
      "epoch": 0.5428585233861616,
      "grad_norm": 0.0,
      "learning_rate": 2.2859489756474682e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 11235
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.9375,
      "epoch": 0.5429068419018168,
      "grad_norm": 0.0,
      "learning_rate": 2.285707383069192e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11236
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.25,
      "epoch": 0.5429551604174719,
      "grad_norm": 0.0,
      "learning_rate": 2.2854657904909162e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11237
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.4375,
      "epoch": 0.5430034789331272,
      "grad_norm": 0.0,
      "learning_rate": 2.28522419791264e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11238
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.9375,
      "epoch": 0.5430517974487824,
      "grad_norm": 0.0,
      "learning_rate": 2.284982605334364e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11239
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.375,
      "epoch": 0.5431001159644375,
      "grad_norm": 0.0,
      "learning_rate": 2.284741012756088e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11240
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.75,
      "epoch": 0.5431484344800928,
      "grad_norm": 0.0,
      "learning_rate": 2.284499420177812e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11241
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.9375,
      "epoch": 0.543196752995748,
      "grad_norm": 0.0,
      "learning_rate": 2.284257827599536e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11242
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.78125,
      "epoch": 0.5432450715114032,
      "grad_norm": 0.0,
      "learning_rate": 2.28401623502126e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.2041158676147461,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11243
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.25,
      "epoch": 0.5432933900270583,
      "grad_norm": 0.0,
      "learning_rate": 2.283774642442984e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11244
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.0625,
      "epoch": 0.5433417085427136,
      "grad_norm": 0.0,
      "learning_rate": 2.2835330498647082e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11245
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.875,
      "epoch": 0.5433900270583688,
      "grad_norm": 0.0,
      "learning_rate": 2.283291457286432e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11246
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.46875,
      "epoch": 0.5434383455740239,
      "grad_norm": 0.0,
      "learning_rate": 2.2830498647081562e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11247
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.75,
      "epoch": 0.5434866640896792,
      "grad_norm": 0.0,
      "learning_rate": 2.28280827212988e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11248
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.21875,
      "epoch": 0.5435349826053344,
      "grad_norm": 0.0,
      "learning_rate": 2.2825666795516042e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11249
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.8125,
      "epoch": 0.5435833011209895,
      "grad_norm": 0.0,
      "learning_rate": 2.282325086973328e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11250
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.15625,
      "epoch": 0.5436316196366447,
      "grad_norm": 0.0,
      "learning_rate": 2.282083494395052e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11251
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.9375,
      "epoch": 0.5436799381523,
      "grad_norm": 0.0,
      "learning_rate": 2.281841901816776e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11252
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.375,
      "epoch": 0.5437282566679552,
      "grad_norm": 0.0,
      "learning_rate": 2.2816003092385e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11253
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.1875,
      "epoch": 0.5437765751836103,
      "grad_norm": 0.0,
      "learning_rate": 2.281358716660224e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11254
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.21875,
      "epoch": 0.5438248936992656,
      "grad_norm": 0.0,
      "learning_rate": 2.2811171240819482e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11255
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.15625,
      "epoch": 0.5438732122149208,
      "grad_norm": 0.0,
      "learning_rate": 2.280875531503672e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11256
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.03125,
      "epoch": 0.5439215307305759,
      "grad_norm": 0.0,
      "learning_rate": 2.2806339389253963e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11257
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.875,
      "epoch": 0.5439698492462312,
      "grad_norm": 0.0,
      "learning_rate": 2.28039234634712e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11258
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.5,
      "epoch": 0.5440181677618864,
      "grad_norm": 0.0,
      "learning_rate": 2.2801507537688443e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11259
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.15625,
      "epoch": 0.5440664862775415,
      "grad_norm": 0.0,
      "learning_rate": 2.279909161190568e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11260
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.09375,
      "epoch": 0.5441148047931967,
      "grad_norm": 0.0,
      "learning_rate": 2.279667568612292e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11261
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.96875,
      "epoch": 0.544163123308852,
      "grad_norm": 0.0,
      "learning_rate": 2.2794259760340161e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11262
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 95.09375,
      "epoch": 0.5442114418245072,
      "grad_norm": 47.10689361953826,
      "learning_rate": 2.27918438345574e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.18857255578041077,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11263
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.90625,
      "epoch": 0.5442597603401623,
      "grad_norm": 0.0,
      "learning_rate": 2.2789427908774641e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11264
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.0,
      "epoch": 0.5443080788558176,
      "grad_norm": 0.0,
      "learning_rate": 2.2787011982991883e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11265
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.15625,
      "epoch": 0.5443563973714728,
      "grad_norm": 0.0,
      "learning_rate": 2.2784596057209122e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11266
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.4375,
      "epoch": 0.5444047158871279,
      "grad_norm": 0.0,
      "learning_rate": 2.2782180131426363e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11267
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.84375,
      "epoch": 0.5444530344027831,
      "grad_norm": 0.0,
      "learning_rate": 2.2779764205643602e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11268
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.96875,
      "epoch": 0.5445013529184384,
      "grad_norm": 0.0,
      "learning_rate": 2.2777348279860843e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11269
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.5625,
      "epoch": 0.5445496714340935,
      "grad_norm": 0.0,
      "learning_rate": 2.2774932354078082e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11270
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.40625,
      "epoch": 0.5445979899497487,
      "grad_norm": 0.0,
      "learning_rate": 2.277251642829532e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11271
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.40625,
      "epoch": 0.544646308465404,
      "grad_norm": 0.0,
      "learning_rate": 2.2770100502512562e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11272
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.3125,
      "epoch": 0.5446946269810592,
      "grad_norm": 0.0,
      "learning_rate": 2.27676845767298e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11273
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.625,
      "epoch": 0.5447429454967143,
      "grad_norm": 0.0,
      "learning_rate": 2.2765268650947042e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11274
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.71875,
      "epoch": 0.5447912640123695,
      "grad_norm": 0.0,
      "learning_rate": 2.2762852725164283e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11275
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.375,
      "epoch": 0.5448395825280248,
      "grad_norm": 0.0,
      "learning_rate": 2.2760436799381522e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11276
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.34375,
      "epoch": 0.5448879010436799,
      "grad_norm": 0.0,
      "learning_rate": 2.2758020873598763e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11277
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.375,
      "epoch": 0.5449362195593351,
      "grad_norm": 0.0,
      "learning_rate": 2.2755604947816002e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11278
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.4375,
      "epoch": 0.5449845380749904,
      "grad_norm": 0.0,
      "learning_rate": 2.2753189022033243e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11279
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.125,
      "epoch": 0.5450328565906455,
      "grad_norm": 0.0,
      "learning_rate": 2.2750773096250485e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11280
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.96875,
      "epoch": 0.5450811751063007,
      "grad_norm": 0.0,
      "learning_rate": 2.274835717046772e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11281
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.40625,
      "epoch": 0.545129493621956,
      "grad_norm": 0.0,
      "learning_rate": 2.2745941244684962e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11282
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.65625,
      "epoch": 0.5451778121376112,
      "grad_norm": 0.0,
      "learning_rate": 2.27435253189022e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11283
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.0,
      "epoch": 0.5452261306532663,
      "grad_norm": 0.0,
      "learning_rate": 2.2741109393119442e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11284
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.75,
      "epoch": 0.5452744491689215,
      "grad_norm": 0.0,
      "learning_rate": 2.2738693467336684e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11285
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.25,
      "epoch": 0.5453227676845768,
      "grad_norm": 0.0,
      "learning_rate": 2.2736277541553922e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11286
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.34375,
      "epoch": 0.5453710862002319,
      "grad_norm": 0.0,
      "learning_rate": 2.2733861615771164e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11287
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.5,
      "epoch": 0.5454194047158871,
      "grad_norm": 0.0,
      "learning_rate": 2.2731445689988402e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11288
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.8125,
      "epoch": 0.5454677232315424,
      "grad_norm": 0.0,
      "learning_rate": 2.2729029764205644e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11289
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.59375,
      "epoch": 0.5455160417471975,
      "grad_norm": 0.0,
      "learning_rate": 2.2726613838422885e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11290
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.4375,
      "epoch": 0.5455643602628527,
      "grad_norm": 0.0,
      "learning_rate": 2.272419791264012e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11291
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.75,
      "epoch": 0.5456126787785079,
      "grad_norm": 0.0,
      "learning_rate": 2.2721781986857363e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11292
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.28125,
      "epoch": 0.5456609972941632,
      "grad_norm": 0.0,
      "learning_rate": 2.27193660610746e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11293
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.84375,
      "epoch": 0.5457093158098183,
      "grad_norm": 0.0,
      "learning_rate": 2.2716950135291843e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11294
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.90625,
      "epoch": 0.5457576343254735,
      "grad_norm": 0.0,
      "learning_rate": 2.2714534209509084e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11295
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.59375,
      "epoch": 0.5458059528411288,
      "grad_norm": 0.0,
      "learning_rate": 2.2712118283726323e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11296
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.25,
      "epoch": 0.5458542713567839,
      "grad_norm": 0.0,
      "learning_rate": 2.2709702357943564e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11297
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.21875,
      "epoch": 0.5459025898724391,
      "grad_norm": 0.0,
      "learning_rate": 2.2707286432160803e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11298
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.21875,
      "epoch": 0.5459509083880943,
      "grad_norm": 0.0,
      "learning_rate": 2.2704870506378044e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11299
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.5625,
      "epoch": 0.5459992269037495,
      "grad_norm": 0.0,
      "learning_rate": 2.2702454580595286e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24251842498779297,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11300
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.5,
      "epoch": 0.5460475454194047,
      "grad_norm": 0.0,
      "learning_rate": 2.2700038654812522e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11301
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.125,
      "epoch": 0.5460958639350599,
      "grad_norm": 0.0,
      "learning_rate": 2.2697622729029763e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 11302
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.125,
      "epoch": 0.5461441824507152,
      "grad_norm": 0.0,
      "learning_rate": 2.2695206803247002e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11303
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.21875,
      "epoch": 0.5461925009663703,
      "grad_norm": 0.0,
      "learning_rate": 2.2692790877464243e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11304
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.0,
      "epoch": 0.5462408194820255,
      "grad_norm": 0.0,
      "learning_rate": 2.2690374951681484e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 11305
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.03125,
      "epoch": 0.5462891379976808,
      "grad_norm": 0.0,
      "learning_rate": 2.2687959025898723e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11306
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.59375,
      "epoch": 0.5463374565133359,
      "grad_norm": 0.0,
      "learning_rate": 2.2685543100115964e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11307
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 85.71875,
      "epoch": 0.5463857750289911,
      "grad_norm": 11.828810315107054,
      "learning_rate": 2.2683127174333203e-07,
      "loss": -0.0,
      "reward": 2.359375,
      "reward_std": 0.27883464097976685,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11308
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.0625,
      "epoch": 0.5464340935446463,
      "grad_norm": 0.0,
      "learning_rate": 2.2680711248550445e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11309
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.59375,
      "epoch": 0.5464824120603015,
      "grad_norm": 0.0,
      "learning_rate": 2.2678295322767686e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11310
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.09375,
      "epoch": 0.5465307305759567,
      "grad_norm": 0.0,
      "learning_rate": 2.2675879396984922e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11311
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.5625,
      "epoch": 0.5465790490916119,
      "grad_norm": 0.0,
      "learning_rate": 2.2673463471202163e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11312
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.78125,
      "epoch": 0.5466273676072672,
      "grad_norm": 0.0,
      "learning_rate": 2.2671047545419402e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11313
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.96875,
      "epoch": 0.5466756861229223,
      "grad_norm": 0.0,
      "learning_rate": 2.2668631619636643e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11314
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.09375,
      "epoch": 0.5467240046385775,
      "grad_norm": 0.0,
      "learning_rate": 2.2666215693853885e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11315
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.84375,
      "epoch": 0.5467723231542327,
      "grad_norm": 0.0,
      "learning_rate": 2.2663799768071124e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11316
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.5625,
      "epoch": 0.5468206416698879,
      "grad_norm": 0.0,
      "learning_rate": 2.2661383842288365e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11317
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.09375,
      "epoch": 0.5468689601855431,
      "grad_norm": 0.0,
      "learning_rate": 2.2658967916505604e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11318
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.34375,
      "epoch": 0.5469172787011983,
      "grad_norm": 0.0,
      "learning_rate": 2.2656551990722845e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11319
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.875,
      "epoch": 0.5469655972168534,
      "grad_norm": 0.0,
      "learning_rate": 2.2654136064940086e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11320
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.3125,
      "epoch": 0.5470139157325087,
      "grad_norm": 0.0,
      "learning_rate": 2.2651720139157322e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11321
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.28125,
      "epoch": 0.5470622342481639,
      "grad_norm": 0.0,
      "learning_rate": 2.2649304213374564e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11322
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.90625,
      "epoch": 0.5471105527638191,
      "grad_norm": 0.0,
      "learning_rate": 2.2646888287591802e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11323
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.1875,
      "epoch": 0.5471588712794743,
      "grad_norm": 0.0,
      "learning_rate": 2.2644472361809044e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11324
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.28125,
      "epoch": 0.5472071897951295,
      "grad_norm": 0.0,
      "learning_rate": 2.2642056436026285e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11325
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.875,
      "epoch": 0.5472555083107847,
      "grad_norm": 0.0,
      "learning_rate": 2.2639640510243524e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11326
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.375,
      "epoch": 0.5473038268264399,
      "grad_norm": 0.0,
      "learning_rate": 2.2637224584460765e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11327
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.34375,
      "epoch": 0.5473521453420951,
      "grad_norm": 0.0,
      "learning_rate": 2.2634808658678004e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11328
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.09375,
      "epoch": 0.5474004638577503,
      "grad_norm": 0.0,
      "learning_rate": 2.2632392732895245e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11329
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.84375,
      "epoch": 0.5474487823734054,
      "grad_norm": 0.0,
      "learning_rate": 2.2629976807112487e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11330
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.5,
      "epoch": 0.5474971008890607,
      "grad_norm": 0.0,
      "learning_rate": 2.2627560881329725e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11331
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.71875,
      "epoch": 0.5475454194047159,
      "grad_norm": 0.0,
      "learning_rate": 2.2625144955546964e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11332
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.125,
      "epoch": 0.5475937379203711,
      "grad_norm": 0.0,
      "learning_rate": 2.2622729029764203e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11333
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.84375,
      "epoch": 0.5476420564360263,
      "grad_norm": 0.0,
      "learning_rate": 2.2620313103981444e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11334
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.40625,
      "epoch": 0.5476903749516815,
      "grad_norm": 0.0,
      "learning_rate": 2.2617897178198686e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11335
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.03125,
      "epoch": 0.5477386934673367,
      "grad_norm": 0.0,
      "learning_rate": 2.2615481252415924e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11336
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.25,
      "epoch": 0.5477870119829918,
      "grad_norm": 0.0,
      "learning_rate": 2.2613065326633166e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11337
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.25,
      "epoch": 0.5478353304986471,
      "grad_norm": 0.0,
      "learning_rate": 2.2610649400850404e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11338
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.0,
      "epoch": 0.5478836490143023,
      "grad_norm": 0.0,
      "learning_rate": 2.2608233475067646e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11339
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.5,
      "epoch": 0.5479319675299574,
      "grad_norm": 0.0,
      "learning_rate": 2.2605817549284887e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11340
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.96875,
      "epoch": 0.5479802860456127,
      "grad_norm": 0.0,
      "learning_rate": 2.2603401623502126e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11341
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.75,
      "epoch": 0.5480286045612679,
      "grad_norm": 0.0,
      "learning_rate": 2.2600985697719365e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11342
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.34375,
      "epoch": 0.5480769230769231,
      "grad_norm": 0.0,
      "learning_rate": 2.2598569771936603e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.260896235704422,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11343
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.25,
      "epoch": 0.5481252415925782,
      "grad_norm": 0.0,
      "learning_rate": 2.2596153846153845e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11344
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.90625,
      "epoch": 0.5481735601082335,
      "grad_norm": 0.0,
      "learning_rate": 2.2593737920371086e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11345
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.1875,
      "epoch": 0.5482218786238887,
      "grad_norm": 0.0,
      "learning_rate": 2.2591321994588325e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11346
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.71875,
      "epoch": 0.5482701971395438,
      "grad_norm": 0.0,
      "learning_rate": 2.2588906068805566e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11347
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.8125,
      "epoch": 0.5483185156551991,
      "grad_norm": 0.0,
      "learning_rate": 2.2586490143022805e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11348
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.96875,
      "epoch": 0.5483668341708543,
      "grad_norm": 0.0,
      "learning_rate": 2.2584074217240046e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11349
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.09375,
      "epoch": 0.5484151526865094,
      "grad_norm": 0.0,
      "learning_rate": 2.2581658291457288e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11350
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.9375,
      "epoch": 0.5484634712021647,
      "grad_norm": 0.0,
      "learning_rate": 2.2579242365674526e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11351
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.03125,
      "epoch": 0.5485117897178199,
      "grad_norm": 0.0,
      "learning_rate": 2.2576826439891768e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11352
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.84375,
      "epoch": 0.5485601082334751,
      "grad_norm": 0.0,
      "learning_rate": 2.2574410514109004e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11353
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.625,
      "epoch": 0.5486084267491302,
      "grad_norm": 0.0,
      "learning_rate": 2.2571994588326245e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11354
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.90625,
      "epoch": 0.5486567452647855,
      "grad_norm": 0.0,
      "learning_rate": 2.2569578662543486e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11355
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.21875,
      "epoch": 0.5487050637804407,
      "grad_norm": 0.0,
      "learning_rate": 2.2567162736760725e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11356
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.9375,
      "epoch": 0.5487533822960958,
      "grad_norm": 0.0,
      "learning_rate": 2.2564746810977966e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11357
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.75,
      "epoch": 0.5488017008117511,
      "grad_norm": 0.0,
      "learning_rate": 2.2562330885195205e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11358
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.0625,
      "epoch": 0.5488500193274063,
      "grad_norm": 0.0,
      "learning_rate": 2.2559914959412447e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11359
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.75,
      "epoch": 0.5488983378430614,
      "grad_norm": 0.0,
      "learning_rate": 2.2557499033629688e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 11360
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.8125,
      "epoch": 0.5489466563587166,
      "grad_norm": 0.0,
      "learning_rate": 2.2555083107846927e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11361
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.40625,
      "epoch": 0.5489949748743719,
      "grad_norm": 0.0,
      "learning_rate": 2.2552667182064168e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11362
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.5625,
      "epoch": 0.5490432933900271,
      "grad_norm": 0.0,
      "learning_rate": 2.2550251256281404e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11363
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.03125,
      "epoch": 0.5490916119056822,
      "grad_norm": 0.0,
      "learning_rate": 2.2547835330498645e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11364
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.625,
      "epoch": 0.5491399304213375,
      "grad_norm": 0.0,
      "learning_rate": 2.2545419404715887e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11365
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.5,
      "epoch": 0.5491882489369927,
      "grad_norm": 0.0,
      "learning_rate": 2.2543003478933125e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11366
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.625,
      "epoch": 0.5492365674526478,
      "grad_norm": 0.0,
      "learning_rate": 2.2540587553150367e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11367
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.28125,
      "epoch": 0.549284885968303,
      "grad_norm": 0.0,
      "learning_rate": 2.2538171627367606e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11368
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.09375,
      "epoch": 0.5493332044839583,
      "grad_norm": 0.0,
      "learning_rate": 2.2535755701584847e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11369
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.5625,
      "epoch": 0.5493815229996134,
      "grad_norm": 0.0,
      "learning_rate": 2.2533339775802088e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11370
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.0,
      "epoch": 0.5494298415152686,
      "grad_norm": 0.0,
      "learning_rate": 2.2530923850019327e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11371
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.1875,
      "epoch": 0.5494781600309239,
      "grad_norm": 0.0,
      "learning_rate": 2.2528507924236568e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11372
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.03125,
      "epoch": 0.5495264785465791,
      "grad_norm": 0.0,
      "learning_rate": 2.2526091998453804e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11373
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.9375,
      "epoch": 0.5495747970622342,
      "grad_norm": 0.0,
      "learning_rate": 2.2523676072671046e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11374
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.75,
      "epoch": 0.5496231155778895,
      "grad_norm": 0.0,
      "learning_rate": 2.2521260146888287e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11375
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.0,
      "epoch": 0.5496714340935447,
      "grad_norm": 0.0,
      "learning_rate": 2.2518844221105526e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11376
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.3125,
      "epoch": 0.5497197526091998,
      "grad_norm": 0.0,
      "learning_rate": 2.2516428295322767e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11377
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.53125,
      "epoch": 0.549768071124855,
      "grad_norm": 0.0,
      "learning_rate": 2.2514012369540006e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11378
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.5625,
      "epoch": 0.5498163896405103,
      "grad_norm": 0.0,
      "learning_rate": 2.2511596443757247e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11379
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.375,
      "epoch": 0.5498647081561654,
      "grad_norm": 45.5138525948815,
      "learning_rate": 2.250918051797449e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.21528217196464539,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11380
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.3125,
      "epoch": 0.5499130266718206,
      "grad_norm": 0.0,
      "learning_rate": 2.2506764592191727e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11381
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.90625,
      "epoch": 0.5499613451874759,
      "grad_norm": 0.0,
      "learning_rate": 2.250434866640897e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11382
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.90625,
      "epoch": 0.5500096637031311,
      "grad_norm": 0.0,
      "learning_rate": 2.2501932740626205e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11383
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.65625,
      "epoch": 0.5500579822187862,
      "grad_norm": 0.0,
      "learning_rate": 2.2499516814843446e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11384
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.78125,
      "epoch": 0.5501063007344414,
      "grad_norm": 0.0,
      "learning_rate": 2.2497100889060688e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11385
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.09375,
      "epoch": 0.5501546192500967,
      "grad_norm": 0.0,
      "learning_rate": 2.2494684963277926e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11386
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.375,
      "epoch": 0.5502029377657518,
      "grad_norm": 0.0,
      "learning_rate": 2.2492269037495168e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11387
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.78125,
      "epoch": 0.550251256281407,
      "grad_norm": 0.0,
      "learning_rate": 2.2489853111712406e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11388
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.9375,
      "epoch": 0.5502995747970623,
      "grad_norm": 0.0,
      "learning_rate": 2.2487437185929648e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11389
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.125,
      "epoch": 0.5503478933127174,
      "grad_norm": 0.0,
      "learning_rate": 2.248502126014689e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11390
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.21875,
      "epoch": 0.5503962118283726,
      "grad_norm": 0.0,
      "learning_rate": 2.2482605334364128e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11391
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.09375,
      "epoch": 0.5504445303440278,
      "grad_norm": 0.0,
      "learning_rate": 2.248018940858137e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11392
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.28125,
      "epoch": 0.5504928488596831,
      "grad_norm": 0.0,
      "learning_rate": 2.2477773482798605e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11393
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.15625,
      "epoch": 0.5505411673753382,
      "grad_norm": 0.0,
      "learning_rate": 2.2475357557015847e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11394
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.75,
      "epoch": 0.5505894858909934,
      "grad_norm": 0.0,
      "learning_rate": 2.2472941631233088e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11395
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.21875,
      "epoch": 0.5506378044066487,
      "grad_norm": 0.0,
      "learning_rate": 2.2470525705450327e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11396
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.78125,
      "epoch": 0.5506861229223038,
      "grad_norm": 0.0,
      "learning_rate": 2.2468109779667568e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11397
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.875,
      "epoch": 0.550734441437959,
      "grad_norm": 0.0,
      "learning_rate": 2.2465693853884807e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11398
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.6875,
      "epoch": 0.5507827599536143,
      "grad_norm": 0.0,
      "learning_rate": 2.2463277928102048e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11399
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.3125,
      "epoch": 0.5508310784692694,
      "grad_norm": 0.0,
      "learning_rate": 2.246086200231929e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11400
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.4375,
      "epoch": 0.5508793969849246,
      "grad_norm": 0.0,
      "learning_rate": 2.2458446076536528e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 11401
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.6875,
      "epoch": 0.5509277155005798,
      "grad_norm": 0.0,
      "learning_rate": 2.245603015075377e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11402
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.5625,
      "epoch": 0.5509760340162351,
      "grad_norm": 0.0,
      "learning_rate": 2.2453614224971008e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11403
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.4375,
      "epoch": 0.5510243525318902,
      "grad_norm": 0.0,
      "learning_rate": 2.2451198299188247e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11404
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.53125,
      "epoch": 0.5510726710475454,
      "grad_norm": 0.0,
      "learning_rate": 2.2448782373405488e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11405
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.875,
      "epoch": 0.5511209895632007,
      "grad_norm": 0.0,
      "learning_rate": 2.2446366447622727e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11406
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.875,
      "epoch": 0.5511693080788558,
      "grad_norm": 0.0,
      "learning_rate": 2.2443950521839968e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11407
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.25,
      "epoch": 0.551217626594511,
      "grad_norm": 0.0,
      "learning_rate": 2.2441534596057207e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11408
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.03125,
      "epoch": 0.5512659451101662,
      "grad_norm": 0.0,
      "learning_rate": 2.2439118670274449e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11409
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.65625,
      "epoch": 0.5513142636258214,
      "grad_norm": 0.0,
      "learning_rate": 2.243670274449169e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11410
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.25,
      "epoch": 0.5513625821414766,
      "grad_norm": 0.0,
      "learning_rate": 2.2434286818708929e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11411
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.0,
      "epoch": 0.5514109006571318,
      "grad_norm": 0.0,
      "learning_rate": 2.243187089292617e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11412
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.53125,
      "epoch": 0.5514592191727871,
      "grad_norm": 0.0,
      "learning_rate": 2.2429454967143409e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11413
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.90625,
      "epoch": 0.5515075376884422,
      "grad_norm": 0.0,
      "learning_rate": 2.2427039041360647e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11414
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.75,
      "epoch": 0.5515558562040974,
      "grad_norm": 0.0,
      "learning_rate": 2.242462311557789e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11415
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.09375,
      "epoch": 0.5516041747197526,
      "grad_norm": 0.0,
      "learning_rate": 2.2422207189795127e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11416
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.28125,
      "epoch": 0.5516524932354078,
      "grad_norm": 0.0,
      "learning_rate": 2.241979126401237e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11417
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.59375,
      "epoch": 0.551700811751063,
      "grad_norm": 0.0,
      "learning_rate": 2.2417375338229608e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11418
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.0625,
      "epoch": 0.5517491302667182,
      "grad_norm": 0.0,
      "learning_rate": 2.241495941244685e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11419
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 85.625,
      "epoch": 0.5517974487823734,
      "grad_norm": 15.86740252569326,
      "learning_rate": 2.241254348666409e-07,
      "loss": -0.0,
      "reward": 1.796875,
      "reward_std": 0.23685522377490997,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11420
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.125,
      "epoch": 0.5518457672980286,
      "grad_norm": 0.0,
      "learning_rate": 2.241012756088133e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11421
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.34375,
      "epoch": 0.5518940858136838,
      "grad_norm": 0.0,
      "learning_rate": 2.240771163509857e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11422
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.78125,
      "epoch": 0.551942404329339,
      "grad_norm": 0.0,
      "learning_rate": 2.240529570931581e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11423
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.625,
      "epoch": 0.5519907228449942,
      "grad_norm": 0.0,
      "learning_rate": 2.2402879783533048e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11424
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.59375,
      "epoch": 0.5520390413606494,
      "grad_norm": 0.0,
      "learning_rate": 2.240046385775029e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11425
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.4375,
      "epoch": 0.5520873598763046,
      "grad_norm": 0.0,
      "learning_rate": 2.2398047931967528e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11426
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.5,
      "epoch": 0.5521356783919598,
      "grad_norm": 0.0,
      "learning_rate": 2.239563200618477e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11427
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.09375,
      "epoch": 0.552183996907615,
      "grad_norm": 0.0,
      "learning_rate": 2.2393216080402008e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11428
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.09375,
      "epoch": 0.5522323154232702,
      "grad_norm": 0.0,
      "learning_rate": 2.239080015461925e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11429
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.5,
      "epoch": 0.5522806339389253,
      "grad_norm": 0.0,
      "learning_rate": 2.238838422883649e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11430
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.34375,
      "epoch": 0.5523289524545806,
      "grad_norm": 0.0,
      "learning_rate": 2.238596830305373e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11431
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.96875,
      "epoch": 0.5523772709702358,
      "grad_norm": 0.0,
      "learning_rate": 2.238355237727097e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11432
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.96875,
      "epoch": 0.552425589485891,
      "grad_norm": 0.0,
      "learning_rate": 2.238113645148821e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11433
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.375,
      "epoch": 0.5524739080015462,
      "grad_norm": 0.0,
      "learning_rate": 2.237872052570545e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11434
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.15625,
      "epoch": 0.5525222265172014,
      "grad_norm": 0.0,
      "learning_rate": 2.237630459992269e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11435
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.15625,
      "epoch": 0.5525705450328566,
      "grad_norm": 0.0,
      "learning_rate": 2.2373888674139928e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11436
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.96875,
      "epoch": 0.5526188635485118,
      "grad_norm": 0.0,
      "learning_rate": 2.237147274835717e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11437
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.84375,
      "epoch": 0.552667182064167,
      "grad_norm": 0.0,
      "learning_rate": 2.2369056822574408e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11438
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.0,
      "epoch": 0.5527155005798222,
      "grad_norm": 0.0,
      "learning_rate": 2.236664089679165e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11439
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.28125,
      "epoch": 0.5527638190954773,
      "grad_norm": 0.0,
      "learning_rate": 2.236422497100889e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11440
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.84375,
      "epoch": 0.5528121376111326,
      "grad_norm": 0.0,
      "learning_rate": 2.236180904522613e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11441
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.75,
      "epoch": 0.5528604561267878,
      "grad_norm": 0.0,
      "learning_rate": 2.235939311944337e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11442
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.53125,
      "epoch": 0.552908774642443,
      "grad_norm": 0.0,
      "learning_rate": 2.235697719366061e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11443
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.0625,
      "epoch": 0.5529570931580982,
      "grad_norm": 0.0,
      "learning_rate": 2.235456126787785e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11444
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.59375,
      "epoch": 0.5530054116737534,
      "grad_norm": 0.0,
      "learning_rate": 2.235214534209509e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11445
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.84375,
      "epoch": 0.5530537301894086,
      "grad_norm": 0.0,
      "learning_rate": 2.2349729416312329e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 11446
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.125,
      "epoch": 0.5531020487050637,
      "grad_norm": 0.0,
      "learning_rate": 2.234731349052957e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11447
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.8125,
      "epoch": 0.553150367220719,
      "grad_norm": 0.0,
      "learning_rate": 2.234489756474681e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11448
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.5625,
      "epoch": 0.5531986857363742,
      "grad_norm": 0.0,
      "learning_rate": 2.234248163896405e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11449
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.5625,
      "epoch": 0.5532470042520293,
      "grad_norm": 0.0,
      "learning_rate": 2.2340065713181291e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11450
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.0,
      "epoch": 0.5532953227676846,
      "grad_norm": 0.0,
      "learning_rate": 2.233764978739853e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11451
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.03125,
      "epoch": 0.5533436412833398,
      "grad_norm": 0.0,
      "learning_rate": 2.2335233861615772e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11452
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.15625,
      "epoch": 0.553391959798995,
      "grad_norm": 0.0,
      "learning_rate": 2.233281793583301e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11453
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.1875,
      "epoch": 0.5534402783146501,
      "grad_norm": 0.0,
      "learning_rate": 2.2330402010050252e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11454
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.28125,
      "epoch": 0.5534885968303054,
      "grad_norm": 0.0,
      "learning_rate": 2.2327986084267493e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11455
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.8125,
      "epoch": 0.5535369153459606,
      "grad_norm": 0.0,
      "learning_rate": 2.232557015848473e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2130674123764038,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11456
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.90625,
      "epoch": 0.5535852338616157,
      "grad_norm": 0.0,
      "learning_rate": 2.232315423270197e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11457
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.9375,
      "epoch": 0.553633552377271,
      "grad_norm": 0.0,
      "learning_rate": 2.232073830691921e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11458
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.3125,
      "epoch": 0.5536818708929262,
      "grad_norm": 0.0,
      "learning_rate": 2.231832238113645e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11459
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.5,
      "epoch": 0.5537301894085813,
      "grad_norm": 0.0,
      "learning_rate": 2.2315906455353692e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11460
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.84375,
      "epoch": 0.5537785079242366,
      "grad_norm": 0.0,
      "learning_rate": 2.231349052957093e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11461
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.03125,
      "epoch": 0.5538268264398918,
      "grad_norm": 0.0,
      "learning_rate": 2.2311074603788172e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11462
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.25,
      "epoch": 0.553875144955547,
      "grad_norm": 0.0,
      "learning_rate": 2.230865867800541e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11463
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.875,
      "epoch": 0.5539234634712021,
      "grad_norm": 0.0,
      "learning_rate": 2.2306242752222652e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11464
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.3125,
      "epoch": 0.5539717819868574,
      "grad_norm": 0.0,
      "learning_rate": 2.2303826826439893e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11465
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.1875,
      "epoch": 0.5540201005025126,
      "grad_norm": 0.0,
      "learning_rate": 2.230141090065713e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11466
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 94.1875,
      "epoch": 0.5540684190181677,
      "grad_norm": 52.090292215070036,
      "learning_rate": 2.229899497487437e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11467
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.4375,
      "epoch": 0.554116737533823,
      "grad_norm": 0.0,
      "learning_rate": 2.229657904909161e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11468
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.25,
      "epoch": 0.5541650560494782,
      "grad_norm": 0.0,
      "learning_rate": 2.229416312330885e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11469
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.28125,
      "epoch": 0.5542133745651333,
      "grad_norm": 0.0,
      "learning_rate": 2.2291747197526092e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11470
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.5625,
      "epoch": 0.5542616930807885,
      "grad_norm": 0.0,
      "learning_rate": 2.228933127174333e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11471
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.5625,
      "epoch": 0.5543100115964438,
      "grad_norm": 0.0,
      "learning_rate": 2.2286915345960572e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11472
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.0,
      "epoch": 0.554358330112099,
      "grad_norm": 0.0,
      "learning_rate": 2.228449942017781e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11473
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.25,
      "epoch": 0.5544066486277541,
      "grad_norm": 0.0,
      "learning_rate": 2.2282083494395052e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11474
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.4375,
      "epoch": 0.5544549671434094,
      "grad_norm": 0.0,
      "learning_rate": 2.2279667568612294e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11475
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.09375,
      "epoch": 0.5545032856590646,
      "grad_norm": 0.0,
      "learning_rate": 2.227725164282953e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11476
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.21875,
      "epoch": 0.5545516041747197,
      "grad_norm": 0.0,
      "learning_rate": 2.227483571704677e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11477
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.875,
      "epoch": 0.554599922690375,
      "grad_norm": 0.0,
      "learning_rate": 2.227241979126401e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11478
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.46875,
      "epoch": 0.5546482412060302,
      "grad_norm": 10.321045459676256,
      "learning_rate": 2.227000386548125e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.3029785752296448,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11479
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.75,
      "epoch": 0.5546965597216853,
      "grad_norm": 0.0,
      "learning_rate": 2.2267587939698493e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11480
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 94.03125,
      "epoch": 0.5547448782373405,
      "grad_norm": 11.353426978934364,
      "learning_rate": 2.2265172013915731e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11481
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.78125,
      "epoch": 0.5547931967529958,
      "grad_norm": 0.0,
      "learning_rate": 2.2262756088132973e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11482
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.75,
      "epoch": 0.554841515268651,
      "grad_norm": 0.0,
      "learning_rate": 2.2260340162350211e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11483
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.09375,
      "epoch": 0.5548898337843061,
      "grad_norm": 0.0,
      "learning_rate": 2.2257924236567453e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11484
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 85.71875,
      "epoch": 0.5549381522999614,
      "grad_norm": 0.0,
      "learning_rate": 2.2255508310784694e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 0.96875,
      "step": 11485
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.0,
      "epoch": 0.5549864708156166,
      "grad_norm": 0.0,
      "learning_rate": 2.225309238500193e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11486
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.84375,
      "epoch": 0.5550347893312717,
      "grad_norm": 0.0,
      "learning_rate": 2.2250676459219172e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11487
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.46875,
      "epoch": 0.5550831078469269,
      "grad_norm": 0.0,
      "learning_rate": 2.224826053343641e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11488
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.75,
      "epoch": 0.5551314263625822,
      "grad_norm": 0.0,
      "learning_rate": 2.2245844607653652e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11489
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 97.09375,
      "epoch": 0.5551797448782373,
      "grad_norm": 0.0,
      "learning_rate": 2.2243428681870893e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11490
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.34375,
      "epoch": 0.5552280633938925,
      "grad_norm": 0.0,
      "learning_rate": 2.2241012756088132e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11491
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.46875,
      "epoch": 0.5552763819095478,
      "grad_norm": 0.0,
      "learning_rate": 2.2238596830305373e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11492
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.15625,
      "epoch": 0.555324700425203,
      "grad_norm": 0.0,
      "learning_rate": 2.2236180904522612e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11493
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.46875,
      "epoch": 0.5553730189408581,
      "grad_norm": 0.0,
      "learning_rate": 2.2233764978739853e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11494
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.90625,
      "epoch": 0.5554213374565133,
      "grad_norm": 0.0,
      "learning_rate": 2.2231349052957095e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11495
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.625,
      "epoch": 0.5554696559721686,
      "grad_norm": 0.0,
      "learning_rate": 2.222893312717433e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11496
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.78125,
      "epoch": 0.5555179744878237,
      "grad_norm": 0.0,
      "learning_rate": 2.2226517201391572e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11497
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.8125,
      "epoch": 0.5555662930034789,
      "grad_norm": 0.0,
      "learning_rate": 2.222410127560881e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11498
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.625,
      "epoch": 0.5556146115191342,
      "grad_norm": 0.0,
      "learning_rate": 2.2221685349826052e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11499
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.65625,
      "epoch": 0.5556629300347893,
      "grad_norm": 0.0,
      "learning_rate": 2.2219269424043293e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11500
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.0,
      "epoch": 0.5557112485504445,
      "grad_norm": 0.0,
      "learning_rate": 2.2216853498260532e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11501
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.9375,
      "epoch": 0.5557595670660997,
      "grad_norm": 0.0,
      "learning_rate": 2.2214437572477773e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11502
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.34375,
      "epoch": 0.555807885581755,
      "grad_norm": 0.0,
      "learning_rate": 2.2212021646695012e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11503
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.21875,
      "epoch": 0.5558562040974101,
      "grad_norm": 0.0,
      "learning_rate": 2.2209605720912254e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11504
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.96875,
      "epoch": 0.5559045226130653,
      "grad_norm": 10.151486100732058,
      "learning_rate": 2.2207189795129495e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.3014557361602783,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11505
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.53125,
      "epoch": 0.5559528411287206,
      "grad_norm": 0.0,
      "learning_rate": 2.2204773869346734e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11506
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 89.59375,
      "epoch": 0.5560011596443757,
      "grad_norm": 66.5069796662709,
      "learning_rate": 2.2202357943563972e-07,
      "loss": 0.0,
      "reward": 2.1875,
      "reward_std": 0.28566449880599976,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11507
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.71875,
      "epoch": 0.5560494781600309,
      "grad_norm": 0.0,
      "learning_rate": 2.219994201778121e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11508
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.0,
      "epoch": 0.5560977966756862,
      "grad_norm": 0.0,
      "learning_rate": 2.2197526091998452e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11509
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.40625,
      "epoch": 0.5561461151913413,
      "grad_norm": 0.0,
      "learning_rate": 2.2195110166215694e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11510
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.84375,
      "epoch": 0.5561944337069965,
      "grad_norm": 0.0,
      "learning_rate": 2.2192694240432933e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11511
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.5625,
      "epoch": 0.5562427522226517,
      "grad_norm": 0.0,
      "learning_rate": 2.2190278314650174e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11512
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.84375,
      "epoch": 0.556291070738307,
      "grad_norm": 0.0,
      "learning_rate": 2.2187862388867413e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11513
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.40625,
      "epoch": 0.5563393892539621,
      "grad_norm": 0.0,
      "learning_rate": 2.2185446463084654e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11514
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 76.53125,
      "epoch": 0.5563877077696173,
      "grad_norm": 89.20162005904363,
      "learning_rate": 2.2183030537301895e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11515
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.3125,
      "epoch": 0.5564360262852726,
      "grad_norm": 0.0,
      "learning_rate": 2.2180614611519134e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11516
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.75,
      "epoch": 0.5564843448009277,
      "grad_norm": 0.0,
      "learning_rate": 2.2178198685736373e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11517
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.65625,
      "epoch": 0.5565326633165829,
      "grad_norm": 0.0,
      "learning_rate": 2.2175782759953611e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11518
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.625,
      "epoch": 0.5565809818322381,
      "grad_norm": 0.0,
      "learning_rate": 2.2173366834170853e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11519
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.71875,
      "epoch": 0.5566293003478933,
      "grad_norm": 0.0,
      "learning_rate": 2.2170950908388094e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11520
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.375,
      "epoch": 0.5566776188635485,
      "grad_norm": 0.0,
      "learning_rate": 2.2168534982605333e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11521
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.4375,
      "epoch": 0.5567259373792037,
      "grad_norm": 0.0,
      "learning_rate": 2.2166119056822574e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11522
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.71875,
      "epoch": 0.556774255894859,
      "grad_norm": 0.0,
      "learning_rate": 2.2163703131039813e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11523
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.28125,
      "epoch": 0.5568225744105141,
      "grad_norm": 0.0,
      "learning_rate": 2.2161287205257054e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11524
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.21875,
      "epoch": 0.5568708929261693,
      "grad_norm": 0.0,
      "learning_rate": 2.2158871279474296e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11525
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.09375,
      "epoch": 0.5569192114418245,
      "grad_norm": 0.0,
      "learning_rate": 2.2156455353691534e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11526
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.875,
      "epoch": 0.5569675299574797,
      "grad_norm": 0.0,
      "learning_rate": 2.2154039427908773e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11527
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.96875,
      "epoch": 0.5570158484731349,
      "grad_norm": 0.0,
      "learning_rate": 2.2151623502126012e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11528
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.40625,
      "epoch": 0.5570641669887901,
      "grad_norm": 0.0,
      "learning_rate": 2.2149207576343253e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11529
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.1875,
      "epoch": 0.5571124855044453,
      "grad_norm": 0.0,
      "learning_rate": 2.2146791650560495e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11530
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.40625,
      "epoch": 0.5571608040201005,
      "grad_norm": 0.0,
      "learning_rate": 2.2144375724777733e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11531
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.65625,
      "epoch": 0.5572091225357557,
      "grad_norm": 0.0,
      "learning_rate": 2.2141959798994975e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11532
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.78125,
      "epoch": 0.557257441051411,
      "grad_norm": 0.0,
      "learning_rate": 2.2139543873212213e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11533
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.25,
      "epoch": 0.5573057595670661,
      "grad_norm": 0.0,
      "learning_rate": 2.2137127947429455e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11534
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.75,
      "epoch": 0.5573540780827213,
      "grad_norm": 0.0,
      "learning_rate": 2.2134712021646696e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11535
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.53125,
      "epoch": 0.5574023965983765,
      "grad_norm": 0.0,
      "learning_rate": 2.2132296095863935e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11536
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.125,
      "epoch": 0.5574507151140317,
      "grad_norm": 0.0,
      "learning_rate": 2.2129880170081176e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 11537
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.6875,
      "epoch": 0.5574990336296869,
      "grad_norm": 0.0,
      "learning_rate": 2.2127464244298412e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11538
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.90625,
      "epoch": 0.5575473521453421,
      "grad_norm": 0.0,
      "learning_rate": 2.2125048318515654e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11539
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.09375,
      "epoch": 0.5575956706609972,
      "grad_norm": 0.0,
      "learning_rate": 2.2122632392732895e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11540
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.3125,
      "epoch": 0.5576439891766525,
      "grad_norm": 0.0,
      "learning_rate": 2.2120216466950134e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11541
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.125,
      "epoch": 0.5576923076923077,
      "grad_norm": 0.0,
      "learning_rate": 2.2117800541167375e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11542
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.84375,
      "epoch": 0.5577406262079629,
      "grad_norm": 0.0,
      "learning_rate": 2.2115384615384614e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11543
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.53125,
      "epoch": 0.5577889447236181,
      "grad_norm": 0.0,
      "learning_rate": 2.2112968689601855e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11544
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 88.03125,
      "epoch": 0.5578372632392733,
      "grad_norm": 24.367108943486897,
      "learning_rate": 2.2110552763819096e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11545
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.6875,
      "epoch": 0.5578855817549285,
      "grad_norm": 0.0,
      "learning_rate": 2.2108136838036335e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11546
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.90625,
      "epoch": 0.5579339002705836,
      "grad_norm": 0.0,
      "learning_rate": 2.2105720912253577e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11547
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.0,
      "epoch": 0.5579822187862389,
      "grad_norm": 0.0,
      "learning_rate": 2.2103304986470813e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11548
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.5625,
      "epoch": 0.5580305373018941,
      "grad_norm": 0.0,
      "learning_rate": 2.2100889060688054e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11549
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.125,
      "epoch": 0.5580788558175492,
      "grad_norm": 0.0,
      "learning_rate": 2.2098473134905295e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 11550
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.15625,
      "epoch": 0.5581271743332045,
      "grad_norm": 0.0,
      "learning_rate": 2.2096057209122534e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11551
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.28125,
      "epoch": 0.5581754928488597,
      "grad_norm": 0.0,
      "learning_rate": 2.2093641283339775e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11552
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.34375,
      "epoch": 0.5582238113645149,
      "grad_norm": 44.913394085878984,
      "learning_rate": 2.2091225357557014e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11553
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.96875,
      "epoch": 0.55827212988017,
      "grad_norm": 0.0,
      "learning_rate": 2.2088809431774256e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11554
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.71875,
      "epoch": 0.5583204483958253,
      "grad_norm": 0.0,
      "learning_rate": 2.2086393505991497e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11555
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.09375,
      "epoch": 0.5583687669114805,
      "grad_norm": 0.0,
      "learning_rate": 2.2083977580208736e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11556
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.875,
      "epoch": 0.5584170854271356,
      "grad_norm": 0.0,
      "learning_rate": 2.2081561654425977e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11557
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.90625,
      "epoch": 0.5584654039427909,
      "grad_norm": 0.0,
      "learning_rate": 2.2079145728643213e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11558
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.46875,
      "epoch": 0.5585137224584461,
      "grad_norm": 0.0,
      "learning_rate": 2.2076729802860454e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11559
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.75,
      "epoch": 0.5585620409741012,
      "grad_norm": 0.0,
      "learning_rate": 2.2074313877077696e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11560
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.71875,
      "epoch": 0.5586103594897565,
      "grad_norm": 0.0,
      "learning_rate": 2.2071897951294934e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11561
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.25,
      "epoch": 0.5586586780054117,
      "grad_norm": 0.0,
      "learning_rate": 2.2069482025512176e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11562
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.21875,
      "epoch": 0.5587069965210669,
      "grad_norm": 0.0,
      "learning_rate": 2.2067066099729415e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11563
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.0,
      "epoch": 0.558755315036722,
      "grad_norm": 0.0,
      "learning_rate": 2.2064650173946656e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11564
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.96875,
      "epoch": 0.5588036335523773,
      "grad_norm": 0.0,
      "learning_rate": 2.2062234248163897e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11565
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.8125,
      "epoch": 0.5588519520680325,
      "grad_norm": 0.0,
      "learning_rate": 2.2059818322381136e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11566
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.875,
      "epoch": 0.5589002705836876,
      "grad_norm": 0.0,
      "learning_rate": 2.2057402396598377e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11567
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.0,
      "epoch": 0.5589485890993429,
      "grad_norm": 0.0,
      "learning_rate": 2.2054986470815613e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11568
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.96875,
      "epoch": 0.5589969076149981,
      "grad_norm": 0.0,
      "learning_rate": 2.2052570545032855e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11569
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.53125,
      "epoch": 0.5590452261306532,
      "grad_norm": 0.0,
      "learning_rate": 2.2050154619250096e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11570
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.625,
      "epoch": 0.5590935446463084,
      "grad_norm": 0.0,
      "learning_rate": 2.2047738693467335e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11571
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.125,
      "epoch": 0.5591418631619637,
      "grad_norm": 0.0,
      "learning_rate": 2.2045322767684576e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11572
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.03125,
      "epoch": 0.5591901816776189,
      "grad_norm": 0.0,
      "learning_rate": 2.2042906841901815e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11573
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.28125,
      "epoch": 0.559238500193274,
      "grad_norm": 0.0,
      "learning_rate": 2.2040490916119056e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11574
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.4375,
      "epoch": 0.5592868187089293,
      "grad_norm": 0.0,
      "learning_rate": 2.2038074990336298e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 11575
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.75,
      "epoch": 0.5593351372245845,
      "grad_norm": 0.0,
      "learning_rate": 2.2035659064553536e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11576
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.1875,
      "epoch": 0.5593834557402396,
      "grad_norm": 0.0,
      "learning_rate": 2.2033243138770778e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11577
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 84.09375,
      "epoch": 0.5594317742558949,
      "grad_norm": 0.0,
      "learning_rate": 2.2030827212988016e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 11578
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.09375,
      "epoch": 0.5594800927715501,
      "grad_norm": 0.0,
      "learning_rate": 2.2028411287205255e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.2041158676147461,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11579
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.3125,
      "epoch": 0.5595284112872052,
      "grad_norm": 0.0,
      "learning_rate": 2.2025995361422497e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11580
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.28125,
      "epoch": 0.5595767298028604,
      "grad_norm": 0.0,
      "learning_rate": 2.2023579435639735e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11581
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.40625,
      "epoch": 0.5596250483185157,
      "grad_norm": 0.0,
      "learning_rate": 2.2021163509856977e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11582
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.71875,
      "epoch": 0.5596733668341709,
      "grad_norm": 0.0,
      "learning_rate": 2.2018747584074215e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11583
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.375,
      "epoch": 0.559721685349826,
      "grad_norm": 0.0,
      "learning_rate": 2.2016331658291457e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11584
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.8125,
      "epoch": 0.5597700038654813,
      "grad_norm": 0.0,
      "learning_rate": 2.2013915732508698e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11585
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.875,
      "epoch": 0.5598183223811365,
      "grad_norm": 0.0,
      "learning_rate": 2.2011499806725937e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11586
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.46875,
      "epoch": 0.5598666408967916,
      "grad_norm": 0.0,
      "learning_rate": 2.2009083880943178e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11587
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.0625,
      "epoch": 0.5599149594124468,
      "grad_norm": 0.0,
      "learning_rate": 2.2006667955160417e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11588
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.15625,
      "epoch": 0.5599632779281021,
      "grad_norm": 0.0,
      "learning_rate": 2.2004252029377656e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11589
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.34375,
      "epoch": 0.5600115964437572,
      "grad_norm": 0.0,
      "learning_rate": 2.2001836103594897e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11590
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.53125,
      "epoch": 0.5600599149594124,
      "grad_norm": 0.0,
      "learning_rate": 2.1999420177812136e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11591
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.96875,
      "epoch": 0.5601082334750677,
      "grad_norm": 0.0,
      "learning_rate": 2.1997004252029377e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11592
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.34375,
      "epoch": 0.5601565519907229,
      "grad_norm": 0.0,
      "learning_rate": 2.1994588326246616e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11593
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.21875,
      "epoch": 0.560204870506378,
      "grad_norm": 0.0,
      "learning_rate": 2.1992172400463857e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11594
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.3125,
      "epoch": 0.5602531890220332,
      "grad_norm": 0.0,
      "learning_rate": 2.1989756474681098e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11595
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.71875,
      "epoch": 0.5603015075376885,
      "grad_norm": 0.0,
      "learning_rate": 2.1987340548898337e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11596
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.15625,
      "epoch": 0.5603498260533436,
      "grad_norm": 0.0,
      "learning_rate": 2.1984924623115579e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11597
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.65625,
      "epoch": 0.5603981445689988,
      "grad_norm": 0.0,
      "learning_rate": 2.1982508697332817e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11598
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.25,
      "epoch": 0.5604464630846541,
      "grad_norm": 0.0,
      "learning_rate": 2.1980092771550056e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11599
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.59375,
      "epoch": 0.5604947816003092,
      "grad_norm": 0.0,
      "learning_rate": 2.1977676845767297e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11600
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.5,
      "epoch": 0.5605431001159644,
      "grad_norm": 0.0,
      "learning_rate": 2.1975260919984536e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11601
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.28125,
      "epoch": 0.5605914186316197,
      "grad_norm": 0.0,
      "learning_rate": 2.1972844994201777e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11602
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.25,
      "epoch": 0.5606397371472749,
      "grad_norm": 0.0,
      "learning_rate": 2.1970429068419016e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11603
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.53125,
      "epoch": 0.56068805566293,
      "grad_norm": 0.0,
      "learning_rate": 2.1968013142636257e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11604
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.46875,
      "epoch": 0.5607363741785852,
      "grad_norm": 0.0,
      "learning_rate": 2.19655972168535e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 11605
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.46875,
      "epoch": 0.5607846926942405,
      "grad_norm": 0.0,
      "learning_rate": 2.1963181291070738e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11606
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.65625,
      "epoch": 0.5608330112098956,
      "grad_norm": 0.0,
      "learning_rate": 2.196076536528798e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11607
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.46875,
      "epoch": 0.5608813297255508,
      "grad_norm": 0.0,
      "learning_rate": 2.1958349439505218e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11608
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.53125,
      "epoch": 0.5609296482412061,
      "grad_norm": 0.0,
      "learning_rate": 2.195593351372246e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11609
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.0,
      "epoch": 0.5609779667568612,
      "grad_norm": 0.0,
      "learning_rate": 2.1953517587939698e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11610
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.34375,
      "epoch": 0.5610262852725164,
      "grad_norm": 0.0,
      "learning_rate": 2.1951101662156936e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11611
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.15625,
      "epoch": 0.5610746037881716,
      "grad_norm": 0.0,
      "learning_rate": 2.1948685736374178e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11612
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.8125,
      "epoch": 0.5611229223038269,
      "grad_norm": 0.0,
      "learning_rate": 2.1946269810591417e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11613
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.4375,
      "epoch": 0.561171240819482,
      "grad_norm": 0.0,
      "learning_rate": 2.1943853884808658e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11614
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.15625,
      "epoch": 0.5612195593351372,
      "grad_norm": 0.0,
      "learning_rate": 2.19414379590259e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11615
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.65625,
      "epoch": 0.5612678778507925,
      "grad_norm": 0.0,
      "learning_rate": 2.1939022033243138e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11616
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.46875,
      "epoch": 0.5613161963664476,
      "grad_norm": 0.0,
      "learning_rate": 2.193660610746038e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11617
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.1875,
      "epoch": 0.5613645148821028,
      "grad_norm": 0.0,
      "learning_rate": 2.1934190181677618e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11618
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.59375,
      "epoch": 0.561412833397758,
      "grad_norm": 0.0,
      "learning_rate": 2.193177425589486e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11619
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.46875,
      "epoch": 0.5614611519134132,
      "grad_norm": 0.0,
      "learning_rate": 2.1929358330112098e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11620
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.0,
      "epoch": 0.5615094704290684,
      "grad_norm": 0.0,
      "learning_rate": 2.1926942404329337e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 11621
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.5,
      "epoch": 0.5615577889447236,
      "grad_norm": 0.0,
      "learning_rate": 2.1924526478546578e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11622
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.78125,
      "epoch": 0.5616061074603789,
      "grad_norm": 0.0,
      "learning_rate": 2.1922110552763817e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11623
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.1875,
      "epoch": 0.561654425976034,
      "grad_norm": 0.0,
      "learning_rate": 2.1919694626981058e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11624
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.75,
      "epoch": 0.5617027444916892,
      "grad_norm": 0.0,
      "learning_rate": 2.19172787011983e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11625
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.15625,
      "epoch": 0.5617510630073445,
      "grad_norm": 0.0,
      "learning_rate": 2.1914862775415538e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11626
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.75,
      "epoch": 0.5617993815229996,
      "grad_norm": 0.0,
      "learning_rate": 2.191244684963278e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11627
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.21875,
      "epoch": 0.5618477000386548,
      "grad_norm": 0.0,
      "learning_rate": 2.1910030923850018e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11628
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.21875,
      "epoch": 0.56189601855431,
      "grad_norm": 0.0,
      "learning_rate": 2.190761499806726e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11629
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.78125,
      "epoch": 0.5619443370699652,
      "grad_norm": 0.0,
      "learning_rate": 2.1905199072284499e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11630
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.9375,
      "epoch": 0.5619926555856204,
      "grad_norm": 0.0,
      "learning_rate": 2.1902783146501737e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11631
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.5625,
      "epoch": 0.5620409741012756,
      "grad_norm": 0.0,
      "learning_rate": 2.1900367220718979e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 11632
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.65625,
      "epoch": 0.5620892926169309,
      "grad_norm": 0.0,
      "learning_rate": 2.1897951294936217e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11633
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.6875,
      "epoch": 0.562137611132586,
      "grad_norm": 0.0,
      "learning_rate": 2.1895535369153459e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11634
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.125,
      "epoch": 0.5621859296482412,
      "grad_norm": 0.0,
      "learning_rate": 2.18931194433707e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11635
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.96875,
      "epoch": 0.5622342481638964,
      "grad_norm": 0.0,
      "learning_rate": 2.189070351758794e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11636
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 97.375,
      "epoch": 0.5622825666795516,
      "grad_norm": 0.0,
      "learning_rate": 2.188828759180518e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 11637
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.0,
      "epoch": 0.5623308851952068,
      "grad_norm": 0.0,
      "learning_rate": 2.188587166602242e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11638
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.28125,
      "epoch": 0.562379203710862,
      "grad_norm": 0.0,
      "learning_rate": 2.188345574023966e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11639
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.46875,
      "epoch": 0.5624275222265172,
      "grad_norm": 0.0,
      "learning_rate": 2.1881039814456902e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11640
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.625,
      "epoch": 0.5624758407421724,
      "grad_norm": 0.0,
      "learning_rate": 2.1878623888674138e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11641
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.78125,
      "epoch": 0.5625241592578276,
      "grad_norm": 0.0,
      "learning_rate": 2.187620796289138e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11642
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.125,
      "epoch": 0.5625724777734828,
      "grad_norm": 0.0,
      "learning_rate": 2.1873792037108618e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11643
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.96875,
      "epoch": 0.562620796289138,
      "grad_norm": 0.0,
      "learning_rate": 2.187137611132586e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11644
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.6875,
      "epoch": 0.5626691148047932,
      "grad_norm": 0.0,
      "learning_rate": 2.1868960185543098e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11645
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.6875,
      "epoch": 0.5627174333204484,
      "grad_norm": 0.0,
      "learning_rate": 2.186654425976034e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11646
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.75,
      "epoch": 0.5627657518361036,
      "grad_norm": 0.0,
      "learning_rate": 2.186412833397758e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11647
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.28125,
      "epoch": 0.5628140703517588,
      "grad_norm": 0.0,
      "learning_rate": 2.186171240819482e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11648
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.875,
      "epoch": 0.562862388867414,
      "grad_norm": 0.0,
      "learning_rate": 2.185929648241206e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11649
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.6875,
      "epoch": 0.5629107073830691,
      "grad_norm": 0.0,
      "learning_rate": 2.1856880556629297e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11650
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.375,
      "epoch": 0.5629590258987244,
      "grad_norm": 0.0,
      "learning_rate": 2.1854464630846538e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11651
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.0625,
      "epoch": 0.5630073444143796,
      "grad_norm": 0.0,
      "learning_rate": 2.185204870506378e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11652
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.125,
      "epoch": 0.5630556629300348,
      "grad_norm": 0.0,
      "learning_rate": 2.1849632779281018e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11653
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.65625,
      "epoch": 0.56310398144569,
      "grad_norm": 0.0,
      "learning_rate": 2.184721685349826e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11654
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.5,
      "epoch": 0.5631522999613452,
      "grad_norm": 0.0,
      "learning_rate": 2.1844800927715498e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11655
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.4375,
      "epoch": 0.5632006184770004,
      "grad_norm": 0.0,
      "learning_rate": 2.184238500193274e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11656
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.90625,
      "epoch": 0.5632489369926555,
      "grad_norm": 0.0,
      "learning_rate": 2.183996907614998e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11657
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.25,
      "epoch": 0.5632972555083108,
      "grad_norm": 0.0,
      "learning_rate": 2.183755315036722e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11658
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.28125,
      "epoch": 0.563345574023966,
      "grad_norm": 0.0,
      "learning_rate": 2.183513722458446e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11659
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.375,
      "epoch": 0.5633938925396211,
      "grad_norm": 0.0,
      "learning_rate": 2.18327212988017e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11660
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.1875,
      "epoch": 0.5634422110552764,
      "grad_norm": 0.0,
      "learning_rate": 2.1830305373018938e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11661
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.1875,
      "epoch": 0.5634905295709316,
      "grad_norm": 0.0,
      "learning_rate": 2.182788944723618e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11662
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.96875,
      "epoch": 0.5635388480865868,
      "grad_norm": 0.0,
      "learning_rate": 2.1825473521453418e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11663
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.71875,
      "epoch": 0.563587166602242,
      "grad_norm": 0.0,
      "learning_rate": 2.182305759567066e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11664
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.875,
      "epoch": 0.5636354851178972,
      "grad_norm": 0.0,
      "learning_rate": 2.1820641669887899e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11665
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.75,
      "epoch": 0.5636838036335524,
      "grad_norm": 0.0,
      "learning_rate": 2.181822574410514e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11666
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.625,
      "epoch": 0.5637321221492075,
      "grad_norm": 0.0,
      "learning_rate": 2.181580981832238e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11667
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.15625,
      "epoch": 0.5637804406648628,
      "grad_norm": 0.0,
      "learning_rate": 2.181339389253962e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11668
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.0,
      "epoch": 0.563828759180518,
      "grad_norm": 0.0,
      "learning_rate": 2.1810977966756861e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11669
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.65625,
      "epoch": 0.5638770776961731,
      "grad_norm": 0.0,
      "learning_rate": 2.18085620409741e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11670
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.78125,
      "epoch": 0.5639253962118284,
      "grad_norm": 0.0,
      "learning_rate": 2.180614611519134e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11671
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.0625,
      "epoch": 0.5639737147274836,
      "grad_norm": 0.0,
      "learning_rate": 2.180373018940858e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 11672
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.4375,
      "epoch": 0.5640220332431388,
      "grad_norm": 0.0,
      "learning_rate": 2.180131426362582e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11673
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.28125,
      "epoch": 0.5640703517587939,
      "grad_norm": 0.0,
      "learning_rate": 2.179889833784306e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11674
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.0625,
      "epoch": 0.5641186702744492,
      "grad_norm": 0.0,
      "learning_rate": 2.17964824120603e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11675
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.1875,
      "epoch": 0.5641669887901044,
      "grad_norm": 0.0,
      "learning_rate": 2.179406648627754e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11676
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.5625,
      "epoch": 0.5642153073057595,
      "grad_norm": 0.0,
      "learning_rate": 2.1791650560494782e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11677
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.65625,
      "epoch": 0.5642636258214148,
      "grad_norm": 0.0,
      "learning_rate": 2.178923463471202e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11678
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.4375,
      "epoch": 0.56431194433707,
      "grad_norm": 0.0,
      "learning_rate": 2.1786818708929262e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11679
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.59375,
      "epoch": 0.5643602628527251,
      "grad_norm": 0.0,
      "learning_rate": 2.17844027831465e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11680
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.40625,
      "epoch": 0.5644085813683803,
      "grad_norm": 0.0,
      "learning_rate": 2.1781986857363742e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11681
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 82.59375,
      "epoch": 0.5644568998840356,
      "grad_norm": 11.378761471024987,
      "learning_rate": 2.177957093158098e-07,
      "loss": 0.0,
      "reward": 1.5,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11682
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.3125,
      "epoch": 0.5645052183996908,
      "grad_norm": 0.0,
      "learning_rate": 2.177715500579822e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11683
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.125,
      "epoch": 0.5645535369153459,
      "grad_norm": 0.0,
      "learning_rate": 2.177473908001546e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11684
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 68.625,
      "epoch": 0.5646018554310012,
      "grad_norm": 0.0,
      "learning_rate": 2.17723231542327e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11685
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.28125,
      "epoch": 0.5646501739466564,
      "grad_norm": 0.0,
      "learning_rate": 2.176990722844994e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11686
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.46875,
      "epoch": 0.5646984924623115,
      "grad_norm": 0.0,
      "learning_rate": 2.1767491302667182e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11687
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.90625,
      "epoch": 0.5647468109779668,
      "grad_norm": 0.0,
      "learning_rate": 2.176507537688442e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11688
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.6875,
      "epoch": 0.564795129493622,
      "grad_norm": 0.0,
      "learning_rate": 2.1762659451101662e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11689
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.8125,
      "epoch": 0.5648434480092771,
      "grad_norm": 0.0,
      "learning_rate": 2.17602435253189e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11690
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.96875,
      "epoch": 0.5648917665249323,
      "grad_norm": 0.0,
      "learning_rate": 2.1757827599536142e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11691
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.65625,
      "epoch": 0.5649400850405876,
      "grad_norm": 0.0,
      "learning_rate": 2.175541167375338e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11692
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.71875,
      "epoch": 0.5649884035562428,
      "grad_norm": 0.0,
      "learning_rate": 2.175299574797062e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11693
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.1875,
      "epoch": 0.5650367220718979,
      "grad_norm": 0.0,
      "learning_rate": 2.175057982218786e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11694
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.90625,
      "epoch": 0.5650850405875532,
      "grad_norm": 0.0,
      "learning_rate": 2.17481638964051e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11695
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.8125,
      "epoch": 0.5651333591032084,
      "grad_norm": 0.0,
      "learning_rate": 2.174574797062234e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11696
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.28125,
      "epoch": 0.5651816776188635,
      "grad_norm": 0.0,
      "learning_rate": 2.1743332044839582e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11697
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.65625,
      "epoch": 0.5652299961345187,
      "grad_norm": 0.0,
      "learning_rate": 2.174091611905682e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11698
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.15625,
      "epoch": 0.565278314650174,
      "grad_norm": 0.0,
      "learning_rate": 2.1738500193274063e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 11699
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.5625,
      "epoch": 0.5653266331658291,
      "grad_norm": 0.0,
      "learning_rate": 2.17360842674913e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11700
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.1875,
      "epoch": 0.5653749516814843,
      "grad_norm": 0.0,
      "learning_rate": 2.1733668341708543e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11701
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 88.28125,
      "epoch": 0.5654232701971396,
      "grad_norm": 8.009069465911377,
      "learning_rate": 2.1731252415925781e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11702
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.5625,
      "epoch": 0.5654715887127948,
      "grad_norm": 0.0,
      "learning_rate": 2.172883649014302e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11703
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.0,
      "epoch": 0.5655199072284499,
      "grad_norm": 0.0,
      "learning_rate": 2.1726420564360261e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11704
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.5625,
      "epoch": 0.5655682257441051,
      "grad_norm": 0.0,
      "learning_rate": 2.17240046385775e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11705
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.65625,
      "epoch": 0.5656165442597604,
      "grad_norm": 0.0,
      "learning_rate": 2.1721588712794742e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11706
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.84375,
      "epoch": 0.5656648627754155,
      "grad_norm": 0.0,
      "learning_rate": 2.1719172787011983e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11707
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.34375,
      "epoch": 0.5657131812910707,
      "grad_norm": 0.0,
      "learning_rate": 2.1716756861229222e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11708
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.9375,
      "epoch": 0.565761499806726,
      "grad_norm": 0.0,
      "learning_rate": 2.1714340935446463e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11709
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.5625,
      "epoch": 0.5658098183223811,
      "grad_norm": 0.0,
      "learning_rate": 2.1711925009663702e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11710
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.3125,
      "epoch": 0.5658581368380363,
      "grad_norm": 0.0,
      "learning_rate": 2.1709509083880943e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11711
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.90625,
      "epoch": 0.5659064553536916,
      "grad_norm": 0.0,
      "learning_rate": 2.1707093158098184e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11712
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.0625,
      "epoch": 0.5659547738693468,
      "grad_norm": 0.0,
      "learning_rate": 2.170467723231542e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11713
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.96875,
      "epoch": 0.5660030923850019,
      "grad_norm": 0.0,
      "learning_rate": 2.1702261306532662e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11714
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.5625,
      "epoch": 0.5660514109006571,
      "grad_norm": 0.0,
      "learning_rate": 2.16998453807499e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11715
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.21875,
      "epoch": 0.5660997294163124,
      "grad_norm": 0.0,
      "learning_rate": 2.1697429454967142e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11716
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.3125,
      "epoch": 0.5661480479319675,
      "grad_norm": 0.0,
      "learning_rate": 2.1695013529184383e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11717
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.21875,
      "epoch": 0.5661963664476227,
      "grad_norm": 0.0,
      "learning_rate": 2.1692597603401622e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11718
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.3125,
      "epoch": 0.566244684963278,
      "grad_norm": 0.0,
      "learning_rate": 2.1690181677618863e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11719
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.96875,
      "epoch": 0.5662930034789331,
      "grad_norm": 0.0,
      "learning_rate": 2.1687765751836102e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11720
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.375,
      "epoch": 0.5663413219945883,
      "grad_norm": 0.0,
      "learning_rate": 2.1685349826053343e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11721
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.53125,
      "epoch": 0.5663896405102435,
      "grad_norm": 0.0,
      "learning_rate": 2.1682933900270585e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11722
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.21875,
      "epoch": 0.5664379590258988,
      "grad_norm": 0.0,
      "learning_rate": 2.168051797448782e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11723
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.8125,
      "epoch": 0.5664862775415539,
      "grad_norm": 0.0,
      "learning_rate": 2.1678102048705062e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11724
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.90625,
      "epoch": 0.5665345960572091,
      "grad_norm": 0.0,
      "learning_rate": 2.16756861229223e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11725
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.96875,
      "epoch": 0.5665829145728644,
      "grad_norm": 0.0,
      "learning_rate": 2.1673270197139542e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11726
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.9375,
      "epoch": 0.5666312330885195,
      "grad_norm": 0.0,
      "learning_rate": 2.1670854271356784e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11727
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.34375,
      "epoch": 0.5666795516041747,
      "grad_norm": 0.0,
      "learning_rate": 2.1668438345574022e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11728
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.25,
      "epoch": 0.56672787011983,
      "grad_norm": 0.0,
      "learning_rate": 2.1666022419791264e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11729
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.6875,
      "epoch": 0.5667761886354851,
      "grad_norm": 0.0,
      "learning_rate": 2.1663606494008502e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11730
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.53125,
      "epoch": 0.5668245071511403,
      "grad_norm": 0.0,
      "learning_rate": 2.1661190568225744e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11731
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.71875,
      "epoch": 0.5668728256667955,
      "grad_norm": 0.0,
      "learning_rate": 2.1658774642442985e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11732
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.59375,
      "epoch": 0.5669211441824508,
      "grad_norm": 0.0,
      "learning_rate": 2.165635871666022e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11733
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.90625,
      "epoch": 0.5669694626981059,
      "grad_norm": 0.0,
      "learning_rate": 2.1653942790877463e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11734
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.9375,
      "epoch": 0.5670177812137611,
      "grad_norm": 0.0,
      "learning_rate": 2.1651526865094701e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11735
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.28125,
      "epoch": 0.5670660997294164,
      "grad_norm": 0.0,
      "learning_rate": 2.1649110939311943e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11736
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.5,
      "epoch": 0.5671144182450715,
      "grad_norm": 0.0,
      "learning_rate": 2.1646695013529184e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11737
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 89.3125,
      "epoch": 0.5671627367607267,
      "grad_norm": 0.0,
      "learning_rate": 2.1644279087746423e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 11738
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.40625,
      "epoch": 0.5672110552763819,
      "grad_norm": 0.0,
      "learning_rate": 2.1641863161963664e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11739
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.40625,
      "epoch": 0.5672593737920371,
      "grad_norm": 0.0,
      "learning_rate": 2.1639447236180903e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11740
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 94.6875,
      "epoch": 0.5673076923076923,
      "grad_norm": 0.0,
      "learning_rate": 2.1637031310398144e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 11741
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.78125,
      "epoch": 0.5673560108233475,
      "grad_norm": 0.0,
      "learning_rate": 2.1634615384615386e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11742
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.0625,
      "epoch": 0.5674043293390028,
      "grad_norm": 0.0,
      "learning_rate": 2.1632199458832622e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11743
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.1875,
      "epoch": 0.5674526478546579,
      "grad_norm": 0.0,
      "learning_rate": 2.1629783533049863e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11744
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.8125,
      "epoch": 0.5675009663703131,
      "grad_norm": 0.0,
      "learning_rate": 2.1627367607267102e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11745
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.0,
      "epoch": 0.5675492848859683,
      "grad_norm": 0.0,
      "learning_rate": 2.1624951681484343e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11746
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.75,
      "epoch": 0.5675976034016235,
      "grad_norm": 0.0,
      "learning_rate": 2.1622535755701584e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11747
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.03125,
      "epoch": 0.5676459219172787,
      "grad_norm": 0.0,
      "learning_rate": 2.1620119829918823e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11748
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.71875,
      "epoch": 0.5676942404329339,
      "grad_norm": 0.0,
      "learning_rate": 2.1617703904136065e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11749
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.625,
      "epoch": 0.567742558948589,
      "grad_norm": 0.0,
      "learning_rate": 2.1615287978353303e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11750
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.875,
      "epoch": 0.5677908774642443,
      "grad_norm": 0.0,
      "learning_rate": 2.1612872052570545e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11751
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.3125,
      "epoch": 0.5678391959798995,
      "grad_norm": 0.0,
      "learning_rate": 2.1610456126787786e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11752
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.09375,
      "epoch": 0.5678875144955547,
      "grad_norm": 0.0,
      "learning_rate": 2.1608040201005022e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11753
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.21875,
      "epoch": 0.5679358330112099,
      "grad_norm": 0.0,
      "learning_rate": 2.1605624275222263e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.260896235704422,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11754
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.90625,
      "epoch": 0.5679841515268651,
      "grad_norm": 0.0,
      "learning_rate": 2.1603208349439502e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11755
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.84375,
      "epoch": 0.5680324700425203,
      "grad_norm": 0.0,
      "learning_rate": 2.1600792423656743e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11756
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.34375,
      "epoch": 0.5680807885581755,
      "grad_norm": 0.0,
      "learning_rate": 2.1598376497873985e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11757
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.25,
      "epoch": 0.5681291070738307,
      "grad_norm": 0.0,
      "learning_rate": 2.1595960572091224e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11758
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.21875,
      "epoch": 0.5681774255894859,
      "grad_norm": 0.0,
      "learning_rate": 2.1593544646308465e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11759
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.96875,
      "epoch": 0.568225744105141,
      "grad_norm": 0.0,
      "learning_rate": 2.1591128720525704e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11760
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.15625,
      "epoch": 0.5682740626207963,
      "grad_norm": 0.0,
      "learning_rate": 2.1588712794742945e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11761
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 96.9375,
      "epoch": 0.5683223811364515,
      "grad_norm": 0.0,
      "learning_rate": 2.1586296868960186e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 11762
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.8125,
      "epoch": 0.5683706996521067,
      "grad_norm": 0.0,
      "learning_rate": 2.1583880943177425e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11763
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.8125,
      "epoch": 0.5684190181677619,
      "grad_norm": 0.0,
      "learning_rate": 2.1581465017394664e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11764
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.59375,
      "epoch": 0.5684673366834171,
      "grad_norm": 0.0,
      "learning_rate": 2.1579049091611903e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11765
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.90625,
      "epoch": 0.5685156551990723,
      "grad_norm": 0.0,
      "learning_rate": 2.1576633165829144e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11766
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.5,
      "epoch": 0.5685639737147274,
      "grad_norm": 0.0,
      "learning_rate": 2.1574217240046385e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11767
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.84375,
      "epoch": 0.5686122922303827,
      "grad_norm": 0.0,
      "learning_rate": 2.1571801314263624e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11768
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.0,
      "epoch": 0.5686606107460379,
      "grad_norm": 0.0,
      "learning_rate": 2.1569385388480865e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11769
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.59375,
      "epoch": 0.568708929261693,
      "grad_norm": 0.0,
      "learning_rate": 2.1566969462698104e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11770
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.25,
      "epoch": 0.5687572477773483,
      "grad_norm": 0.0,
      "learning_rate": 2.1564553536915345e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11771
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.53125,
      "epoch": 0.5688055662930035,
      "grad_norm": 0.0,
      "learning_rate": 2.1562137611132587e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11772
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.84375,
      "epoch": 0.5688538848086587,
      "grad_norm": 0.0,
      "learning_rate": 2.1559721685349825e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11773
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.75,
      "epoch": 0.5689022033243138,
      "grad_norm": 0.0,
      "learning_rate": 2.1557305759567064e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11774
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.21875,
      "epoch": 0.5689505218399691,
      "grad_norm": 0.0,
      "learning_rate": 2.1554889833784303e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11775
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.84375,
      "epoch": 0.5689988403556243,
      "grad_norm": 0.0,
      "learning_rate": 2.1552473908001544e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11776
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.40625,
      "epoch": 0.5690471588712794,
      "grad_norm": 0.0,
      "learning_rate": 2.1550057982218786e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11777
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.75,
      "epoch": 0.5690954773869347,
      "grad_norm": 0.0,
      "learning_rate": 2.1547642056436024e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11778
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.875,
      "epoch": 0.5691437959025899,
      "grad_norm": 0.0,
      "learning_rate": 2.1545226130653266e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11779
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.8125,
      "epoch": 0.569192114418245,
      "grad_norm": 0.0,
      "learning_rate": 2.1542810204870504e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11780
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.1875,
      "epoch": 0.5692404329339003,
      "grad_norm": 0.0,
      "learning_rate": 2.1540394279087746e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11781
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.125,
      "epoch": 0.5692887514495555,
      "grad_norm": 0.0,
      "learning_rate": 2.1537978353304987e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11782
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.1875,
      "epoch": 0.5693370699652107,
      "grad_norm": 0.0,
      "learning_rate": 2.1535562427522226e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11783
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.21875,
      "epoch": 0.5693853884808658,
      "grad_norm": 0.0,
      "learning_rate": 2.1533146501739465e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11784
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 85.40625,
      "epoch": 0.5694337069965211,
      "grad_norm": 21.021796637392328,
      "learning_rate": 2.1530730575956703e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2561880350112915,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11785
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.5625,
      "epoch": 0.5694820255121763,
      "grad_norm": 0.0,
      "learning_rate": 2.1528314650173945e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11786
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.71875,
      "epoch": 0.5695303440278314,
      "grad_norm": 0.0,
      "learning_rate": 2.1525898724391186e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11787
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.09375,
      "epoch": 0.5695786625434867,
      "grad_norm": 0.0,
      "learning_rate": 2.1523482798608425e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11788
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.3125,
      "epoch": 0.5696269810591419,
      "grad_norm": 0.0,
      "learning_rate": 2.1521066872825666e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11789
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.09375,
      "epoch": 0.569675299574797,
      "grad_norm": 0.0,
      "learning_rate": 2.1518650947042905e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11790
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.59375,
      "epoch": 0.5697236180904522,
      "grad_norm": 0.0,
      "learning_rate": 2.1516235021260146e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11791
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.0,
      "epoch": 0.5697719366061075,
      "grad_norm": 0.0,
      "learning_rate": 2.1513819095477388e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11792
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.59375,
      "epoch": 0.5698202551217627,
      "grad_norm": 0.0,
      "learning_rate": 2.1511403169694626e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11793
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.6875,
      "epoch": 0.5698685736374178,
      "grad_norm": 0.0,
      "learning_rate": 2.1508987243911868e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11794
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.5,
      "epoch": 0.5699168921530731,
      "grad_norm": 0.0,
      "learning_rate": 2.1506571318129104e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11795
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.15625,
      "epoch": 0.5699652106687283,
      "grad_norm": 0.0,
      "learning_rate": 2.1504155392346345e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11796
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.34375,
      "epoch": 0.5700135291843834,
      "grad_norm": 0.0,
      "learning_rate": 2.1501739466563586e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11797
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.5,
      "epoch": 0.5700618477000386,
      "grad_norm": 0.0,
      "learning_rate": 2.1499323540780825e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11798
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.03125,
      "epoch": 0.5701101662156939,
      "grad_norm": 0.0,
      "learning_rate": 2.1496907614998066e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11799
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.25,
      "epoch": 0.570158484731349,
      "grad_norm": 0.0,
      "learning_rate": 2.1494491689215305e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11800
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.78125,
      "epoch": 0.5702068032470042,
      "grad_norm": 0.0,
      "learning_rate": 2.1492075763432547e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11801
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.9375,
      "epoch": 0.5702551217626595,
      "grad_norm": 0.0,
      "learning_rate": 2.1489659837649788e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11802
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.34375,
      "epoch": 0.5703034402783147,
      "grad_norm": 0.0,
      "learning_rate": 2.1487243911867027e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11803
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.34375,
      "epoch": 0.5703517587939698,
      "grad_norm": 0.0,
      "learning_rate": 2.1484827986084268e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11804
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.8125,
      "epoch": 0.570400077309625,
      "grad_norm": 0.0,
      "learning_rate": 2.1482412060301504e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11805
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.90625,
      "epoch": 0.5704483958252803,
      "grad_norm": 0.0,
      "learning_rate": 2.1479996134518745e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11806
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.46875,
      "epoch": 0.5704967143409354,
      "grad_norm": 0.0,
      "learning_rate": 2.1477580208735987e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11807
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.15625,
      "epoch": 0.5705450328565906,
      "grad_norm": 0.0,
      "learning_rate": 2.1475164282953226e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11808
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.0,
      "epoch": 0.5705933513722459,
      "grad_norm": 0.0,
      "learning_rate": 2.1472748357170467e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11809
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.375,
      "epoch": 0.570641669887901,
      "grad_norm": 0.0,
      "learning_rate": 2.1470332431387706e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11810
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.4375,
      "epoch": 0.5706899884035562,
      "grad_norm": 0.0,
      "learning_rate": 2.1467916505604947e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11811
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.90625,
      "epoch": 0.5707383069192115,
      "grad_norm": 0.0,
      "learning_rate": 2.1465500579822188e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11812
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.9375,
      "epoch": 0.5707866254348667,
      "grad_norm": 0.0,
      "learning_rate": 2.1463084654039427e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11813
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.8125,
      "epoch": 0.5708349439505218,
      "grad_norm": 0.0,
      "learning_rate": 2.1460668728256668e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11814
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.59375,
      "epoch": 0.570883262466177,
      "grad_norm": 0.0,
      "learning_rate": 2.1458252802473904e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11815
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.40625,
      "epoch": 0.5709315809818323,
      "grad_norm": 0.0,
      "learning_rate": 2.1455836876691146e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11816
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.46875,
      "epoch": 0.5709798994974874,
      "grad_norm": 0.0,
      "learning_rate": 2.1453420950908387e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11817
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.875,
      "epoch": 0.5710282180131426,
      "grad_norm": 0.0,
      "learning_rate": 2.1451005025125626e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11818
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.34375,
      "epoch": 0.5710765365287979,
      "grad_norm": 0.0,
      "learning_rate": 2.1448589099342867e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11819
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.59375,
      "epoch": 0.571124855044453,
      "grad_norm": 0.0,
      "learning_rate": 2.1446173173560106e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11820
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.8125,
      "epoch": 0.5711731735601082,
      "grad_norm": 0.0,
      "learning_rate": 2.1443757247777347e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11821
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.78125,
      "epoch": 0.5712214920757634,
      "grad_norm": 0.0,
      "learning_rate": 2.144134132199459e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11822
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.21875,
      "epoch": 0.5712698105914187,
      "grad_norm": 0.0,
      "learning_rate": 2.1438925396211827e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11823
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 87.125,
      "epoch": 0.5713181291070738,
      "grad_norm": 0.0,
      "learning_rate": 2.143650947042907e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 11824
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.46875,
      "epoch": 0.571366447622729,
      "grad_norm": 0.0,
      "learning_rate": 2.1434093544646305e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11825
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.625,
      "epoch": 0.5714147661383843,
      "grad_norm": 0.0,
      "learning_rate": 2.1431677618863546e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11826
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.71875,
      "epoch": 0.5714630846540394,
      "grad_norm": 0.0,
      "learning_rate": 2.1429261693080788e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11827
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.59375,
      "epoch": 0.5715114031696946,
      "grad_norm": 0.0,
      "learning_rate": 2.1426845767298026e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11828
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.21875,
      "epoch": 0.5715597216853499,
      "grad_norm": 0.0,
      "learning_rate": 2.1424429841515268e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11829
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.46875,
      "epoch": 0.571608040201005,
      "grad_norm": 0.0,
      "learning_rate": 2.1422013915732506e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11830
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.28125,
      "epoch": 0.5716563587166602,
      "grad_norm": 0.0,
      "learning_rate": 2.1419597989949748e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11831
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.625,
      "epoch": 0.5717046772323154,
      "grad_norm": 0.0,
      "learning_rate": 2.141718206416699e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11832
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.8125,
      "epoch": 0.5717529957479707,
      "grad_norm": 0.0,
      "learning_rate": 2.1414766138384228e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11833
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.96875,
      "epoch": 0.5718013142636258,
      "grad_norm": 0.0,
      "learning_rate": 2.141235021260147e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11834
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.96875,
      "epoch": 0.571849632779281,
      "grad_norm": 0.0,
      "learning_rate": 2.1409934286818708e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11835
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.5,
      "epoch": 0.5718979512949363,
      "grad_norm": 0.0,
      "learning_rate": 2.1407518361035947e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11836
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.625,
      "epoch": 0.5719462698105914,
      "grad_norm": 0.0,
      "learning_rate": 2.1405102435253188e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11837
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.375,
      "epoch": 0.5719945883262466,
      "grad_norm": 0.0,
      "learning_rate": 2.1402686509470427e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11838
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.03125,
      "epoch": 0.5720429068419018,
      "grad_norm": 0.0,
      "learning_rate": 2.1400270583687668e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11839
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.59375,
      "epoch": 0.572091225357557,
      "grad_norm": 0.0,
      "learning_rate": 2.1397854657904907e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11840
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.84375,
      "epoch": 0.5721395438732122,
      "grad_norm": 0.0,
      "learning_rate": 2.1395438732122148e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11841
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.0625,
      "epoch": 0.5721878623888674,
      "grad_norm": 0.0,
      "learning_rate": 2.139302280633939e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11842
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.78125,
      "epoch": 0.5722361809045227,
      "grad_norm": 0.0,
      "learning_rate": 2.1390606880556628e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11843
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.84375,
      "epoch": 0.5722844994201778,
      "grad_norm": 0.0,
      "learning_rate": 2.138819095477387e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11844
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.53125,
      "epoch": 0.572332817935833,
      "grad_norm": 0.0,
      "learning_rate": 2.1385775028991108e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11845
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.71875,
      "epoch": 0.5723811364514882,
      "grad_norm": 0.0,
      "learning_rate": 2.1383359103208347e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11846
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.1875,
      "epoch": 0.5724294549671434,
      "grad_norm": 0.0,
      "learning_rate": 2.1380943177425588e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11847
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.28125,
      "epoch": 0.5724777734827986,
      "grad_norm": 0.0,
      "learning_rate": 2.1378527251642827e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11848
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.625,
      "epoch": 0.5725260919984538,
      "grad_norm": 0.0,
      "learning_rate": 2.1376111325860068e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11849
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.96875,
      "epoch": 0.572574410514109,
      "grad_norm": 0.0,
      "learning_rate": 2.1373695400077307e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11850
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.8125,
      "epoch": 0.5726227290297642,
      "grad_norm": 0.0,
      "learning_rate": 2.1371279474294549e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11851
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 72.71875,
      "epoch": 0.5726710475454194,
      "grad_norm": 0.0,
      "learning_rate": 2.136886354851179e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11852
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.53125,
      "epoch": 0.5727193660610747,
      "grad_norm": 0.0,
      "learning_rate": 2.1366447622729029e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11853
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.9375,
      "epoch": 0.5727676845767298,
      "grad_norm": 0.0,
      "learning_rate": 2.136403169694627e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11854
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.0,
      "epoch": 0.572816003092385,
      "grad_norm": 0.0,
      "learning_rate": 2.136161577116351e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11855
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.59375,
      "epoch": 0.5728643216080402,
      "grad_norm": 0.0,
      "learning_rate": 2.1359199845380747e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11856
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.53125,
      "epoch": 0.5729126401236954,
      "grad_norm": 0.0,
      "learning_rate": 2.135678391959799e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11857
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.5625,
      "epoch": 0.5729609586393506,
      "grad_norm": 0.0,
      "learning_rate": 2.1354367993815227e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11858
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.46875,
      "epoch": 0.5730092771550058,
      "grad_norm": 0.0,
      "learning_rate": 2.135195206803247e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11859
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.96875,
      "epoch": 0.573057595670661,
      "grad_norm": 0.0,
      "learning_rate": 2.1349536142249708e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11860
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.59375,
      "epoch": 0.5731059141863162,
      "grad_norm": 7.930489634911782,
      "learning_rate": 2.134712021646695e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2883424758911133,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11861
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.46875,
      "epoch": 0.5731542327019714,
      "grad_norm": 0.0,
      "learning_rate": 2.134470429068419e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11862
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.875,
      "epoch": 0.5732025512176266,
      "grad_norm": 110.36999531305094,
      "learning_rate": 2.134228836490143e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.3129216432571411,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11863
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.125,
      "epoch": 0.5732508697332818,
      "grad_norm": 0.0,
      "learning_rate": 2.133987243911867e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11864
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 71.46875,
      "epoch": 0.573299188248937,
      "grad_norm": 0.0,
      "learning_rate": 2.133745651333591e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11865
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.6875,
      "epoch": 0.5733475067645922,
      "grad_norm": 0.0,
      "learning_rate": 2.133504058755315e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11866
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.96875,
      "epoch": 0.5733958252802474,
      "grad_norm": 0.0,
      "learning_rate": 2.133262466177039e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11867
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.03125,
      "epoch": 0.5734441437959026,
      "grad_norm": 0.0,
      "learning_rate": 2.1330208735987628e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11868
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.96875,
      "epoch": 0.5734924623115578,
      "grad_norm": 0.0,
      "learning_rate": 2.132779281020487e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11869
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.625,
      "epoch": 0.5735407808272129,
      "grad_norm": 0.0,
      "learning_rate": 2.1325376884422108e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11870
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.28125,
      "epoch": 0.5735890993428682,
      "grad_norm": 0.0,
      "learning_rate": 2.132296095863935e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 11871
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.1875,
      "epoch": 0.5736374178585234,
      "grad_norm": 0.0,
      "learning_rate": 2.132054503285659e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11872
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.5625,
      "epoch": 0.5736857363741786,
      "grad_norm": 0.0,
      "learning_rate": 2.131812910707383e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11873
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.625,
      "epoch": 0.5737340548898338,
      "grad_norm": 0.0,
      "learning_rate": 2.131571318129107e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11874
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.59375,
      "epoch": 0.573782373405489,
      "grad_norm": 0.0,
      "learning_rate": 2.131329725550831e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 11875
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.53125,
      "epoch": 0.5738306919211442,
      "grad_norm": 0.0,
      "learning_rate": 2.131088132972555e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11876
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.3125,
      "epoch": 0.5738790104367993,
      "grad_norm": 0.0,
      "learning_rate": 2.130846540394279e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 11877
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.46875,
      "epoch": 0.5739273289524546,
      "grad_norm": 0.0,
      "learning_rate": 2.1306049478160028e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11878
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 69.34375,
      "epoch": 0.5739756474681098,
      "grad_norm": 0.0,
      "learning_rate": 2.130363355237727e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 11879
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.53125,
      "epoch": 0.5740239659837649,
      "grad_norm": 0.0,
      "learning_rate": 2.1301217626594508e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11880
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.9375,
      "epoch": 0.5740722844994202,
      "grad_norm": 0.0,
      "learning_rate": 2.129880170081175e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11881
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.03125,
      "epoch": 0.5741206030150754,
      "grad_norm": 0.0,
      "learning_rate": 2.129638577502899e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11882
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.9375,
      "epoch": 0.5741689215307306,
      "grad_norm": 0.0,
      "learning_rate": 2.129396984924623e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11883
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.1875,
      "epoch": 0.5742172400463857,
      "grad_norm": 0.0,
      "learning_rate": 2.129155392346347e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11884
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.78125,
      "epoch": 0.574265558562041,
      "grad_norm": 0.0,
      "learning_rate": 2.128913799768071e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11885
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.3125,
      "epoch": 0.5743138770776962,
      "grad_norm": 0.0,
      "learning_rate": 2.128672207189795e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 11886
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 75.34375,
      "epoch": 0.5743621955933513,
      "grad_norm": 0.0,
      "learning_rate": 2.128430614611519e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 11887
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.0625,
      "epoch": 0.5744105141090066,
      "grad_norm": 0.0,
      "learning_rate": 2.1281890220332429e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11888
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.3125,
      "epoch": 0.5744588326246618,
      "grad_norm": 0.0,
      "learning_rate": 2.127947429454967e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11889
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.0625,
      "epoch": 0.5745071511403169,
      "grad_norm": 0.0,
      "learning_rate": 2.127705836876691e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11890
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.6875,
      "epoch": 0.5745554696559722,
      "grad_norm": 0.0,
      "learning_rate": 2.127464244298415e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11891
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.46875,
      "epoch": 0.5746037881716274,
      "grad_norm": 0.0,
      "learning_rate": 2.1272226517201391e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11892
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.25,
      "epoch": 0.5746521066872826,
      "grad_norm": 0.0,
      "learning_rate": 2.126981059141863e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 11893
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.0625,
      "epoch": 0.5747004252029377,
      "grad_norm": 0.0,
      "learning_rate": 2.1267394665635872e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11894
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.03125,
      "epoch": 0.574748743718593,
      "grad_norm": 0.0,
      "learning_rate": 2.126497873985311e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 11895
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.46875,
      "epoch": 0.5747970622342482,
      "grad_norm": 0.0,
      "learning_rate": 2.1262562814070352e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11896
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.15625,
      "epoch": 0.5748453807499033,
      "grad_norm": 0.0,
      "learning_rate": 2.1260146888287593e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11897
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.15625,
      "epoch": 0.5748936992655586,
      "grad_norm": 0.0,
      "learning_rate": 2.125773096250483e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11898
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 67.65625,
      "epoch": 0.5749420177812138,
      "grad_norm": 0.0,
      "learning_rate": 2.125531503672207e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11899
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.96875,
      "epoch": 0.5749903362968689,
      "grad_norm": 0.0,
      "learning_rate": 2.125289911093931e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11900
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.65625,
      "epoch": 0.5750386548125241,
      "grad_norm": 0.0,
      "learning_rate": 2.125048318515655e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 11901
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.125,
      "epoch": 0.5750869733281794,
      "grad_norm": 0.0,
      "learning_rate": 2.1248067259373792e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11902
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 68.9375,
      "epoch": 0.5751352918438346,
      "grad_norm": 0.0,
      "learning_rate": 2.124565133359103e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 11903
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.34375,
      "epoch": 0.5751836103594897,
      "grad_norm": 0.0,
      "learning_rate": 2.1243235407808272e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11904
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.8125,
      "epoch": 0.575231928875145,
      "grad_norm": 0.0,
      "learning_rate": 2.124081948202551e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11905
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.0625,
      "epoch": 0.5752802473908002,
      "grad_norm": 0.0,
      "learning_rate": 2.1238403556242752e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11906
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.53125,
      "epoch": 0.5753285659064553,
      "grad_norm": 0.0,
      "learning_rate": 2.1235987630459993e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11907
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.78125,
      "epoch": 0.5753768844221105,
      "grad_norm": 0.0,
      "learning_rate": 2.123357170467723e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 11908
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.3125,
      "epoch": 0.5754252029377658,
      "grad_norm": 0.0,
      "learning_rate": 2.123115577889447e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11909
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.625,
      "epoch": 0.5754735214534209,
      "grad_norm": 0.0,
      "learning_rate": 2.122873985311171e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11910
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.71875,
      "epoch": 0.5755218399690761,
      "grad_norm": 0.0,
      "learning_rate": 2.122632392732895e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11911
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.0,
      "epoch": 0.5755701584847314,
      "grad_norm": 0.0,
      "learning_rate": 2.1223908001546192e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11912
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.1875,
      "epoch": 0.5756184770003866,
      "grad_norm": 0.0,
      "learning_rate": 2.122149207576343e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11913
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.84375,
      "epoch": 0.5756667955160417,
      "grad_norm": 0.0,
      "learning_rate": 2.1219076149980672e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11914
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.4375,
      "epoch": 0.575715114031697,
      "grad_norm": 0.0,
      "learning_rate": 2.121666022419791e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11915
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.1875,
      "epoch": 0.5757634325473522,
      "grad_norm": 0.0,
      "learning_rate": 2.1214244298415152e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 11916
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 72.25,
      "epoch": 0.5758117510630073,
      "grad_norm": 0.0,
      "learning_rate": 2.1211828372632394e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11917
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.125,
      "epoch": 0.5758600695786625,
      "grad_norm": 0.0,
      "learning_rate": 2.120941244684963e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11918
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.28125,
      "epoch": 0.5759083880943178,
      "grad_norm": 0.0,
      "learning_rate": 2.120699652106687e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11919
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.4375,
      "epoch": 0.5759567066099729,
      "grad_norm": 0.0,
      "learning_rate": 2.120458059528411e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11920
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.59375,
      "epoch": 0.5760050251256281,
      "grad_norm": 0.0,
      "learning_rate": 2.120216466950135e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11921
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.0,
      "epoch": 0.5760533436412834,
      "grad_norm": 0.0,
      "learning_rate": 2.1199748743718593e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11922
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.8125,
      "epoch": 0.5761016621569386,
      "grad_norm": 0.0,
      "learning_rate": 2.1197332817935831e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11923
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.4375,
      "epoch": 0.5761499806725937,
      "grad_norm": 0.0,
      "learning_rate": 2.1194916892153073e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11924
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.40625,
      "epoch": 0.5761982991882489,
      "grad_norm": 0.0,
      "learning_rate": 2.1192500966370311e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11925
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.125,
      "epoch": 0.5762466177039042,
      "grad_norm": 0.0,
      "learning_rate": 2.1190085040587553e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11926
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.46875,
      "epoch": 0.5762949362195593,
      "grad_norm": 0.0,
      "learning_rate": 2.1187669114804794e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11927
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.28125,
      "epoch": 0.5763432547352145,
      "grad_norm": 0.0,
      "learning_rate": 2.118525318902203e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11928
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.6875,
      "epoch": 0.5763915732508698,
      "grad_norm": 0.0,
      "learning_rate": 2.1182837263239272e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11929
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.875,
      "epoch": 0.5764398917665249,
      "grad_norm": 0.0,
      "learning_rate": 2.118042133745651e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11930
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.03125,
      "epoch": 0.5764882102821801,
      "grad_norm": 0.0,
      "learning_rate": 2.1178005411673752e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11931
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.375,
      "epoch": 0.5765365287978353,
      "grad_norm": 0.0,
      "learning_rate": 2.1175589485890993e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11932
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.90625,
      "epoch": 0.5765848473134906,
      "grad_norm": 0.0,
      "learning_rate": 2.1173173560108232e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 11933
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.59375,
      "epoch": 0.5766331658291457,
      "grad_norm": 0.0,
      "learning_rate": 2.1170757634325473e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11934
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.625,
      "epoch": 0.5766814843448009,
      "grad_norm": 0.0,
      "learning_rate": 2.1168341708542712e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11935
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 64.625,
      "epoch": 0.5767298028604562,
      "grad_norm": 0.0,
      "learning_rate": 2.1165925782759953e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11936
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.96875,
      "epoch": 0.5767781213761113,
      "grad_norm": 0.0,
      "learning_rate": 2.1163509856977195e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11937
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 71.28125,
      "epoch": 0.5768264398917665,
      "grad_norm": 0.0,
      "learning_rate": 2.1161093931194433e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 11938
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.25,
      "epoch": 0.5768747584074218,
      "grad_norm": 134.86847524525456,
      "learning_rate": 2.1158678005411672e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.25194472074508667,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11939
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.21875,
      "epoch": 0.5769230769230769,
      "grad_norm": 0.0,
      "learning_rate": 2.115626207962891e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11940
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.125,
      "epoch": 0.5769713954387321,
      "grad_norm": 0.0,
      "learning_rate": 2.1153846153846152e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11941
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.1875,
      "epoch": 0.5770197139543873,
      "grad_norm": 0.0,
      "learning_rate": 2.1151430228063393e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11942
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.625,
      "epoch": 0.5770680324700426,
      "grad_norm": 0.0,
      "learning_rate": 2.1149014302280632e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11943
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.59375,
      "epoch": 0.5771163509856977,
      "grad_norm": 0.0,
      "learning_rate": 2.1146598376497874e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11944
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.5625,
      "epoch": 0.5771646695013529,
      "grad_norm": 0.0,
      "learning_rate": 2.1144182450715112e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11945
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.96875,
      "epoch": 0.5772129880170082,
      "grad_norm": 0.0,
      "learning_rate": 2.1141766524932354e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11946
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.0,
      "epoch": 0.5772613065326633,
      "grad_norm": 0.0,
      "learning_rate": 2.1139350599149595e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11947
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.03125,
      "epoch": 0.5773096250483185,
      "grad_norm": 0.0,
      "learning_rate": 2.1136934673366834e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 11948
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.65625,
      "epoch": 0.5773579435639737,
      "grad_norm": 0.0,
      "learning_rate": 2.1134518747584072e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 11949
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 57.15625,
      "epoch": 0.5774062620796289,
      "grad_norm": 0.0,
      "learning_rate": 2.113210282180131e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 11950
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.3125,
      "epoch": 0.5774545805952841,
      "grad_norm": 0.0,
      "learning_rate": 2.1129686896018552e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11951
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.625,
      "epoch": 0.5775028991109393,
      "grad_norm": 0.0,
      "learning_rate": 2.1127270970235794e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11952
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.375,
      "epoch": 0.5775512176265946,
      "grad_norm": 0.0,
      "learning_rate": 2.1124855044453033e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11953
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.34375,
      "epoch": 0.5775995361422497,
      "grad_norm": 0.0,
      "learning_rate": 2.1122439118670274e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11954
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.8125,
      "epoch": 0.5776478546579049,
      "grad_norm": 0.0,
      "learning_rate": 2.1120023192887513e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11955
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.625,
      "epoch": 0.5776961731735601,
      "grad_norm": 0.0,
      "learning_rate": 2.1117607267104754e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11956
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 63.5,
      "epoch": 0.5777444916892153,
      "grad_norm": 0.0,
      "learning_rate": 2.1115191341321995e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11957
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.65625,
      "epoch": 0.5777928102048705,
      "grad_norm": 0.0,
      "learning_rate": 2.1112775415539234e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11958
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.0,
      "epoch": 0.5778411287205257,
      "grad_norm": 0.0,
      "learning_rate": 2.1110359489756473e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11959
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.21875,
      "epoch": 0.5778894472361809,
      "grad_norm": 0.0,
      "learning_rate": 2.1107943563973711e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11960
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.5625,
      "epoch": 0.5779377657518361,
      "grad_norm": 0.0,
      "learning_rate": 2.1105527638190953e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11961
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.375,
      "epoch": 0.5779860842674913,
      "grad_norm": 0.0,
      "learning_rate": 2.1103111712408194e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11962
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.5625,
      "epoch": 0.5780344027831466,
      "grad_norm": 0.0,
      "learning_rate": 2.1100695786625433e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11963
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 71.84375,
      "epoch": 0.5780827212988017,
      "grad_norm": 0.0,
      "learning_rate": 2.1098279860842674e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 11964
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.6875,
      "epoch": 0.5781310398144569,
      "grad_norm": 0.0,
      "learning_rate": 2.1095863935059913e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11965
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 73.5625,
      "epoch": 0.5781793583301121,
      "grad_norm": 12.332818532277246,
      "learning_rate": 2.1093448009277154e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.249358132481575,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11966
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.15625,
      "epoch": 0.5782276768457673,
      "grad_norm": 0.0,
      "learning_rate": 2.1091032083494396e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11967
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.21875,
      "epoch": 0.5782759953614225,
      "grad_norm": 0.0,
      "learning_rate": 2.1088616157711634e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11968
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.0625,
      "epoch": 0.5783243138770777,
      "grad_norm": 0.0,
      "learning_rate": 2.1086200231928876e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 11969
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.3125,
      "epoch": 0.5783726323927328,
      "grad_norm": 0.0,
      "learning_rate": 2.1083784306146112e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11970
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.09375,
      "epoch": 0.5784209509083881,
      "grad_norm": 0.0,
      "learning_rate": 2.1081368380363353e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11971
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.25,
      "epoch": 0.5784692694240433,
      "grad_norm": 0.0,
      "learning_rate": 2.1078952454580595e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11972
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.4375,
      "epoch": 0.5785175879396985,
      "grad_norm": 0.0,
      "learning_rate": 2.1076536528797833e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11973
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.78125,
      "epoch": 0.5785659064553537,
      "grad_norm": 0.0,
      "learning_rate": 2.1074120603015075e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 11974
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.125,
      "epoch": 0.5786142249710089,
      "grad_norm": 0.0,
      "learning_rate": 2.1071704677232313e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 11975
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.84375,
      "epoch": 0.5786625434866641,
      "grad_norm": 0.0,
      "learning_rate": 2.1069288751449555e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11976
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.46875,
      "epoch": 0.5787108620023192,
      "grad_norm": 0.0,
      "learning_rate": 2.1066872825666796e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11977
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.3125,
      "epoch": 0.5787591805179745,
      "grad_norm": 0.0,
      "learning_rate": 2.1064456899884035e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11978
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.0625,
      "epoch": 0.5788074990336297,
      "grad_norm": 0.0,
      "learning_rate": 2.1062040974101276e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 11979
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.75,
      "epoch": 0.5788558175492848,
      "grad_norm": 0.0,
      "learning_rate": 2.1059625048318512e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880052089691,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 11980
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.25,
      "epoch": 0.5789041360649401,
      "grad_norm": 0.0,
      "learning_rate": 2.1057209122535754e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11981
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.5625,
      "epoch": 0.5789524545805953,
      "grad_norm": 0.0,
      "learning_rate": 2.1054793196752995e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11982
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.96875,
      "epoch": 0.5790007730962505,
      "grad_norm": 0.0,
      "learning_rate": 2.1052377270970234e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11983
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.0,
      "epoch": 0.5790490916119057,
      "grad_norm": 0.0,
      "learning_rate": 2.1049961345187475e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 11984
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 55.9375,
      "epoch": 0.5790974101275609,
      "grad_norm": 0.0,
      "learning_rate": 2.1047545419404714e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 11985
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.03125,
      "epoch": 0.5791457286432161,
      "grad_norm": 0.0,
      "learning_rate": 2.1045129493621955e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 11986
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.59375,
      "epoch": 0.5791940471588712,
      "grad_norm": 0.0,
      "learning_rate": 2.1042713567839197e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 11987
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.5,
      "epoch": 0.5792423656745265,
      "grad_norm": 0.0,
      "learning_rate": 2.1040297642056435e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 11988
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 66.34375,
      "epoch": 0.5792906841901817,
      "grad_norm": 83.14169185477073,
      "learning_rate": 2.1037881716273677e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.2993341088294983,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 11989
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.90625,
      "epoch": 0.5793390027058368,
      "grad_norm": 0.0,
      "learning_rate": 2.1035465790490913e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 11990
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.59375,
      "epoch": 0.5793873212214921,
      "grad_norm": 0.0,
      "learning_rate": 2.1033049864708154e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 11991
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.46875,
      "epoch": 0.5794356397371473,
      "grad_norm": 0.0,
      "learning_rate": 2.1030633938925395e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11992
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.8125,
      "epoch": 0.5794839582528025,
      "grad_norm": 0.0,
      "learning_rate": 2.1028218013142634e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 11993
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.65625,
      "epoch": 0.5795322767684576,
      "grad_norm": 0.0,
      "learning_rate": 2.1025802087359875e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 11994
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.5625,
      "epoch": 0.5795805952841129,
      "grad_norm": 0.0,
      "learning_rate": 2.1023386161577114e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 11995
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.15625,
      "epoch": 0.5796289137997681,
      "grad_norm": 0.0,
      "learning_rate": 2.1020970235794356e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 11996
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.125,
      "epoch": 0.5796772323154232,
      "grad_norm": 0.0,
      "learning_rate": 2.1018554310011597e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 11997
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.875,
      "epoch": 0.5797255508310785,
      "grad_norm": 0.0,
      "learning_rate": 2.1016138384228836e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 11998
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.0625,
      "epoch": 0.5797738693467337,
      "grad_norm": 0.0,
      "learning_rate": 2.1013722458446077e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 11999
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.4375,
      "epoch": 0.5798221878623888,
      "grad_norm": 0.0,
      "learning_rate": 2.1011306532663313e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12000
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 74.09375,
      "epoch": 0.579870506378044,
      "grad_norm": 178.1232590339401,
      "learning_rate": 2.1008890606880554e-07,
      "loss": 0.0,
      "reward": 1.484375,
      "reward_std": 0.15580293536186218,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12001
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 70.0,
      "epoch": 0.5799188248936993,
      "grad_norm": 10.74139436743851,
      "learning_rate": 2.1006474681097796e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.24831002950668335,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12002
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.40625,
      "epoch": 0.5799671434093545,
      "grad_norm": 0.0,
      "learning_rate": 2.1004058755315035e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 12003
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.25,
      "epoch": 0.5800154619250096,
      "grad_norm": 0.0,
      "learning_rate": 2.1001642829532276e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12004
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 46.96875,
      "epoch": 0.5800637804406649,
      "grad_norm": 0.0,
      "learning_rate": 2.0999226903749515e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 12005
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.65625,
      "epoch": 0.5801120989563201,
      "grad_norm": 0.0,
      "learning_rate": 2.0996810977966756e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12006
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 61.0,
      "epoch": 0.5801604174719752,
      "grad_norm": 0.0,
      "learning_rate": 2.0994395052183997e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12007
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.375,
      "epoch": 0.5802087359876305,
      "grad_norm": 0.0,
      "learning_rate": 2.0991979126401236e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12008
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.25,
      "epoch": 0.5802570545032857,
      "grad_norm": 0.0,
      "learning_rate": 2.0989563200618477e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12009
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.71875,
      "epoch": 0.5803053730189408,
      "grad_norm": 0.0,
      "learning_rate": 2.0987147274835713e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12010
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.96875,
      "epoch": 0.580353691534596,
      "grad_norm": 0.0,
      "learning_rate": 2.0984731349052955e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12011
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.0625,
      "epoch": 0.5804020100502513,
      "grad_norm": 0.0,
      "learning_rate": 2.0982315423270196e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12012
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.375,
      "epoch": 0.5804503285659065,
      "grad_norm": 0.0,
      "learning_rate": 2.0979899497487435e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12013
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.5625,
      "epoch": 0.5804986470815616,
      "grad_norm": 0.0,
      "learning_rate": 2.0977483571704676e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12014
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.6875,
      "epoch": 0.5805469655972169,
      "grad_norm": 0.0,
      "learning_rate": 2.0975067645921915e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12015
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 65.71875,
      "epoch": 0.5805952841128721,
      "grad_norm": 0.0,
      "learning_rate": 2.0972651720139156e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12016
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.5,
      "epoch": 0.5806436026285272,
      "grad_norm": 0.0,
      "learning_rate": 2.0970235794356398e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12017
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.46875,
      "epoch": 0.5806919211441824,
      "grad_norm": 0.0,
      "learning_rate": 2.0967819868573636e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12018
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.25,
      "epoch": 0.5807402396598377,
      "grad_norm": 0.0,
      "learning_rate": 2.0965403942790878e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12019
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.0625,
      "epoch": 0.5807885581754928,
      "grad_norm": 0.0,
      "learning_rate": 2.0962988017008117e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12020
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.0625,
      "epoch": 0.580836876691148,
      "grad_norm": 0.0,
      "learning_rate": 2.0960572091225355e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12021
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 44.90625,
      "epoch": 0.5808851952068033,
      "grad_norm": 0.0,
      "learning_rate": 2.0958156165442597e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12022
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.6875,
      "epoch": 0.5809335137224585,
      "grad_norm": 0.0,
      "learning_rate": 2.0955740239659835e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12023
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.8125,
      "epoch": 0.5809818322381136,
      "grad_norm": 0.0,
      "learning_rate": 2.0953324313877077e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12024
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.03125,
      "epoch": 0.5810301507537688,
      "grad_norm": 0.0,
      "learning_rate": 2.0950908388094315e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12025
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.625,
      "epoch": 0.5810784692694241,
      "grad_norm": 0.0,
      "learning_rate": 2.0948492462311557e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12026
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.6875,
      "epoch": 0.5811267877850792,
      "grad_norm": 0.0,
      "learning_rate": 2.0946076536528798e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12027
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.375,
      "epoch": 0.5811751063007344,
      "grad_norm": 0.0,
      "learning_rate": 2.0943660610746037e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12028
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.03125,
      "epoch": 0.5812234248163897,
      "grad_norm": 0.0,
      "learning_rate": 2.0941244684963278e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12029
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.625,
      "epoch": 0.5812717433320448,
      "grad_norm": 0.0,
      "learning_rate": 2.0938828759180517e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12030
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.4375,
      "epoch": 0.5813200618477,
      "grad_norm": 0.0,
      "learning_rate": 2.0936412833397756e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12031
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.84375,
      "epoch": 0.5813683803633553,
      "grad_norm": 0.0,
      "learning_rate": 2.0933996907614997e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12032
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.3125,
      "epoch": 0.5814166988790105,
      "grad_norm": 0.0,
      "learning_rate": 2.0931580981832236e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12033
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.71875,
      "epoch": 0.5814650173946656,
      "grad_norm": 0.0,
      "learning_rate": 2.0929165056049477e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12034
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.375,
      "epoch": 0.5815133359103208,
      "grad_norm": 0.0,
      "learning_rate": 2.0926749130266716e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12035
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.25,
      "epoch": 0.5815616544259761,
      "grad_norm": 0.0,
      "learning_rate": 2.0924333204483957e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880350112915,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12036
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.46875,
      "epoch": 0.5816099729416312,
      "grad_norm": 0.0,
      "learning_rate": 2.0921917278701198e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12037
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 60.0625,
      "epoch": 0.5816582914572864,
      "grad_norm": 10.896688519222156,
      "learning_rate": 2.0919501352918437e-07,
      "loss": -0.0,
      "reward": 2.359375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12038
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 54.125,
      "epoch": 0.5817066099729417,
      "grad_norm": 0.0,
      "learning_rate": 2.0917085427135679e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 12039
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.34375,
      "epoch": 0.5817549284885968,
      "grad_norm": 0.0,
      "learning_rate": 2.0914669501352917e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12040
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.46875,
      "epoch": 0.581803247004252,
      "grad_norm": 0.0,
      "learning_rate": 2.0912253575570159e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12041
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.53125,
      "epoch": 0.5818515655199072,
      "grad_norm": 0.0,
      "learning_rate": 2.0909837649787397e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12042
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.59375,
      "epoch": 0.5818998840355625,
      "grad_norm": 0.0,
      "learning_rate": 2.0907421724004636e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12043
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 53.25,
      "epoch": 0.5819482025512176,
      "grad_norm": 0.0,
      "learning_rate": 2.0905005798221877e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12044
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.15625,
      "epoch": 0.5819965210668728,
      "grad_norm": 0.0,
      "learning_rate": 2.0902589872439116e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12045
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.34375,
      "epoch": 0.5820448395825281,
      "grad_norm": 0.0,
      "learning_rate": 2.0900173946656358e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12046
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.875,
      "epoch": 0.5820931580981832,
      "grad_norm": 0.0,
      "learning_rate": 2.08977580208736e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12047
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.375,
      "epoch": 0.5821414766138384,
      "grad_norm": 0.0,
      "learning_rate": 2.0895342095090838e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12048
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.5,
      "epoch": 0.5821897951294936,
      "grad_norm": 0.0,
      "learning_rate": 2.089292616930808e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12049
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.8125,
      "epoch": 0.5822381136451488,
      "grad_norm": 0.0,
      "learning_rate": 2.0890510243525318e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12050
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.96875,
      "epoch": 0.582286432160804,
      "grad_norm": 0.0,
      "learning_rate": 2.088809431774256e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12051
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.96875,
      "epoch": 0.5823347506764592,
      "grad_norm": 0.0,
      "learning_rate": 2.0885678391959798e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12052
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.25,
      "epoch": 0.5823830691921145,
      "grad_norm": 0.0,
      "learning_rate": 2.0883262466177036e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12053
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.875,
      "epoch": 0.5824313877077696,
      "grad_norm": 0.0,
      "learning_rate": 2.0880846540394278e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12054
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.09375,
      "epoch": 0.5824797062234248,
      "grad_norm": 0.0,
      "learning_rate": 2.0878430614611517e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12055
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 57.9375,
      "epoch": 0.58252802473908,
      "grad_norm": 0.0,
      "learning_rate": 2.0876014688828758e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12056
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.34375,
      "epoch": 0.5825763432547352,
      "grad_norm": 0.0,
      "learning_rate": 2.0873598763046e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12057
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.5625,
      "epoch": 0.5826246617703904,
      "grad_norm": 0.0,
      "learning_rate": 2.0871182837263238e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12058
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.21875,
      "epoch": 0.5826729802860456,
      "grad_norm": 0.0,
      "learning_rate": 2.086876691148048e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12059
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.28125,
      "epoch": 0.5827212988017008,
      "grad_norm": 0.0,
      "learning_rate": 2.0866350985697718e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12060
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.59375,
      "epoch": 0.582769617317356,
      "grad_norm": 0.0,
      "learning_rate": 2.086393505991496e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12061
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.65625,
      "epoch": 0.5828179358330112,
      "grad_norm": 0.0,
      "learning_rate": 2.0861519134132198e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12062
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.1875,
      "epoch": 0.5828662543486665,
      "grad_norm": 0.0,
      "learning_rate": 2.0859103208349437e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12063
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.0,
      "epoch": 0.5829145728643216,
      "grad_norm": 0.0,
      "learning_rate": 2.0856687282566678e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12064
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.03125,
      "epoch": 0.5829628913799768,
      "grad_norm": 0.0,
      "learning_rate": 2.0854271356783917e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12065
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.875,
      "epoch": 0.583011209895632,
      "grad_norm": 0.0,
      "learning_rate": 2.0851855431001158e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12066
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.75,
      "epoch": 0.5830595284112872,
      "grad_norm": 0.0,
      "learning_rate": 2.08494395052184e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12067
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.84375,
      "epoch": 0.5831078469269424,
      "grad_norm": 0.0,
      "learning_rate": 2.0847023579435638e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12068
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.375,
      "epoch": 0.5831561654425976,
      "grad_norm": 0.0,
      "learning_rate": 2.084460765365288e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12069
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.46875,
      "epoch": 0.5832044839582528,
      "grad_norm": 0.0,
      "learning_rate": 2.0842191727870118e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12070
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.9375,
      "epoch": 0.583252802473908,
      "grad_norm": 0.0,
      "learning_rate": 2.083977580208736e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12071
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 52.40625,
      "epoch": 0.5833011209895632,
      "grad_norm": 0.0,
      "learning_rate": 2.08373598763046e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12072
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.875,
      "epoch": 0.5833494395052184,
      "grad_norm": 0.0,
      "learning_rate": 2.0834943950521837e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12073
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 56.78125,
      "epoch": 0.5833977580208736,
      "grad_norm": 0.0,
      "learning_rate": 2.0832528024739079e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12074
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.5,
      "epoch": 0.5834460765365288,
      "grad_norm": 0.0,
      "learning_rate": 2.0830112098956317e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12075
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.1875,
      "epoch": 0.583494395052184,
      "grad_norm": 0.0,
      "learning_rate": 2.082769617317356e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12076
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.84375,
      "epoch": 0.5835427135678392,
      "grad_norm": 0.0,
      "learning_rate": 2.08252802473908e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12077
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.21875,
      "epoch": 0.5835910320834944,
      "grad_norm": 0.0,
      "learning_rate": 2.082286432160804e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12078
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.9375,
      "epoch": 0.5836393505991496,
      "grad_norm": 0.0,
      "learning_rate": 2.082044839582528e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12079
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.34375,
      "epoch": 0.5836876691148047,
      "grad_norm": 0.0,
      "learning_rate": 2.081803247004252e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12080
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.71875,
      "epoch": 0.58373598763046,
      "grad_norm": 0.0,
      "learning_rate": 2.081561654425976e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12081
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.875,
      "epoch": 0.5837843061461152,
      "grad_norm": 0.0,
      "learning_rate": 2.0813200618477002e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12082
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.75,
      "epoch": 0.5838326246617704,
      "grad_norm": 0.0,
      "learning_rate": 2.0810784692694238e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12083
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.03125,
      "epoch": 0.5838809431774256,
      "grad_norm": 0.0,
      "learning_rate": 2.080836876691148e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12084
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.75,
      "epoch": 0.5839292616930808,
      "grad_norm": 0.0,
      "learning_rate": 2.0805952841128718e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12085
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 50.40625,
      "epoch": 0.583977580208736,
      "grad_norm": 0.0,
      "learning_rate": 2.080353691534596e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12086
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.6875,
      "epoch": 0.5840258987243911,
      "grad_norm": 0.0,
      "learning_rate": 2.08011209895632e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12087
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 43.0625,
      "epoch": 0.5840742172400464,
      "grad_norm": 0.0,
      "learning_rate": 2.079870506378044e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12088
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.28125,
      "epoch": 0.5841225357557016,
      "grad_norm": 0.0,
      "learning_rate": 2.079628913799768e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12089
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.65625,
      "epoch": 0.5841708542713567,
      "grad_norm": 0.0,
      "learning_rate": 2.079387321221492e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12090
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.0,
      "epoch": 0.584219172787012,
      "grad_norm": 0.0,
      "learning_rate": 2.079145728643216e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12091
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.0,
      "epoch": 0.5842674913026672,
      "grad_norm": 0.0,
      "learning_rate": 2.0789041360649402e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12092
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.71875,
      "epoch": 0.5843158098183224,
      "grad_norm": 0.0,
      "learning_rate": 2.0786625434866638e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12093
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.5,
      "epoch": 0.5843641283339776,
      "grad_norm": 0.0,
      "learning_rate": 2.078420950908388e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12094
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.03125,
      "epoch": 0.5844124468496328,
      "grad_norm": 0.0,
      "learning_rate": 2.0781793583301118e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12095
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 49.09375,
      "epoch": 0.584460765365288,
      "grad_norm": 0.0,
      "learning_rate": 2.077937765751836e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12096
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.28125,
      "epoch": 0.5845090838809431,
      "grad_norm": 0.0,
      "learning_rate": 2.07769617317356e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12097
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.84375,
      "epoch": 0.5845574023965984,
      "grad_norm": 0.0,
      "learning_rate": 2.077454580595284e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12098
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.6875,
      "epoch": 0.5846057209122536,
      "grad_norm": 0.0,
      "learning_rate": 2.077212988017008e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12099
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.0625,
      "epoch": 0.5846540394279087,
      "grad_norm": 0.0,
      "learning_rate": 2.076971395438732e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12100
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.34375,
      "epoch": 0.584702357943564,
      "grad_norm": 0.0,
      "learning_rate": 2.076729802860456e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12101
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.6875,
      "epoch": 0.5847506764592192,
      "grad_norm": 0.0,
      "learning_rate": 2.0764882102821802e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12102
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.65625,
      "epoch": 0.5847989949748744,
      "grad_norm": 0.0,
      "learning_rate": 2.0762466177039038e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12103
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.65625,
      "epoch": 0.5848473134905295,
      "grad_norm": 0.0,
      "learning_rate": 2.076005025125628e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12104
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 48.4375,
      "epoch": 0.5848956320061848,
      "grad_norm": 0.0,
      "learning_rate": 2.0757634325473519e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12105
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.90625,
      "epoch": 0.58494395052184,
      "grad_norm": 0.0,
      "learning_rate": 2.075521839969076e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12106
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.84375,
      "epoch": 0.5849922690374951,
      "grad_norm": 0.0,
      "learning_rate": 2.0752802473908e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12107
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.0,
      "epoch": 0.5850405875531504,
      "grad_norm": 0.0,
      "learning_rate": 2.075038654812524e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12108
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.71875,
      "epoch": 0.5850889060688056,
      "grad_norm": 0.0,
      "learning_rate": 2.0747970622342481e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12109
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.875,
      "epoch": 0.5851372245844607,
      "grad_norm": 0.0,
      "learning_rate": 2.074555469655972e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12110
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 46.34375,
      "epoch": 0.585185543100116,
      "grad_norm": 0.0,
      "learning_rate": 2.0743138770776961e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12111
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 53.28125,
      "epoch": 0.5852338616157712,
      "grad_norm": 0.0,
      "learning_rate": 2.0740722844994203e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 12112
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.5,
      "epoch": 0.5852821801314264,
      "grad_norm": 0.0,
      "learning_rate": 2.073830691921144e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12113
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.625,
      "epoch": 0.5853304986470815,
      "grad_norm": 0.0,
      "learning_rate": 2.073589099342868e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12114
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.0625,
      "epoch": 0.5853788171627368,
      "grad_norm": 0.0,
      "learning_rate": 2.073347506764592e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12115
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.96875,
      "epoch": 0.585427135678392,
      "grad_norm": 0.0,
      "learning_rate": 2.073105914186316e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12116
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.5,
      "epoch": 0.5854754541940471,
      "grad_norm": 0.0,
      "learning_rate": 2.0728643216080402e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12117
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.40625,
      "epoch": 0.5855237727097024,
      "grad_norm": 0.0,
      "learning_rate": 2.072622729029764e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12118
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.0625,
      "epoch": 0.5855720912253576,
      "grad_norm": 0.0,
      "learning_rate": 2.0723811364514882e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12119
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 45.59375,
      "epoch": 0.5856204097410127,
      "grad_norm": 0.0,
      "learning_rate": 2.072139543873212e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12120
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.75,
      "epoch": 0.5856687282566679,
      "grad_norm": 0.0,
      "learning_rate": 2.0718979512949362e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12121
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.90625,
      "epoch": 0.5857170467723232,
      "grad_norm": 0.0,
      "learning_rate": 2.0716563587166603e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12122
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.09375,
      "epoch": 0.5857653652879784,
      "grad_norm": 0.0,
      "learning_rate": 2.0714147661383842e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12123
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.1875,
      "epoch": 0.5858136838036335,
      "grad_norm": 0.0,
      "learning_rate": 2.071173173560108e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12124
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.53125,
      "epoch": 0.5858620023192888,
      "grad_norm": 0.0,
      "learning_rate": 2.070931580981832e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12125
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.125,
      "epoch": 0.585910320834944,
      "grad_norm": 0.0,
      "learning_rate": 2.070689988403556e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12126
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.96875,
      "epoch": 0.5859586393505991,
      "grad_norm": 0.0,
      "learning_rate": 2.0704483958252802e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12127
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 60.40625,
      "epoch": 0.5860069578662543,
      "grad_norm": 37.26205275610087,
      "learning_rate": 2.070206803247004e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12128
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.78125,
      "epoch": 0.5860552763819096,
      "grad_norm": 0.0,
      "learning_rate": 2.0699652106687282e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12129
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.28125,
      "epoch": 0.5861035948975647,
      "grad_norm": 0.0,
      "learning_rate": 2.069723618090452e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12130
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.6875,
      "epoch": 0.5861519134132199,
      "grad_norm": 0.0,
      "learning_rate": 2.0694820255121762e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12131
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.71875,
      "epoch": 0.5862002319288752,
      "grad_norm": 0.0,
      "learning_rate": 2.0692404329339004e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12132
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.28125,
      "epoch": 0.5862485504445304,
      "grad_norm": 0.0,
      "learning_rate": 2.0689988403556242e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12133
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.75,
      "epoch": 0.5862968689601855,
      "grad_norm": 0.0,
      "learning_rate": 2.068757247777348e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12134
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.59375,
      "epoch": 0.5863451874758407,
      "grad_norm": 0.0,
      "learning_rate": 2.068515655199072e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12135
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.25,
      "epoch": 0.586393505991496,
      "grad_norm": 0.0,
      "learning_rate": 2.068274062620796e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12136
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.71875,
      "epoch": 0.5864418245071511,
      "grad_norm": 0.0,
      "learning_rate": 2.0680324700425202e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12137
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.15625,
      "epoch": 0.5864901430228063,
      "grad_norm": 0.0,
      "learning_rate": 2.067790877464244e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12138
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.75,
      "epoch": 0.5865384615384616,
      "grad_norm": 0.0,
      "learning_rate": 2.0675492848859683e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12139
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.6875,
      "epoch": 0.5865867800541167,
      "grad_norm": 0.0,
      "learning_rate": 2.067307692307692e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12140
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 48.15625,
      "epoch": 0.5866350985697719,
      "grad_norm": 0.0,
      "learning_rate": 2.0670660997294163e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12141
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 68.0,
      "epoch": 0.5866834170854272,
      "grad_norm": 0.0,
      "learning_rate": 2.0668245071511404e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 12142
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.96875,
      "epoch": 0.5867317356010824,
      "grad_norm": 0.0,
      "learning_rate": 2.0665829145728643e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12143
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 55.09375,
      "epoch": 0.5867800541167375,
      "grad_norm": 0.0,
      "learning_rate": 2.0663413219945884e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 12144
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.21875,
      "epoch": 0.5868283726323927,
      "grad_norm": 0.0,
      "learning_rate": 2.066099729416312e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12145
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.09375,
      "epoch": 0.586876691148048,
      "grad_norm": 0.0,
      "learning_rate": 2.0658581368380361e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 12146
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.65625,
      "epoch": 0.5869250096637031,
      "grad_norm": 0.0,
      "learning_rate": 2.0656165442597603e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12147
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.03125,
      "epoch": 0.5869733281793583,
      "grad_norm": 0.0,
      "learning_rate": 2.0653749516814842e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12148
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.8125,
      "epoch": 0.5870216466950136,
      "grad_norm": 0.0,
      "learning_rate": 2.0651333591032083e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12149
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.90625,
      "epoch": 0.5870699652106687,
      "grad_norm": 0.0,
      "learning_rate": 2.0648917665249322e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12150
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.0,
      "epoch": 0.5871182837263239,
      "grad_norm": 0.0,
      "learning_rate": 2.0646501739466563e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12151
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 53.25,
      "epoch": 0.5871666022419791,
      "grad_norm": 19.965202563302213,
      "learning_rate": 2.0644085813683804e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12152
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.65625,
      "epoch": 0.5872149207576344,
      "grad_norm": 0.0,
      "learning_rate": 2.0641669887901043e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12153
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.59375,
      "epoch": 0.5872632392732895,
      "grad_norm": 0.0,
      "learning_rate": 2.0639253962118284e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12154
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.3125,
      "epoch": 0.5873115577889447,
      "grad_norm": 0.0,
      "learning_rate": 2.063683803633552e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12155
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.9375,
      "epoch": 0.5873598763046,
      "grad_norm": 0.0,
      "learning_rate": 2.0634422110552762e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12156
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.25,
      "epoch": 0.5874081948202551,
      "grad_norm": 0.0,
      "learning_rate": 2.0632006184770003e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12157
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.03125,
      "epoch": 0.5874565133359103,
      "grad_norm": 0.0,
      "learning_rate": 2.0629590258987242e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12158
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.65625,
      "epoch": 0.5875048318515655,
      "grad_norm": 0.0,
      "learning_rate": 2.0627174333204483e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12159
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.71875,
      "epoch": 0.5875531503672207,
      "grad_norm": 0.0,
      "learning_rate": 2.0624758407421722e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12160
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.5625,
      "epoch": 0.5876014688828759,
      "grad_norm": 0.0,
      "learning_rate": 2.0622342481638963e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12161
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.78125,
      "epoch": 0.5876497873985311,
      "grad_norm": 0.0,
      "learning_rate": 2.0619926555856205e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12162
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.15625,
      "epoch": 0.5876981059141864,
      "grad_norm": 0.0,
      "learning_rate": 2.0617510630073443e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12163
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.46875,
      "epoch": 0.5877464244298415,
      "grad_norm": 0.0,
      "learning_rate": 2.0615094704290685e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 12164
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.5625,
      "epoch": 0.5877947429454967,
      "grad_norm": 0.0,
      "learning_rate": 2.061267877850792e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12165
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.53125,
      "epoch": 0.587843061461152,
      "grad_norm": 0.0,
      "learning_rate": 2.0610262852725162e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12166
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.15625,
      "epoch": 0.5878913799768071,
      "grad_norm": 0.0,
      "learning_rate": 2.0607846926942404e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12167
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.03125,
      "epoch": 0.5879396984924623,
      "grad_norm": 0.0,
      "learning_rate": 2.0605431001159642e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12168
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.5,
      "epoch": 0.5879880170081175,
      "grad_norm": 0.0,
      "learning_rate": 2.0603015075376884e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12169
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.40625,
      "epoch": 0.5880363355237727,
      "grad_norm": 0.0,
      "learning_rate": 2.0600599149594122e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12170
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.09375,
      "epoch": 0.5880846540394279,
      "grad_norm": 0.0,
      "learning_rate": 2.0598183223811364e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12171
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.28125,
      "epoch": 0.5881329725550831,
      "grad_norm": 0.0,
      "learning_rate": 2.0595767298028605e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12172
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.96875,
      "epoch": 0.5881812910707384,
      "grad_norm": 0.0,
      "learning_rate": 2.0593351372245844e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12173
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.75,
      "epoch": 0.5882296095863935,
      "grad_norm": 0.0,
      "learning_rate": 2.0590935446463085e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12174
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.0625,
      "epoch": 0.5882779281020487,
      "grad_norm": 0.0,
      "learning_rate": 2.058851952068032e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12175
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.34375,
      "epoch": 0.5883262466177039,
      "grad_norm": 0.0,
      "learning_rate": 2.0586103594897563e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12176
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.28125,
      "epoch": 0.5883745651333591,
      "grad_norm": 0.0,
      "learning_rate": 2.0583687669114804e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12177
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.21875,
      "epoch": 0.5884228836490143,
      "grad_norm": 0.0,
      "learning_rate": 2.0581271743332043e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12178
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.03125,
      "epoch": 0.5884712021646695,
      "grad_norm": 0.0,
      "learning_rate": 2.0578855817549284e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12179
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.0625,
      "epoch": 0.5885195206803246,
      "grad_norm": 0.0,
      "learning_rate": 2.0576439891766523e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12180
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 46.96875,
      "epoch": 0.5885678391959799,
      "grad_norm": 0.0,
      "learning_rate": 2.0574023965983764e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12181
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 46.40625,
      "epoch": 0.5886161577116351,
      "grad_norm": 0.0,
      "learning_rate": 2.0571608040201006e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12182
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 51.59375,
      "epoch": 0.5886644762272903,
      "grad_norm": 0.0,
      "learning_rate": 2.0569192114418244e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12183
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.25,
      "epoch": 0.5887127947429455,
      "grad_norm": 0.0,
      "learning_rate": 2.0566776188635486e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12184
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.0,
      "epoch": 0.5887611132586007,
      "grad_norm": 0.0,
      "learning_rate": 2.0564360262852722e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12185
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.71875,
      "epoch": 0.5888094317742559,
      "grad_norm": 0.0,
      "learning_rate": 2.0561944337069963e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12186
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.53125,
      "epoch": 0.588857750289911,
      "grad_norm": 0.0,
      "learning_rate": 2.0559528411287204e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12187
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.28125,
      "epoch": 0.5889060688055663,
      "grad_norm": 0.0,
      "learning_rate": 2.0557112485504443e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12188
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.46875,
      "epoch": 0.5889543873212215,
      "grad_norm": 0.0,
      "learning_rate": 2.0554696559721684e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12189
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.1875,
      "epoch": 0.5890027058368766,
      "grad_norm": 0.0,
      "learning_rate": 2.0552280633938923e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12190
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 57.875,
      "epoch": 0.5890510243525319,
      "grad_norm": 72.83939178970441,
      "learning_rate": 2.0549864708156165e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.31963658332824707,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12191
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.03125,
      "epoch": 0.5890993428681871,
      "grad_norm": 0.0,
      "learning_rate": 2.0547448782373406e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12192
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.3125,
      "epoch": 0.5891476613838423,
      "grad_norm": 0.0,
      "learning_rate": 2.0545032856590645e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12193
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.15625,
      "epoch": 0.5891959798994975,
      "grad_norm": 0.0,
      "learning_rate": 2.0542616930807886e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12194
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.59375,
      "epoch": 0.5892442984151527,
      "grad_norm": 0.0,
      "learning_rate": 2.0540201005025125e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12195
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.46875,
      "epoch": 0.5892926169308079,
      "grad_norm": 0.0,
      "learning_rate": 2.0537785079242363e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12196
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.375,
      "epoch": 0.589340935446463,
      "grad_norm": 0.0,
      "learning_rate": 2.0535369153459605e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12197
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.59375,
      "epoch": 0.5893892539621183,
      "grad_norm": 0.0,
      "learning_rate": 2.0532953227676844e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12198
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.46875,
      "epoch": 0.5894375724777735,
      "grad_norm": 0.0,
      "learning_rate": 2.0530537301894085e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12199
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.78125,
      "epoch": 0.5894858909934286,
      "grad_norm": 0.0,
      "learning_rate": 2.0528121376111324e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12200
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.125,
      "epoch": 0.5895342095090839,
      "grad_norm": 0.0,
      "learning_rate": 2.0525705450328565e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12201
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.71875,
      "epoch": 0.5895825280247391,
      "grad_norm": 0.0,
      "learning_rate": 2.0523289524545806e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12202
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.3125,
      "epoch": 0.5896308465403943,
      "grad_norm": 0.0,
      "learning_rate": 2.0520873598763045e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12203
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.28125,
      "epoch": 0.5896791650560494,
      "grad_norm": 0.0,
      "learning_rate": 2.0518457672980286e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12204
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.6875,
      "epoch": 0.5897274835717047,
      "grad_norm": 0.0,
      "learning_rate": 2.0516041747197525e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12205
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.65625,
      "epoch": 0.5897758020873599,
      "grad_norm": 0.0,
      "learning_rate": 2.0513625821414764e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12206
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.5,
      "epoch": 0.589824120603015,
      "grad_norm": 0.0,
      "learning_rate": 2.0511209895632005e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12207
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.40625,
      "epoch": 0.5898724391186703,
      "grad_norm": 0.0,
      "learning_rate": 2.0508793969849244e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12208
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.6875,
      "epoch": 0.5899207576343255,
      "grad_norm": 0.0,
      "learning_rate": 2.0506378044066485e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12209
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.5,
      "epoch": 0.5899690761499806,
      "grad_norm": 0.0,
      "learning_rate": 2.0503962118283724e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12210
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.875,
      "epoch": 0.5900173946656359,
      "grad_norm": 0.0,
      "learning_rate": 2.0501546192500965e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12211
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 48.53125,
      "epoch": 0.5900657131812911,
      "grad_norm": 0.0,
      "learning_rate": 2.0499130266718207e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12212
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.65625,
      "epoch": 0.5901140316969463,
      "grad_norm": 0.0,
      "learning_rate": 2.0496714340935445e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12213
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.3125,
      "epoch": 0.5901623502126014,
      "grad_norm": 0.0,
      "learning_rate": 2.0494298415152687e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12214
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.34375,
      "epoch": 0.5902106687282567,
      "grad_norm": 0.0,
      "learning_rate": 2.0491882489369925e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12215
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.65625,
      "epoch": 0.5902589872439119,
      "grad_norm": 0.0,
      "learning_rate": 2.0489466563587164e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12216
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.71875,
      "epoch": 0.590307305759567,
      "grad_norm": 0.0,
      "learning_rate": 2.0487050637804406e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12217
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.125,
      "epoch": 0.5903556242752223,
      "grad_norm": 0.0,
      "learning_rate": 2.0484634712021644e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12218
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.6875,
      "epoch": 0.5904039427908775,
      "grad_norm": 0.0,
      "learning_rate": 2.0482218786238886e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12219
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.84375,
      "epoch": 0.5904522613065326,
      "grad_norm": 0.0,
      "learning_rate": 2.0479802860456124e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12220
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.5,
      "epoch": 0.5905005798221878,
      "grad_norm": 0.0,
      "learning_rate": 2.0477386934673366e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12221
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.0,
      "epoch": 0.5905488983378431,
      "grad_norm": 0.0,
      "learning_rate": 2.0474971008890607e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12222
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.375,
      "epoch": 0.5905972168534983,
      "grad_norm": 0.0,
      "learning_rate": 2.0472555083107846e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12223
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.84375,
      "epoch": 0.5906455353691534,
      "grad_norm": 0.0,
      "learning_rate": 2.0470139157325087e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12224
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 47.625,
      "epoch": 0.5906938538848087,
      "grad_norm": 0.0,
      "learning_rate": 2.0467723231542326e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12225
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 46.40625,
      "epoch": 0.5907421724004639,
      "grad_norm": 0.0,
      "learning_rate": 2.0465307305759567e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 12226
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.65625,
      "epoch": 0.590790490916119,
      "grad_norm": 0.0,
      "learning_rate": 2.0462891379976806e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880052089691,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12227
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.03125,
      "epoch": 0.5908388094317742,
      "grad_norm": 0.0,
      "learning_rate": 2.0460475454194045e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12228
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.0,
      "epoch": 0.5908871279474295,
      "grad_norm": 0.0,
      "learning_rate": 2.0458059528411286e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12229
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.8125,
      "epoch": 0.5909354464630846,
      "grad_norm": 0.0,
      "learning_rate": 2.0455643602628525e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12230
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 55.53125,
      "epoch": 0.5909837649787398,
      "grad_norm": 0.0,
      "learning_rate": 2.0453227676845766e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12231
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.28125,
      "epoch": 0.5910320834943951,
      "grad_norm": 0.0,
      "learning_rate": 2.0450811751063007e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12232
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.65625,
      "epoch": 0.5910804020100503,
      "grad_norm": 0.0,
      "learning_rate": 2.0448395825280246e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12233
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.34375,
      "epoch": 0.5911287205257054,
      "grad_norm": 0.0,
      "learning_rate": 2.0445979899497488e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12234
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.34375,
      "epoch": 0.5911770390413607,
      "grad_norm": 0.0,
      "learning_rate": 2.0443563973714726e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12235
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.8125,
      "epoch": 0.5912253575570159,
      "grad_norm": 0.0,
      "learning_rate": 2.0441148047931968e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12236
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.03125,
      "epoch": 0.591273676072671,
      "grad_norm": 0.0,
      "learning_rate": 2.0438732122149206e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12237
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.0625,
      "epoch": 0.5913219945883262,
      "grad_norm": 0.0,
      "learning_rate": 2.0436316196366445e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12238
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.09375,
      "epoch": 0.5913703131039815,
      "grad_norm": 0.0,
      "learning_rate": 2.0433900270583686e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12239
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.34375,
      "epoch": 0.5914186316196366,
      "grad_norm": 0.0,
      "learning_rate": 2.0431484344800925e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12240
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.15625,
      "epoch": 0.5914669501352918,
      "grad_norm": 0.0,
      "learning_rate": 2.0429068419018167e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12241
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 43.96875,
      "epoch": 0.5915152686509471,
      "grad_norm": 0.0,
      "learning_rate": 2.0426652493235408e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12242
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.03125,
      "epoch": 0.5915635871666023,
      "grad_norm": 0.0,
      "learning_rate": 2.0424236567452647e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12243
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.40625,
      "epoch": 0.5916119056822574,
      "grad_norm": 0.0,
      "learning_rate": 2.0421820641669888e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12244
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.09375,
      "epoch": 0.5916602241979126,
      "grad_norm": 0.0,
      "learning_rate": 2.0419404715887127e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12245
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.625,
      "epoch": 0.5917085427135679,
      "grad_norm": 0.0,
      "learning_rate": 2.0416988790104368e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12246
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.15625,
      "epoch": 0.591756861229223,
      "grad_norm": 0.0,
      "learning_rate": 2.0414572864321607e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12247
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.6875,
      "epoch": 0.5918051797448782,
      "grad_norm": 0.0,
      "learning_rate": 2.0412156938538845e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12248
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.6875,
      "epoch": 0.5918534982605335,
      "grad_norm": 0.0,
      "learning_rate": 2.0409741012756087e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12249
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 60.71875,
      "epoch": 0.5919018167761886,
      "grad_norm": 0.0,
      "learning_rate": 2.0407325086973326e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 12250
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.90625,
      "epoch": 0.5919501352918438,
      "grad_norm": 0.0,
      "learning_rate": 2.0404909161190567e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12251
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.59375,
      "epoch": 0.591998453807499,
      "grad_norm": 0.0,
      "learning_rate": 2.0402493235407808e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12252
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.125,
      "epoch": 0.5920467723231543,
      "grad_norm": 0.0,
      "learning_rate": 2.0400077309625047e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12253
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.3125,
      "epoch": 0.5920950908388094,
      "grad_norm": 0.0,
      "learning_rate": 2.0397661383842288e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12254
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.78125,
      "epoch": 0.5921434093544646,
      "grad_norm": 0.0,
      "learning_rate": 2.0395245458059527e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12255
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.03125,
      "epoch": 0.5921917278701199,
      "grad_norm": 0.0,
      "learning_rate": 2.0392829532276768e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12256
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 54.21875,
      "epoch": 0.592240046385775,
      "grad_norm": 0.0,
      "learning_rate": 2.039041360649401e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 12257
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.59375,
      "epoch": 0.5922883649014302,
      "grad_norm": 0.0,
      "learning_rate": 2.0387997680711246e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12258
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.46875,
      "epoch": 0.5923366834170855,
      "grad_norm": 0.0,
      "learning_rate": 2.0385581754928487e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12259
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.9375,
      "epoch": 0.5923850019327406,
      "grad_norm": 0.0,
      "learning_rate": 2.0383165829145726e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12260
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.28125,
      "epoch": 0.5924333204483958,
      "grad_norm": 0.0,
      "learning_rate": 2.0380749903362967e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 12261
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.0,
      "epoch": 0.592481638964051,
      "grad_norm": 0.0,
      "learning_rate": 2.0378333977580209e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12262
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.15625,
      "epoch": 0.5925299574797063,
      "grad_norm": 0.0,
      "learning_rate": 2.0375918051797447e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 12263
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.03125,
      "epoch": 0.5925782759953614,
      "grad_norm": 0.0,
      "learning_rate": 2.037350212601469e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12264
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.78125,
      "epoch": 0.5926265945110166,
      "grad_norm": 0.0,
      "learning_rate": 2.0371086200231927e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12265
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.15625,
      "epoch": 0.5926749130266719,
      "grad_norm": 0.0,
      "learning_rate": 2.036867027444917e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12266
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.65625,
      "epoch": 0.592723231542327,
      "grad_norm": 0.0,
      "learning_rate": 2.036625434866641e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12267
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.59375,
      "epoch": 0.5927715500579822,
      "grad_norm": 0.0,
      "learning_rate": 2.0363838422883646e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12268
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.34375,
      "epoch": 0.5928198685736374,
      "grad_norm": 0.0,
      "learning_rate": 2.0361422497100888e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12269
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.4375,
      "epoch": 0.5928681870892926,
      "grad_norm": 0.0,
      "learning_rate": 2.0359006571318126e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.260896235704422,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12270
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.3125,
      "epoch": 0.5929165056049478,
      "grad_norm": 0.0,
      "learning_rate": 2.0356590645535368e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12271
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.78125,
      "epoch": 0.592964824120603,
      "grad_norm": 0.0,
      "learning_rate": 2.035417471975261e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12272
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.03125,
      "epoch": 0.5930131426362583,
      "grad_norm": 0.0,
      "learning_rate": 2.0351758793969848e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12273
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.46875,
      "epoch": 0.5930614611519134,
      "grad_norm": 0.0,
      "learning_rate": 2.034934286818709e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12274
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.03125,
      "epoch": 0.5931097796675686,
      "grad_norm": 0.0,
      "learning_rate": 2.0346926942404328e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12275
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 46.46875,
      "epoch": 0.5931580981832238,
      "grad_norm": 0.0,
      "learning_rate": 2.034451101662157e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12276
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.96875,
      "epoch": 0.593206416698879,
      "grad_norm": 0.0,
      "learning_rate": 2.034209509083881e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12277
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 58.6875,
      "epoch": 0.5932547352145342,
      "grad_norm": 0.0,
      "learning_rate": 2.0339679165056047e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 12278
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 58.90625,
      "epoch": 0.5933030537301894,
      "grad_norm": 0.0,
      "learning_rate": 2.0337263239273288e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 12279
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.90625,
      "epoch": 0.5933513722458446,
      "grad_norm": 0.0,
      "learning_rate": 2.0334847313490527e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.2651395797729492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12280
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.625,
      "epoch": 0.5933996907614998,
      "grad_norm": 0.0,
      "learning_rate": 2.0332431387707768e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12281
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.125,
      "epoch": 0.593448009277155,
      "grad_norm": 0.0,
      "learning_rate": 2.033001546192501e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12282
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.28125,
      "epoch": 0.5934963277928103,
      "grad_norm": 0.0,
      "learning_rate": 2.0327599536142248e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12283
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.96875,
      "epoch": 0.5935446463084654,
      "grad_norm": 0.0,
      "learning_rate": 2.032518361035949e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12284
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.53125,
      "epoch": 0.5935929648241206,
      "grad_norm": 0.0,
      "learning_rate": 2.0322767684576728e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12285
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.3125,
      "epoch": 0.5936412833397758,
      "grad_norm": 0.0,
      "learning_rate": 2.032035175879397e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12286
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.15625,
      "epoch": 0.593689601855431,
      "grad_norm": 0.0,
      "learning_rate": 2.031793583301121e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12287
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.15625,
      "epoch": 0.5937379203710862,
      "grad_norm": 0.0,
      "learning_rate": 2.0315519907228447e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12288
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.625,
      "epoch": 0.5937862388867414,
      "grad_norm": 0.0,
      "learning_rate": 2.0313103981445688e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12289
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.84375,
      "epoch": 0.5938345574023965,
      "grad_norm": 0.0,
      "learning_rate": 2.0310688055662927e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12290
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 47.75,
      "epoch": 0.5938828759180518,
      "grad_norm": 0.0,
      "learning_rate": 2.0308272129880168e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12291
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.625,
      "epoch": 0.593931194433707,
      "grad_norm": 0.0,
      "learning_rate": 2.030585620409741e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12292
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.625,
      "epoch": 0.5939795129493622,
      "grad_norm": 0.0,
      "learning_rate": 2.0303440278314649e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12293
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.65625,
      "epoch": 0.5940278314650174,
      "grad_norm": 0.0,
      "learning_rate": 2.030102435253189e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12294
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.40625,
      "epoch": 0.5940761499806726,
      "grad_norm": 0.0,
      "learning_rate": 2.0298608426749129e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12295
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.65625,
      "epoch": 0.5941244684963278,
      "grad_norm": 0.0,
      "learning_rate": 2.029619250096637e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12296
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.34375,
      "epoch": 0.594172787011983,
      "grad_norm": 0.0,
      "learning_rate": 2.0293776575183611e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12297
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.5625,
      "epoch": 0.5942211055276382,
      "grad_norm": 0.0,
      "learning_rate": 2.029136064940085e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12298
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.9375,
      "epoch": 0.5942694240432934,
      "grad_norm": 0.0,
      "learning_rate": 2.028894472361809e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12299
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.1875,
      "epoch": 0.5943177425589485,
      "grad_norm": 0.0,
      "learning_rate": 2.0286528797835328e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12300
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.625,
      "epoch": 0.5943660610746038,
      "grad_norm": 0.0,
      "learning_rate": 2.028411287205257e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12301
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.03125,
      "epoch": 0.594414379590259,
      "grad_norm": 0.0,
      "learning_rate": 2.028169694626981e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12302
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 53.0,
      "epoch": 0.5944626981059142,
      "grad_norm": 0.0,
      "learning_rate": 2.027928102048705e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12303
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 60.25,
      "epoch": 0.5945110166215694,
      "grad_norm": 384.43483421811885,
      "learning_rate": 2.027686509470429e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.17598573863506317,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12304
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.40625,
      "epoch": 0.5945593351372246,
      "grad_norm": 0.0,
      "learning_rate": 2.027444916892153e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12305
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.40625,
      "epoch": 0.5946076536528798,
      "grad_norm": 0.0,
      "learning_rate": 2.027203324313877e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12306
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 51.40625,
      "epoch": 0.5946559721685349,
      "grad_norm": 0.0,
      "learning_rate": 2.0269617317356012e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 12307
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.71875,
      "epoch": 0.5947042906841902,
      "grad_norm": 0.0,
      "learning_rate": 2.026720139157325e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12308
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.21875,
      "epoch": 0.5947526091998454,
      "grad_norm": 0.0,
      "learning_rate": 2.026478546579049e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12309
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.15625,
      "epoch": 0.5948009277155005,
      "grad_norm": 0.0,
      "learning_rate": 2.0262369540007728e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12310
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.25,
      "epoch": 0.5948492462311558,
      "grad_norm": 0.0,
      "learning_rate": 2.025995361422497e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12311
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.78125,
      "epoch": 0.594897564746811,
      "grad_norm": 0.0,
      "learning_rate": 2.025753768844221e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12312
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.34375,
      "epoch": 0.5949458832624662,
      "grad_norm": 0.0,
      "learning_rate": 2.025512176265945e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12313
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 51.90625,
      "epoch": 0.5949942017781213,
      "grad_norm": 0.0,
      "learning_rate": 2.025270583687669e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 12314
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.375,
      "epoch": 0.5950425202937766,
      "grad_norm": 0.0,
      "learning_rate": 2.025028991109393e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12315
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 56.21875,
      "epoch": 0.5950908388094318,
      "grad_norm": 0.0,
      "learning_rate": 2.024787398531117e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 12316
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.03125,
      "epoch": 0.5951391573250869,
      "grad_norm": 0.0,
      "learning_rate": 2.0245458059528412e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12317
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 52.71875,
      "epoch": 0.5951874758407422,
      "grad_norm": 0.0,
      "learning_rate": 2.024304213374565e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 12318
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 51.34375,
      "epoch": 0.5952357943563974,
      "grad_norm": 0.0,
      "learning_rate": 2.024062620796289e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 12319
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.40625,
      "epoch": 0.5952841128720525,
      "grad_norm": 0.0,
      "learning_rate": 2.0238210282180128e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12320
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.0,
      "epoch": 0.5953324313877078,
      "grad_norm": 0.0,
      "learning_rate": 2.023579435639737e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12321
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 47.25,
      "epoch": 0.595380749903363,
      "grad_norm": 0.0,
      "learning_rate": 2.023337843061461e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12322
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.75,
      "epoch": 0.5954290684190182,
      "grad_norm": 0.0,
      "learning_rate": 2.023096250483185e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12323
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.9375,
      "epoch": 0.5954773869346733,
      "grad_norm": 0.0,
      "learning_rate": 2.022854657904909e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12324
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 52.9375,
      "epoch": 0.5955257054503286,
      "grad_norm": 0.0,
      "learning_rate": 2.022613065326633e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 12325
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.0,
      "epoch": 0.5955740239659838,
      "grad_norm": 0.0,
      "learning_rate": 2.022371472748357e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 12326
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.46875,
      "epoch": 0.5956223424816389,
      "grad_norm": 0.0,
      "learning_rate": 2.0221298801700813e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12327
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.25,
      "epoch": 0.5956706609972942,
      "grad_norm": 0.0,
      "learning_rate": 2.021888287591805e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12328
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 64.40625,
      "epoch": 0.5957189795129494,
      "grad_norm": 0.0,
      "learning_rate": 2.0216466950135293e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 12329
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.375,
      "epoch": 0.5957672980286045,
      "grad_norm": 0.0,
      "learning_rate": 2.021405102435253e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12330
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.46875,
      "epoch": 0.5958156165442597,
      "grad_norm": 0.0,
      "learning_rate": 2.021163509856977e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12331
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 53.4375,
      "epoch": 0.595863935059915,
      "grad_norm": 0.0,
      "learning_rate": 2.0209219172787011e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12332
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.8125,
      "epoch": 0.5959122535755702,
      "grad_norm": 0.0,
      "learning_rate": 2.020680324700425e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12333
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.8125,
      "epoch": 0.5959605720912253,
      "grad_norm": 0.0,
      "learning_rate": 2.0204387321221492e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12334
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.5625,
      "epoch": 0.5960088906068806,
      "grad_norm": 0.0,
      "learning_rate": 2.020197139543873e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12335
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.90625,
      "epoch": 0.5960572091225358,
      "grad_norm": 0.0,
      "learning_rate": 2.0199555469655972e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12336
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.15625,
      "epoch": 0.5961055276381909,
      "grad_norm": 0.0,
      "learning_rate": 2.0197139543873213e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12337
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.03125,
      "epoch": 0.5961538461538461,
      "grad_norm": 0.0,
      "learning_rate": 2.0194723618090452e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12338
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.09375,
      "epoch": 0.5962021646695014,
      "grad_norm": 0.0,
      "learning_rate": 2.0192307692307693e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12339
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.1875,
      "epoch": 0.5962504831851565,
      "grad_norm": 0.0,
      "learning_rate": 2.018989176652493e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12340
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.5,
      "epoch": 0.5962988017008117,
      "grad_norm": 0.0,
      "learning_rate": 2.018747584074217e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12341
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 49.53125,
      "epoch": 0.596347120216467,
      "grad_norm": 0.0,
      "learning_rate": 2.0185059914959412e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12342
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.5,
      "epoch": 0.5963954387321222,
      "grad_norm": 0.0,
      "learning_rate": 2.018264398917665e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12343
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.625,
      "epoch": 0.5964437572477773,
      "grad_norm": 0.0,
      "learning_rate": 2.0180228063393892e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12344
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.875,
      "epoch": 0.5964920757634325,
      "grad_norm": 0.0,
      "learning_rate": 2.017781213761113e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12345
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.28125,
      "epoch": 0.5965403942790878,
      "grad_norm": 0.0,
      "learning_rate": 2.0175396211828372e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12346
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.59375,
      "epoch": 0.5965887127947429,
      "grad_norm": 0.0,
      "learning_rate": 2.0172980286045613e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12347
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 56.40625,
      "epoch": 0.5966370313103981,
      "grad_norm": 53.65490684794717,
      "learning_rate": 2.0170564360262852e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12348
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.125,
      "epoch": 0.5966853498260534,
      "grad_norm": 0.0,
      "learning_rate": 2.0168148434480093e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12349
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.25,
      "epoch": 0.5967336683417085,
      "grad_norm": 0.0,
      "learning_rate": 2.016573250869733e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 12350
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.6875,
      "epoch": 0.5967819868573637,
      "grad_norm": 0.0,
      "learning_rate": 2.016331658291457e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12351
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.0,
      "epoch": 0.596830305373019,
      "grad_norm": 0.0,
      "learning_rate": 2.0160900657131812e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12352
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.5,
      "epoch": 0.5968786238886742,
      "grad_norm": 0.0,
      "learning_rate": 2.015848473134905e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12353
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.9375,
      "epoch": 0.5969269424043293,
      "grad_norm": 0.0,
      "learning_rate": 2.0156068805566292e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12354
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.625,
      "epoch": 0.5969752609199845,
      "grad_norm": 0.0,
      "learning_rate": 2.015365287978353e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12355
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.21875,
      "epoch": 0.5970235794356398,
      "grad_norm": 0.0,
      "learning_rate": 2.0151236954000772e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12356
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.75,
      "epoch": 0.5970718979512949,
      "grad_norm": 0.0,
      "learning_rate": 2.0148821028218014e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12357
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.53125,
      "epoch": 0.5971202164669501,
      "grad_norm": 0.0,
      "learning_rate": 2.0146405102435252e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12358
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.21875,
      "epoch": 0.5971685349826054,
      "grad_norm": 0.0,
      "learning_rate": 2.0143989176652494e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12359
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.40625,
      "epoch": 0.5972168534982605,
      "grad_norm": 0.0,
      "learning_rate": 2.014157325086973e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12360
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 58.5625,
      "epoch": 0.5972651720139157,
      "grad_norm": 0.0,
      "learning_rate": 2.013915732508697e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12361
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.75,
      "epoch": 0.597313490529571,
      "grad_norm": 0.0,
      "learning_rate": 2.0136741399304213e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12362
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.28125,
      "epoch": 0.5973618090452262,
      "grad_norm": 0.0,
      "learning_rate": 2.013432547352145e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12363
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.75,
      "epoch": 0.5974101275608813,
      "grad_norm": 0.0,
      "learning_rate": 2.0131909547738693e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12364
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.90625,
      "epoch": 0.5974584460765365,
      "grad_norm": 0.0,
      "learning_rate": 2.0129493621955931e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12365
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.8125,
      "epoch": 0.5975067645921918,
      "grad_norm": 0.0,
      "learning_rate": 2.0127077696173173e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12366
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.15625,
      "epoch": 0.5975550831078469,
      "grad_norm": 0.0,
      "learning_rate": 2.0124661770390414e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 12367
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.96875,
      "epoch": 0.5976034016235021,
      "grad_norm": 0.0,
      "learning_rate": 2.0122245844607653e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12368
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.84375,
      "epoch": 0.5976517201391573,
      "grad_norm": 0.0,
      "learning_rate": 2.0119829918824894e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12369
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.5,
      "epoch": 0.5977000386548125,
      "grad_norm": 0.0,
      "learning_rate": 2.0117413993042133e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12370
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.0,
      "epoch": 0.5977483571704677,
      "grad_norm": 0.0,
      "learning_rate": 2.0114998067259372e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12371
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.75,
      "epoch": 0.5977966756861229,
      "grad_norm": 0.0,
      "learning_rate": 2.0112582141476613e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12372
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.25,
      "epoch": 0.5978449942017782,
      "grad_norm": 0.0,
      "learning_rate": 2.0110166215693852e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12373
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.03125,
      "epoch": 0.5978933127174333,
      "grad_norm": 0.0,
      "learning_rate": 2.0107750289911093e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12374
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 56.8125,
      "epoch": 0.5979416312330885,
      "grad_norm": 0.0,
      "learning_rate": 2.0105334364128332e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12375
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 57.25,
      "epoch": 0.5979899497487438,
      "grad_norm": 48.09600296281366,
      "learning_rate": 2.0102918438345573e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.24251843988895416,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12376
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.25,
      "epoch": 0.5980382682643989,
      "grad_norm": 0.0,
      "learning_rate": 2.0100502512562815e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12377
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.40625,
      "epoch": 0.5980865867800541,
      "grad_norm": 0.0,
      "learning_rate": 2.0098086586780053e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12378
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.625,
      "epoch": 0.5981349052957093,
      "grad_norm": 0.0,
      "learning_rate": 2.0095670660997295e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12379
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.3125,
      "epoch": 0.5981832238113645,
      "grad_norm": 0.0,
      "learning_rate": 2.0093254735214533e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12380
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.875,
      "epoch": 0.5982315423270197,
      "grad_norm": 0.0,
      "learning_rate": 2.0090838809431772e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12381
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 64.09375,
      "epoch": 0.5982798608426749,
      "grad_norm": 36.199300202062496,
      "learning_rate": 2.0088422883649013e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 12382
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.0625,
      "epoch": 0.5983281793583302,
      "grad_norm": 0.0,
      "learning_rate": 2.0086006957866252e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12383
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 61.5625,
      "epoch": 0.5983764978739853,
      "grad_norm": 0.0,
      "learning_rate": 2.0083591032083493e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12384
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.78125,
      "epoch": 0.5984248163896405,
      "grad_norm": 0.0,
      "learning_rate": 2.0081175106300732e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12385
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 57.1875,
      "epoch": 0.5984731349052957,
      "grad_norm": 0.0,
      "learning_rate": 2.0078759180517974e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12386
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 63.09375,
      "epoch": 0.5985214534209509,
      "grad_norm": 0.0,
      "learning_rate": 2.0076343254735215e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12387
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.875,
      "epoch": 0.5985697719366061,
      "grad_norm": 0.0,
      "learning_rate": 2.0073927328952454e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 12388
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.84375,
      "epoch": 0.5986180904522613,
      "grad_norm": 0.0,
      "learning_rate": 2.0071511403169695e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12389
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 71.09375,
      "epoch": 0.5986664089679165,
      "grad_norm": 0.0,
      "learning_rate": 2.0069095477386934e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12390
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.75,
      "epoch": 0.5987147274835717,
      "grad_norm": 0.0,
      "learning_rate": 2.0066679551604172e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12391
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.6875,
      "epoch": 0.5987630459992269,
      "grad_norm": 0.0,
      "learning_rate": 2.0064263625821414e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12392
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.90625,
      "epoch": 0.5988113645148821,
      "grad_norm": 0.0,
      "learning_rate": 2.0061847700038652e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12393
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.8125,
      "epoch": 0.5988596830305373,
      "grad_norm": 0.0,
      "learning_rate": 2.0059431774255894e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12394
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.21875,
      "epoch": 0.5989080015461925,
      "grad_norm": 0.0,
      "learning_rate": 2.0057015848473133e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12395
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.5625,
      "epoch": 0.5989563200618477,
      "grad_norm": 0.0,
      "learning_rate": 2.0054599922690374e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12396
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.09375,
      "epoch": 0.5990046385775029,
      "grad_norm": 0.0,
      "learning_rate": 2.0052183996907615e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12397
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.71875,
      "epoch": 0.5990529570931581,
      "grad_norm": 0.0,
      "learning_rate": 2.0049768071124854e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12398
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.125,
      "epoch": 0.5991012756088133,
      "grad_norm": 0.0,
      "learning_rate": 2.0047352145342095e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12399
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 67.71875,
      "epoch": 0.5991495941244684,
      "grad_norm": 0.0,
      "learning_rate": 2.0044936219559334e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12400
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.15625,
      "epoch": 0.5991979126401237,
      "grad_norm": 0.0,
      "learning_rate": 2.0042520293776575e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12401
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.28125,
      "epoch": 0.5992462311557789,
      "grad_norm": 0.0,
      "learning_rate": 2.0040104367993814e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12402
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 62.375,
      "epoch": 0.5992945496714341,
      "grad_norm": 0.0,
      "learning_rate": 2.0037688442211053e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12403
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.40625,
      "epoch": 0.5993428681870893,
      "grad_norm": 0.0,
      "learning_rate": 2.0035272516428294e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12404
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.03125,
      "epoch": 0.5993911867027445,
      "grad_norm": 0.0,
      "learning_rate": 2.0032856590645533e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12405
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.375,
      "epoch": 0.5994395052183997,
      "grad_norm": 0.0,
      "learning_rate": 2.0030440664862774e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12406
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.625,
      "epoch": 0.5994878237340548,
      "grad_norm": 0.0,
      "learning_rate": 2.0028024739080016e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12407
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.40625,
      "epoch": 0.5995361422497101,
      "grad_norm": 0.0,
      "learning_rate": 2.0025608813297254e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12408
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.46875,
      "epoch": 0.5995844607653653,
      "grad_norm": 0.0,
      "learning_rate": 2.0023192887514496e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12409
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.5,
      "epoch": 0.5996327792810204,
      "grad_norm": 0.0,
      "learning_rate": 2.0020776961731734e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12410
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.375,
      "epoch": 0.5996810977966757,
      "grad_norm": 0.0,
      "learning_rate": 2.0018361035948976e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12411
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 55.8125,
      "epoch": 0.5997294163123309,
      "grad_norm": 125.7361728997569,
      "learning_rate": 2.0015945110166215e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.2961388826370239,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12412
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 69.125,
      "epoch": 0.5997777348279861,
      "grad_norm": 0.0,
      "learning_rate": 2.0013529184383453e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 12413
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.84375,
      "epoch": 0.5998260533436413,
      "grad_norm": 0.0,
      "learning_rate": 2.0011113258600695e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12414
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.875,
      "epoch": 0.5998743718592965,
      "grad_norm": 0.0,
      "learning_rate": 2.0008697332817933e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12415
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.375,
      "epoch": 0.5999226903749517,
      "grad_norm": 0.0,
      "learning_rate": 2.0006281407035175e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12416
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.59375,
      "epoch": 0.5999710088906068,
      "grad_norm": 0.0,
      "learning_rate": 2.0003865481252416e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12417
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.40625,
      "epoch": 0.6000193274062621,
      "grad_norm": 0.0,
      "learning_rate": 2.0001449555469655e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12418
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.21875,
      "epoch": 0.6000676459219173,
      "grad_norm": 0.0,
      "learning_rate": 1.9999033629686896e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12419
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.90625,
      "epoch": 0.6001159644375724,
      "grad_norm": 0.0,
      "learning_rate": 1.9996617703904135e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12420
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.90625,
      "epoch": 0.6001642829532277,
      "grad_norm": 0.0,
      "learning_rate": 1.9994201778121376e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12421
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.21875,
      "epoch": 0.6002126014688829,
      "grad_norm": 0.0,
      "learning_rate": 1.9991785852338615e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12422
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.75,
      "epoch": 0.6002609199845381,
      "grad_norm": 0.0,
      "learning_rate": 1.9989369926555854e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12423
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 58.5,
      "epoch": 0.6003092385001932,
      "grad_norm": 157.9207188124512,
      "learning_rate": 1.9986954000773095e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.3129216432571411,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12424
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.6875,
      "epoch": 0.6003575570158485,
      "grad_norm": 0.0,
      "learning_rate": 1.9984538074990334e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12425
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.34375,
      "epoch": 0.6004058755315037,
      "grad_norm": 0.0,
      "learning_rate": 1.9982122149207575e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12426
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.3125,
      "epoch": 0.6004541940471588,
      "grad_norm": 0.0,
      "learning_rate": 1.9979706223424816e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12427
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.3125,
      "epoch": 0.6005025125628141,
      "grad_norm": 0.0,
      "learning_rate": 1.9977290297642055e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12428
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.875,
      "epoch": 0.6005508310784693,
      "grad_norm": 0.0,
      "learning_rate": 1.9974874371859297e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12429
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.53125,
      "epoch": 0.6005991495941244,
      "grad_norm": 0.0,
      "learning_rate": 1.9972458446076535e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12430
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 62.34375,
      "epoch": 0.6006474681097796,
      "grad_norm": 0.0,
      "learning_rate": 1.9970042520293777e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 12431
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.75,
      "epoch": 0.6006957866254349,
      "grad_norm": 0.0,
      "learning_rate": 1.9967626594511018e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12432
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.65625,
      "epoch": 0.6007441051410901,
      "grad_norm": 0.0,
      "learning_rate": 1.9965210668728254e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12433
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.5625,
      "epoch": 0.6007924236567452,
      "grad_norm": 0.0,
      "learning_rate": 1.9962794742945495e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12434
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.09375,
      "epoch": 0.6008407421724005,
      "grad_norm": 0.0,
      "learning_rate": 1.9960378817162734e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12435
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.90625,
      "epoch": 0.6008890606880557,
      "grad_norm": 0.0,
      "learning_rate": 1.9957962891379976e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12436
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 59.96875,
      "epoch": 0.6009373792037108,
      "grad_norm": 0.0,
      "learning_rate": 1.9955546965597217e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 12437
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.59375,
      "epoch": 0.600985697719366,
      "grad_norm": 0.0,
      "learning_rate": 1.9953131039814456e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12438
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.78125,
      "epoch": 0.6010340162350213,
      "grad_norm": 0.0,
      "learning_rate": 1.9950715114031697e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12439
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.15625,
      "epoch": 0.6010823347506764,
      "grad_norm": 0.0,
      "learning_rate": 1.9948299188248936e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12440
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.1875,
      "epoch": 0.6011306532663316,
      "grad_norm": 0.0,
      "learning_rate": 1.9945883262466177e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12441
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.96875,
      "epoch": 0.6011789717819869,
      "grad_norm": 0.0,
      "learning_rate": 1.9943467336683418e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12442
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.71875,
      "epoch": 0.6012272902976421,
      "grad_norm": 0.0,
      "learning_rate": 1.9941051410900654e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12443
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 53.96875,
      "epoch": 0.6012756088132972,
      "grad_norm": 0.0,
      "learning_rate": 1.9938635485117896e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 12444
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.53125,
      "epoch": 0.6013239273289525,
      "grad_norm": 0.0,
      "learning_rate": 1.9936219559335135e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12445
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 62.90625,
      "epoch": 0.6013722458446077,
      "grad_norm": 6.18458597509796,
      "learning_rate": 1.9933803633552376e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.40631967782974243,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 0.96875,
      "step": 12446
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.96875,
      "epoch": 0.6014205643602628,
      "grad_norm": 0.0,
      "learning_rate": 1.9931387707769617e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12447
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.03125,
      "epoch": 0.601468882875918,
      "grad_norm": 0.0,
      "learning_rate": 1.9928971781986856e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12448
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 65.5625,
      "epoch": 0.6015172013915733,
      "grad_norm": 179.32406987372562,
      "learning_rate": 1.9926555856204097e-07,
      "loss": -0.0,
      "reward": 2.359375,
      "reward_std": 0.22738909721374512,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12449
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.65625,
      "epoch": 0.6015655199072284,
      "grad_norm": 0.0,
      "learning_rate": 1.9924139930421336e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12450
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.09375,
      "epoch": 0.6016138384228836,
      "grad_norm": 0.0,
      "learning_rate": 1.9921724004638577e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12451
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 65.5,
      "epoch": 0.6016621569385389,
      "grad_norm": 0.0,
      "learning_rate": 1.991930807885582e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 12452
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.46875,
      "epoch": 0.6017104754541941,
      "grad_norm": 0.0,
      "learning_rate": 1.9916892153073055e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12453
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 49.96875,
      "epoch": 0.6017587939698492,
      "grad_norm": 0.0,
      "learning_rate": 1.9914476227290296e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12454
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 41.0625,
      "epoch": 0.6018071124855044,
      "grad_norm": 0.0,
      "learning_rate": 1.9912060301507535e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12455
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.0,
      "epoch": 0.6018554310011597,
      "grad_norm": 0.0,
      "learning_rate": 1.9909644375724776e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12456
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.5625,
      "epoch": 0.6019037495168148,
      "grad_norm": 0.0,
      "learning_rate": 1.9907228449942018e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12457
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.09375,
      "epoch": 0.60195206803247,
      "grad_norm": 0.0,
      "learning_rate": 1.9904812524159256e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12458
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.0625,
      "epoch": 0.6020003865481253,
      "grad_norm": 0.0,
      "learning_rate": 1.9902396598376498e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12459
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 49.5,
      "epoch": 0.6020487050637804,
      "grad_norm": 0.0,
      "learning_rate": 1.9899980672593736e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12460
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.65625,
      "epoch": 0.6020970235794356,
      "grad_norm": 0.0,
      "learning_rate": 1.9897564746810978e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12461
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.21875,
      "epoch": 0.6021453420950909,
      "grad_norm": 0.0,
      "learning_rate": 1.989514882102822e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12462
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.4375,
      "epoch": 0.6021936606107461,
      "grad_norm": 0.0,
      "learning_rate": 1.9892732895245455e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12463
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.5625,
      "epoch": 0.6022419791264012,
      "grad_norm": 0.0,
      "learning_rate": 1.9890316969462697e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12464
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.15625,
      "epoch": 0.6022902976420564,
      "grad_norm": 0.0,
      "learning_rate": 1.9887901043679935e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12465
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.75,
      "epoch": 0.6023386161577117,
      "grad_norm": 0.0,
      "learning_rate": 1.9885485117897177e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12466
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.0625,
      "epoch": 0.6023869346733668,
      "grad_norm": 0.0,
      "learning_rate": 1.9883069192114418e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12467
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.375,
      "epoch": 0.602435253189022,
      "grad_norm": 0.0,
      "learning_rate": 1.9880653266331657e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12468
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.65625,
      "epoch": 0.6024835717046773,
      "grad_norm": 0.0,
      "learning_rate": 1.9878237340548898e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12469
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 65.0625,
      "epoch": 0.6025318902203324,
      "grad_norm": 0.0,
      "learning_rate": 1.9875821414766137e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 12470
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.65625,
      "epoch": 0.6025802087359876,
      "grad_norm": 0.0,
      "learning_rate": 1.9873405488983378e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12471
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.90625,
      "epoch": 0.6026285272516428,
      "grad_norm": 0.0,
      "learning_rate": 1.987098956320062e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12472
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.125,
      "epoch": 0.6026768457672981,
      "grad_norm": 0.0,
      "learning_rate": 1.9868573637417856e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12473
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.5,
      "epoch": 0.6027251642829532,
      "grad_norm": 0.0,
      "learning_rate": 1.9866157711635097e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12474
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 55.0,
      "epoch": 0.6027734827986084,
      "grad_norm": 26.2438755007487,
      "learning_rate": 1.9863741785852336e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.17598573863506317,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12475
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 57.0625,
      "epoch": 0.6028218013142637,
      "grad_norm": 0.0,
      "learning_rate": 1.9861325860069577e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 12476
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.4375,
      "epoch": 0.6028701198299188,
      "grad_norm": 0.0,
      "learning_rate": 1.9858909934286818e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12477
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.96875,
      "epoch": 0.602918438345574,
      "grad_norm": 0.0,
      "learning_rate": 1.9856494008504057e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12478
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.0625,
      "epoch": 0.6029667568612292,
      "grad_norm": 0.0,
      "learning_rate": 1.9854078082721299e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12479
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.5625,
      "epoch": 0.6030150753768844,
      "grad_norm": 0.0,
      "learning_rate": 1.9851662156938537e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12480
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.375,
      "epoch": 0.6030633938925396,
      "grad_norm": 0.0,
      "learning_rate": 1.9849246231155779e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12481
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.71875,
      "epoch": 0.6031117124081948,
      "grad_norm": 0.0,
      "learning_rate": 1.984683030537302e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12482
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.59375,
      "epoch": 0.6031600309238501,
      "grad_norm": 0.0,
      "learning_rate": 1.984441437959026e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12483
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.71875,
      "epoch": 0.6032083494395052,
      "grad_norm": 0.0,
      "learning_rate": 1.9841998453807497e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12484
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.25,
      "epoch": 0.6032566679551604,
      "grad_norm": 0.0,
      "learning_rate": 1.9839582528024736e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12485
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.8125,
      "epoch": 0.6033049864708157,
      "grad_norm": 0.0,
      "learning_rate": 1.9837166602241977e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12486
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.09375,
      "epoch": 0.6033533049864708,
      "grad_norm": 0.0,
      "learning_rate": 1.983475067645922e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12487
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.90625,
      "epoch": 0.603401623502126,
      "grad_norm": 0.0,
      "learning_rate": 1.9832334750676458e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12488
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 65.78125,
      "epoch": 0.6034499420177812,
      "grad_norm": 0.0,
      "learning_rate": 1.98299188248937e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 12489
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 65.90625,
      "epoch": 0.6034982605334364,
      "grad_norm": 0.0,
      "learning_rate": 1.9827502899110938e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12490
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.625,
      "epoch": 0.6035465790490916,
      "grad_norm": 0.0,
      "learning_rate": 1.982508697332818e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12491
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.34375,
      "epoch": 0.6035948975647468,
      "grad_norm": 0.0,
      "learning_rate": 1.982267104754542e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12492
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.03125,
      "epoch": 0.6036432160804021,
      "grad_norm": 0.0,
      "learning_rate": 1.982025512176266e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12493
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 64.5625,
      "epoch": 0.6036915345960572,
      "grad_norm": 0.0,
      "learning_rate": 1.9817839195979898e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12494
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 59.375,
      "epoch": 0.6037398531117124,
      "grad_norm": 0.0,
      "learning_rate": 1.9815423270197137e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 12495
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 61.9375,
      "epoch": 0.6037881716273676,
      "grad_norm": 0.0,
      "learning_rate": 1.9813007344414378e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 12496
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.125,
      "epoch": 0.6038364901430228,
      "grad_norm": 0.0,
      "learning_rate": 1.981059141863162e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12497
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.5,
      "epoch": 0.603884808658678,
      "grad_norm": 0.0,
      "learning_rate": 1.9808175492848858e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12498
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.5,
      "epoch": 0.6039331271743332,
      "grad_norm": 0.0,
      "learning_rate": 1.98057595670661e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12499
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.84375,
      "epoch": 0.6039814456899883,
      "grad_norm": 0.0,
      "learning_rate": 1.9803343641283338e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12500
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 59.0,
      "epoch": 0.6040297642056436,
      "grad_norm": 0.0,
      "learning_rate": 1.980092771550058e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12501
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.28125,
      "epoch": 0.6040780827212988,
      "grad_norm": 0.0,
      "learning_rate": 1.979851178971782e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12502
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.75,
      "epoch": 0.604126401236954,
      "grad_norm": 0.0,
      "learning_rate": 1.979609586393506e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12503
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.15625,
      "epoch": 0.6041747197526092,
      "grad_norm": 0.0,
      "learning_rate": 1.97936799381523e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12504
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.40625,
      "epoch": 0.6042230382682644,
      "grad_norm": 0.0,
      "learning_rate": 1.9791264012369537e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12505
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 60.71875,
      "epoch": 0.6042713567839196,
      "grad_norm": 0.0,
      "learning_rate": 1.9788848086586778e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12506
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.03125,
      "epoch": 0.6043196752995748,
      "grad_norm": 0.0,
      "learning_rate": 1.978643216080402e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12507
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 57.59375,
      "epoch": 0.60436799381523,
      "grad_norm": 0.0,
      "learning_rate": 1.9784016235021258e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 12508
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.625,
      "epoch": 0.6044163123308852,
      "grad_norm": 0.0,
      "learning_rate": 1.97816003092385e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12509
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 60.78125,
      "epoch": 0.6044646308465403,
      "grad_norm": 0.0,
      "learning_rate": 1.9779184383455738e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12510
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.59375,
      "epoch": 0.6045129493621956,
      "grad_norm": 0.0,
      "learning_rate": 1.977676845767298e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12511
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.15625,
      "epoch": 0.6045612678778508,
      "grad_norm": 0.0,
      "learning_rate": 1.977435253189022e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12512
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.59375,
      "epoch": 0.604609586393506,
      "grad_norm": 0.0,
      "learning_rate": 1.977193660610746e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12513
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.84375,
      "epoch": 0.6046579049091612,
      "grad_norm": 0.0,
      "learning_rate": 1.97695206803247e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12514
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.6875,
      "epoch": 0.6047062234248164,
      "grad_norm": 0.0,
      "learning_rate": 1.9767104754541937e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12515
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.28125,
      "epoch": 0.6047545419404716,
      "grad_norm": 0.0,
      "learning_rate": 1.9764688828759179e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 12516
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.375,
      "epoch": 0.6048028604561267,
      "grad_norm": 0.0,
      "learning_rate": 1.976227290297642e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12517
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.0625,
      "epoch": 0.604851178971782,
      "grad_norm": 0.0,
      "learning_rate": 1.975985697719366e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12518
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.53125,
      "epoch": 0.6048994974874372,
      "grad_norm": 0.0,
      "learning_rate": 1.97574410514109e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12519
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.34375,
      "epoch": 0.6049478160030923,
      "grad_norm": 0.0,
      "learning_rate": 1.975502512562814e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12520
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.71875,
      "epoch": 0.6049961345187476,
      "grad_norm": 0.0,
      "learning_rate": 1.975260919984538e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12521
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.09375,
      "epoch": 0.6050444530344028,
      "grad_norm": 0.0,
      "learning_rate": 1.9750193274062622e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12522
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.5625,
      "epoch": 0.605092771550058,
      "grad_norm": 0.0,
      "learning_rate": 1.974777734827986e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12523
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.09375,
      "epoch": 0.6051410900657131,
      "grad_norm": 0.0,
      "learning_rate": 1.9745361422497102e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12524
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.1875,
      "epoch": 0.6051894085813684,
      "grad_norm": 0.0,
      "learning_rate": 1.9742945496714338e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12525
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.5,
      "epoch": 0.6052377270970236,
      "grad_norm": 0.0,
      "learning_rate": 1.974052957093158e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12526
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.40625,
      "epoch": 0.6052860456126787,
      "grad_norm": 0.0,
      "learning_rate": 1.973811364514882e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12527
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.40625,
      "epoch": 0.605334364128334,
      "grad_norm": 0.0,
      "learning_rate": 1.973569771936606e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12528
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.3125,
      "epoch": 0.6053826826439892,
      "grad_norm": 0.0,
      "learning_rate": 1.97332817935833e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12529
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.03125,
      "epoch": 0.6054310011596443,
      "grad_norm": 0.0,
      "learning_rate": 1.973086586780054e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12530
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.53125,
      "epoch": 0.6054793196752996,
      "grad_norm": 0.0,
      "learning_rate": 1.972844994201778e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12531
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.65625,
      "epoch": 0.6055276381909548,
      "grad_norm": 0.0,
      "learning_rate": 1.9726034016235022e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12532
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 61.75,
      "epoch": 0.60557595670661,
      "grad_norm": 0.0,
      "learning_rate": 1.972361809045226e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 12533
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.09375,
      "epoch": 0.6056242752222651,
      "grad_norm": 0.0,
      "learning_rate": 1.9721202164669502e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12534
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.0625,
      "epoch": 0.6056725937379204,
      "grad_norm": 0.0,
      "learning_rate": 1.9718786238886738e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12535
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.78125,
      "epoch": 0.6057209122535756,
      "grad_norm": 0.0,
      "learning_rate": 1.971637031310398e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12536
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.46875,
      "epoch": 0.6057692307692307,
      "grad_norm": 0.0,
      "learning_rate": 1.971395438732122e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12537
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.75,
      "epoch": 0.605817549284886,
      "grad_norm": 0.0,
      "learning_rate": 1.971153846153846e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12538
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.0625,
      "epoch": 0.6058658678005412,
      "grad_norm": 0.0,
      "learning_rate": 1.97091225357557e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12539
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 68.375,
      "epoch": 0.6059141863161963,
      "grad_norm": 0.0,
      "learning_rate": 1.970670660997294e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12540
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.0,
      "epoch": 0.6059625048318515,
      "grad_norm": 0.0,
      "learning_rate": 1.970429068419018e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12541
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 45.375,
      "epoch": 0.6060108233475068,
      "grad_norm": 0.0,
      "learning_rate": 1.9701874758407422e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 12542
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.34375,
      "epoch": 0.606059141863162,
      "grad_norm": 0.0,
      "learning_rate": 1.969945883262466e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12543
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.71875,
      "epoch": 0.6061074603788171,
      "grad_norm": 0.0,
      "learning_rate": 1.9697042906841902e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12544
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.15625,
      "epoch": 0.6061557788944724,
      "grad_norm": 0.0,
      "learning_rate": 1.9694626981059138e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12545
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.5625,
      "epoch": 0.6062040974101276,
      "grad_norm": 0.0,
      "learning_rate": 1.969221105527638e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12546
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.84375,
      "epoch": 0.6062524159257827,
      "grad_norm": 0.0,
      "learning_rate": 1.968979512949362e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12547
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.5625,
      "epoch": 0.606300734441438,
      "grad_norm": 0.0,
      "learning_rate": 1.968737920371086e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12548
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.46875,
      "epoch": 0.6063490529570932,
      "grad_norm": 0.0,
      "learning_rate": 1.96849632779281e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12549
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.5625,
      "epoch": 0.6063973714727483,
      "grad_norm": 0.0,
      "learning_rate": 1.968254735214534e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12550
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 63.15625,
      "epoch": 0.6064456899884035,
      "grad_norm": 0.0,
      "learning_rate": 1.9680131426362581e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12551
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.15625,
      "epoch": 0.6064940085040588,
      "grad_norm": 0.0,
      "learning_rate": 1.9677715500579823e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12552
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.15625,
      "epoch": 0.606542327019714,
      "grad_norm": 0.0,
      "learning_rate": 1.9675299574797061e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12553
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.15625,
      "epoch": 0.6065906455353691,
      "grad_norm": 0.0,
      "learning_rate": 1.9672883649014303e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12554
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.28125,
      "epoch": 0.6066389640510244,
      "grad_norm": 0.0,
      "learning_rate": 1.9670467723231542e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12555
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.4375,
      "epoch": 0.6066872825666796,
      "grad_norm": 0.0,
      "learning_rate": 1.966805179744878e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12556
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.5625,
      "epoch": 0.6067356010823347,
      "grad_norm": 0.0,
      "learning_rate": 1.9665635871666022e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12557
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.375,
      "epoch": 0.6067839195979899,
      "grad_norm": 0.0,
      "learning_rate": 1.966321994588326e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12558
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.03125,
      "epoch": 0.6068322381136452,
      "grad_norm": 0.0,
      "learning_rate": 1.9660804020100502e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12559
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.3125,
      "epoch": 0.6068805566293003,
      "grad_norm": 0.0,
      "learning_rate": 1.965838809431774e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12560
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.59375,
      "epoch": 0.6069288751449555,
      "grad_norm": 0.0,
      "learning_rate": 1.9655972168534982e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12561
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.34375,
      "epoch": 0.6069771936606108,
      "grad_norm": 0.0,
      "learning_rate": 1.9653556242752223e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12562
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.09375,
      "epoch": 0.607025512176266,
      "grad_norm": 0.0,
      "learning_rate": 1.9651140316969462e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12563
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.375,
      "epoch": 0.6070738306919211,
      "grad_norm": 0.0,
      "learning_rate": 1.9648724391186703e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12564
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.0625,
      "epoch": 0.6071221492075763,
      "grad_norm": 0.0,
      "learning_rate": 1.9646308465403942e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12565
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.53125,
      "epoch": 0.6071704677232316,
      "grad_norm": 0.0,
      "learning_rate": 1.964389253962118e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 12566
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.5,
      "epoch": 0.6072187862388867,
      "grad_norm": 0.0,
      "learning_rate": 1.9641476613838422e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12567
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.0,
      "epoch": 0.6072671047545419,
      "grad_norm": 0.0,
      "learning_rate": 1.963906068805566e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12568
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.5625,
      "epoch": 0.6073154232701972,
      "grad_norm": 0.0,
      "learning_rate": 1.9636644762272902e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12569
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.375,
      "epoch": 0.6073637417858523,
      "grad_norm": 0.0,
      "learning_rate": 1.963422883649014e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12570
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.46875,
      "epoch": 0.6074120603015075,
      "grad_norm": 0.0,
      "learning_rate": 1.9631812910707382e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12571
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.78125,
      "epoch": 0.6074603788171627,
      "grad_norm": 0.0,
      "learning_rate": 1.9629396984924624e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12572
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.5625,
      "epoch": 0.607508697332818,
      "grad_norm": 0.0,
      "learning_rate": 1.9626981059141862e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12573
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.25,
      "epoch": 0.6075570158484731,
      "grad_norm": 0.0,
      "learning_rate": 1.9624565133359104e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12574
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.15625,
      "epoch": 0.6076053343641283,
      "grad_norm": 0.0,
      "learning_rate": 1.9622149207576342e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12575
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.5625,
      "epoch": 0.6076536528797836,
      "grad_norm": 0.0,
      "learning_rate": 1.961973328179358e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12576
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.0625,
      "epoch": 0.6077019713954387,
      "grad_norm": 0.0,
      "learning_rate": 1.9617317356010822e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12577
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.46875,
      "epoch": 0.6077502899110939,
      "grad_norm": 0.0,
      "learning_rate": 1.961490143022806e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12578
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.6875,
      "epoch": 0.6077986084267492,
      "grad_norm": 0.0,
      "learning_rate": 1.9612485504445302e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12579
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.4375,
      "epoch": 0.6078469269424043,
      "grad_norm": 0.0,
      "learning_rate": 1.961006957866254e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12580
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.8125,
      "epoch": 0.6078952454580595,
      "grad_norm": 0.0,
      "learning_rate": 1.9607653652879783e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12581
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.1875,
      "epoch": 0.6079435639737147,
      "grad_norm": 0.0,
      "learning_rate": 1.9605237727097024e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12582
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.65625,
      "epoch": 0.60799188248937,
      "grad_norm": 0.0,
      "learning_rate": 1.9602821801314263e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.260896235704422,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12583
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.03125,
      "epoch": 0.6080402010050251,
      "grad_norm": 0.0,
      "learning_rate": 1.9600405875531504e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12584
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.78125,
      "epoch": 0.6080885195206803,
      "grad_norm": 0.0,
      "learning_rate": 1.9597989949748743e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12585
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.0625,
      "epoch": 0.6081368380363356,
      "grad_norm": 0.0,
      "learning_rate": 1.9595574023965984e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12586
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.59375,
      "epoch": 0.6081851565519907,
      "grad_norm": 0.0,
      "learning_rate": 1.9593158098183223e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12587
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.875,
      "epoch": 0.6082334750676459,
      "grad_norm": 0.0,
      "learning_rate": 1.9590742172400461e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12588
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.46875,
      "epoch": 0.6082817935833011,
      "grad_norm": 0.0,
      "learning_rate": 1.9588326246617703e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12589
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.375,
      "epoch": 0.6083301120989563,
      "grad_norm": 0.0,
      "learning_rate": 1.9585910320834942e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12590
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.875,
      "epoch": 0.6083784306146115,
      "grad_norm": 0.0,
      "learning_rate": 1.9583494395052183e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12591
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.625,
      "epoch": 0.6084267491302667,
      "grad_norm": 0.0,
      "learning_rate": 1.9581078469269424e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12592
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.125,
      "epoch": 0.608475067645922,
      "grad_norm": 0.0,
      "learning_rate": 1.9578662543486663e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12593
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.28125,
      "epoch": 0.6085233861615771,
      "grad_norm": 0.0,
      "learning_rate": 1.9576246617703904e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12594
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 49.75,
      "epoch": 0.6085717046772323,
      "grad_norm": 0.0,
      "learning_rate": 1.9573830691921143e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12595
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 56.28125,
      "epoch": 0.6086200231928875,
      "grad_norm": 0.0,
      "learning_rate": 1.9571414766138384e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 12596
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.75,
      "epoch": 0.6086683417085427,
      "grad_norm": 0.0,
      "learning_rate": 1.9568998840355623e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12597
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.25,
      "epoch": 0.6087166602241979,
      "grad_norm": 0.0,
      "learning_rate": 1.9566582914572862e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12598
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.65625,
      "epoch": 0.6087649787398531,
      "grad_norm": 0.0,
      "learning_rate": 1.9564166988790103e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12599
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.625,
      "epoch": 0.6088132972555083,
      "grad_norm": 0.0,
      "learning_rate": 1.9561751063007342e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12600
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.125,
      "epoch": 0.6088616157711635,
      "grad_norm": 0.0,
      "learning_rate": 1.9559335137224583e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 12601
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.125,
      "epoch": 0.6089099342868187,
      "grad_norm": 0.0,
      "learning_rate": 1.9556919211441825e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12602
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.5,
      "epoch": 0.608958252802474,
      "grad_norm": 0.0,
      "learning_rate": 1.9554503285659063e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12603
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.5,
      "epoch": 0.6090065713181291,
      "grad_norm": 0.0,
      "learning_rate": 1.9552087359876305e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12604
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 63.1875,
      "epoch": 0.6090548898337843,
      "grad_norm": 0.0,
      "learning_rate": 1.9549671434093543e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 12605
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.65625,
      "epoch": 0.6091032083494395,
      "grad_norm": 0.0,
      "learning_rate": 1.9547255508310785e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12606
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.84375,
      "epoch": 0.6091515268650947,
      "grad_norm": 0.0,
      "learning_rate": 1.9544839582528026e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12607
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 42.625,
      "epoch": 0.6091998453807499,
      "grad_norm": 0.0,
      "learning_rate": 1.9542423656745262e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12608
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.40625,
      "epoch": 0.6092481638964051,
      "grad_norm": 0.0,
      "learning_rate": 1.9540007730962504e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12609
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.21875,
      "epoch": 0.6092964824120602,
      "grad_norm": 0.0,
      "learning_rate": 1.9537591805179742e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12610
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.96875,
      "epoch": 0.6093448009277155,
      "grad_norm": 0.0,
      "learning_rate": 1.9535175879396984e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12611
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 61.71875,
      "epoch": 0.6093931194433707,
      "grad_norm": 362.3074482789151,
      "learning_rate": 1.9532759953614225e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.15580293536186218,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12612
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.21875,
      "epoch": 0.609441437959026,
      "grad_norm": 0.0,
      "learning_rate": 1.9530344027831464e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12613
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.96875,
      "epoch": 0.6094897564746811,
      "grad_norm": 0.0,
      "learning_rate": 1.9527928102048705e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12614
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.21875,
      "epoch": 0.6095380749903363,
      "grad_norm": 0.0,
      "learning_rate": 1.9525512176265944e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12615
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.1875,
      "epoch": 0.6095863935059915,
      "grad_norm": 0.0,
      "learning_rate": 1.9523096250483185e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12616
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.8125,
      "epoch": 0.6096347120216467,
      "grad_norm": 0.0,
      "learning_rate": 1.9520680324700427e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12617
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.9375,
      "epoch": 0.6096830305373019,
      "grad_norm": 0.0,
      "learning_rate": 1.9518264398917663e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12618
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.46875,
      "epoch": 0.6097313490529571,
      "grad_norm": 0.0,
      "learning_rate": 1.9515848473134904e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12619
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.9375,
      "epoch": 0.6097796675686122,
      "grad_norm": 0.0,
      "learning_rate": 1.9513432547352143e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12620
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.46875,
      "epoch": 0.6098279860842675,
      "grad_norm": 0.0,
      "learning_rate": 1.9511016621569384e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12621
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.46875,
      "epoch": 0.6098763045999227,
      "grad_norm": 0.0,
      "learning_rate": 1.9508600695786625e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12622
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.65625,
      "epoch": 0.6099246231155779,
      "grad_norm": 0.0,
      "learning_rate": 1.9506184770003864e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12623
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.03125,
      "epoch": 0.6099729416312331,
      "grad_norm": 0.0,
      "learning_rate": 1.9503768844221106e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12624
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.75,
      "epoch": 0.6100212601468883,
      "grad_norm": 0.0,
      "learning_rate": 1.9501352918438344e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12625
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.3125,
      "epoch": 0.6100695786625435,
      "grad_norm": 0.0,
      "learning_rate": 1.9498936992655586e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12626
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.375,
      "epoch": 0.6101178971781986,
      "grad_norm": 0.0,
      "learning_rate": 1.9496521066872827e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12627
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.21875,
      "epoch": 0.6101662156938539,
      "grad_norm": 0.0,
      "learning_rate": 1.9494105141090063e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12628
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.25,
      "epoch": 0.6102145342095091,
      "grad_norm": 0.0,
      "learning_rate": 1.9491689215307304e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12629
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.4375,
      "epoch": 0.6102628527251642,
      "grad_norm": 0.0,
      "learning_rate": 1.9489273289524543e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12630
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.375,
      "epoch": 0.6103111712408195,
      "grad_norm": 0.0,
      "learning_rate": 1.9486857363741785e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12631
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.625,
      "epoch": 0.6103594897564747,
      "grad_norm": 0.0,
      "learning_rate": 1.9484441437959026e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12632
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 69.90625,
      "epoch": 0.6104078082721299,
      "grad_norm": 0.0,
      "learning_rate": 1.9482025512176265e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 12633
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.09375,
      "epoch": 0.610456126787785,
      "grad_norm": 0.0,
      "learning_rate": 1.9479609586393506e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12634
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.65625,
      "epoch": 0.6105044453034403,
      "grad_norm": 0.0,
      "learning_rate": 1.9477193660610745e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12635
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.96875,
      "epoch": 0.6105527638190955,
      "grad_norm": 0.0,
      "learning_rate": 1.9474777734827986e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12636
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.34375,
      "epoch": 0.6106010823347506,
      "grad_norm": 0.0,
      "learning_rate": 1.9472361809045227e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12637
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.875,
      "epoch": 0.6106494008504059,
      "grad_norm": 0.0,
      "learning_rate": 1.9469945883262463e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12638
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.5,
      "epoch": 0.6106977193660611,
      "grad_norm": 0.0,
      "learning_rate": 1.9467529957479705e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 12639
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.625,
      "epoch": 0.6107460378817162,
      "grad_norm": 0.0,
      "learning_rate": 1.9465114031696944e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12640
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.90625,
      "epoch": 0.6107943563973715,
      "grad_norm": 0.0,
      "learning_rate": 1.9462698105914185e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12641
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.46875,
      "epoch": 0.6108426749130267,
      "grad_norm": 0.0,
      "learning_rate": 1.9460282180131426e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12642
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.6875,
      "epoch": 0.6108909934286819,
      "grad_norm": 0.0,
      "learning_rate": 1.9457866254348665e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12643
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.84375,
      "epoch": 0.610939311944337,
      "grad_norm": 0.0,
      "learning_rate": 1.9455450328565906e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12644
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.84375,
      "epoch": 0.6109876304599923,
      "grad_norm": 0.0,
      "learning_rate": 1.9453034402783145e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12645
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.40625,
      "epoch": 0.6110359489756475,
      "grad_norm": 0.0,
      "learning_rate": 1.9450618477000386e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 12646
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.28125,
      "epoch": 0.6110842674913026,
      "grad_norm": 0.0,
      "learning_rate": 1.9448202551217628e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12647
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.65625,
      "epoch": 0.6111325860069579,
      "grad_norm": 0.0,
      "learning_rate": 1.9445786625434864e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12648
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.09375,
      "epoch": 0.6111809045226131,
      "grad_norm": 0.0,
      "learning_rate": 1.9443370699652105e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12649
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.46875,
      "epoch": 0.6112292230382682,
      "grad_norm": 0.0,
      "learning_rate": 1.9440954773869344e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12650
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.03125,
      "epoch": 0.6112775415539234,
      "grad_norm": 0.0,
      "learning_rate": 1.9438538848086585e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12651
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 58.25,
      "epoch": 0.6113258600695787,
      "grad_norm": 66.35947582444354,
      "learning_rate": 1.9436122922303827e-07,
      "loss": -0.0,
      "reward": 2.390625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12652
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.40625,
      "epoch": 0.6113741785852339,
      "grad_norm": 0.0,
      "learning_rate": 1.9433706996521065e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12653
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.4375,
      "epoch": 0.611422497100889,
      "grad_norm": 0.0,
      "learning_rate": 1.9431291070738307e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12654
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.25,
      "epoch": 0.6114708156165443,
      "grad_norm": 0.0,
      "learning_rate": 1.9428875144955545e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12655
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.8125,
      "epoch": 0.6115191341321995,
      "grad_norm": 0.0,
      "learning_rate": 1.9426459219172787e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12656
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.9375,
      "epoch": 0.6115674526478546,
      "grad_norm": 0.0,
      "learning_rate": 1.9424043293390028e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12657
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.875,
      "epoch": 0.6116157711635098,
      "grad_norm": 0.0,
      "learning_rate": 1.9421627367607267e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12658
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.875,
      "epoch": 0.6116640896791651,
      "grad_norm": 0.0,
      "learning_rate": 1.9419211441824506e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12659
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.65625,
      "epoch": 0.6117124081948202,
      "grad_norm": 0.0,
      "learning_rate": 1.9416795516041744e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12660
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.6875,
      "epoch": 0.6117607267104754,
      "grad_norm": 0.0,
      "learning_rate": 1.9414379590258986e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12661
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.65625,
      "epoch": 0.6118090452261307,
      "grad_norm": 0.0,
      "learning_rate": 1.9411963664476227e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12662
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.875,
      "epoch": 0.6118573637417859,
      "grad_norm": 0.0,
      "learning_rate": 1.9409547738693466e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12663
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.84375,
      "epoch": 0.611905682257441,
      "grad_norm": 0.0,
      "learning_rate": 1.9407131812910707e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12664
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.78125,
      "epoch": 0.6119540007730963,
      "grad_norm": 0.0,
      "learning_rate": 1.9404715887127946e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12665
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.25,
      "epoch": 0.6120023192887515,
      "grad_norm": 0.0,
      "learning_rate": 1.9402299961345187e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12666
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.40625,
      "epoch": 0.6120506378044066,
      "grad_norm": 0.0,
      "learning_rate": 1.9399884035562429e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12667
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.96875,
      "epoch": 0.6120989563200618,
      "grad_norm": 0.0,
      "learning_rate": 1.9397468109779667e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12668
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.21875,
      "epoch": 0.6121472748357171,
      "grad_norm": 0.0,
      "learning_rate": 1.9395052183996906e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12669
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.34375,
      "epoch": 0.6121955933513722,
      "grad_norm": 0.0,
      "learning_rate": 1.9392636258214145e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12670
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.3125,
      "epoch": 0.6122439118670274,
      "grad_norm": 0.0,
      "learning_rate": 1.9390220332431386e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12671
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.59375,
      "epoch": 0.6122922303826827,
      "grad_norm": 0.0,
      "learning_rate": 1.9387804406648627e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12672
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.65625,
      "epoch": 0.6123405488983379,
      "grad_norm": 0.0,
      "learning_rate": 1.9385388480865866e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12673
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.8125,
      "epoch": 0.612388867413993,
      "grad_norm": 0.0,
      "learning_rate": 1.9382972555083108e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12674
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.375,
      "epoch": 0.6124371859296482,
      "grad_norm": 0.0,
      "learning_rate": 1.9380556629300346e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12675
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.21875,
      "epoch": 0.6124855044453035,
      "grad_norm": 0.0,
      "learning_rate": 1.9378140703517588e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12676
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.9375,
      "epoch": 0.6125338229609586,
      "grad_norm": 0.0,
      "learning_rate": 1.937572477773483e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12677
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.46875,
      "epoch": 0.6125821414766138,
      "grad_norm": 0.0,
      "learning_rate": 1.9373308851952068e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12678
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.8125,
      "epoch": 0.6126304599922691,
      "grad_norm": 0.0,
      "learning_rate": 1.9370892926169306e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12679
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 65.40625,
      "epoch": 0.6126787785079242,
      "grad_norm": 77.11092199208166,
      "learning_rate": 1.9368477000386545e-07,
      "loss": 0.0,
      "reward": 1.546875,
      "reward_std": 0.188242569565773,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12680
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.84375,
      "epoch": 0.6127270970235794,
      "grad_norm": 0.0,
      "learning_rate": 1.9366061074603786e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12681
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.3125,
      "epoch": 0.6127754155392346,
      "grad_norm": 0.0,
      "learning_rate": 1.9363645148821028e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12682
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.25,
      "epoch": 0.6128237340548899,
      "grad_norm": 0.0,
      "learning_rate": 1.9361229223038267e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12683
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.8125,
      "epoch": 0.612872052570545,
      "grad_norm": 0.0,
      "learning_rate": 1.9358813297255508e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12684
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.15625,
      "epoch": 0.6129203710862002,
      "grad_norm": 0.0,
      "learning_rate": 1.9356397371472747e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12685
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.65625,
      "epoch": 0.6129686896018555,
      "grad_norm": 0.0,
      "learning_rate": 1.9353981445689988e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12686
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.03125,
      "epoch": 0.6130170081175106,
      "grad_norm": 0.0,
      "learning_rate": 1.935156551990723e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12687
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.5625,
      "epoch": 0.6130653266331658,
      "grad_norm": 0.0,
      "learning_rate": 1.9349149594124468e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12688
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.59375,
      "epoch": 0.613113645148821,
      "grad_norm": 0.0,
      "learning_rate": 1.934673366834171e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12689
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.25,
      "epoch": 0.6131619636644762,
      "grad_norm": 0.0,
      "learning_rate": 1.9344317742558946e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12690
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 66.8125,
      "epoch": 0.6132102821801314,
      "grad_norm": 17.559241143826714,
      "learning_rate": 1.9341901816776187e-07,
      "loss": -0.0,
      "reward": 1.640625,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12691
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.4375,
      "epoch": 0.6132586006957866,
      "grad_norm": 0.0,
      "learning_rate": 1.9339485890993428e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12692
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.1875,
      "epoch": 0.6133069192114419,
      "grad_norm": 0.0,
      "learning_rate": 1.9337069965210667e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12693
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.5625,
      "epoch": 0.613355237727097,
      "grad_norm": 0.0,
      "learning_rate": 1.9334654039427908e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12694
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 54.90625,
      "epoch": 0.6134035562427522,
      "grad_norm": 12.86075772040934,
      "learning_rate": 1.9332238113645147e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 12695
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.21875,
      "epoch": 0.6134518747584075,
      "grad_norm": 0.0,
      "learning_rate": 1.9329822187862388e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12696
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 67.375,
      "epoch": 0.6135001932740626,
      "grad_norm": 0.0,
      "learning_rate": 1.932740626207963e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 12697
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.46875,
      "epoch": 0.6135485117897178,
      "grad_norm": 0.0,
      "learning_rate": 1.9324990336296868e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 12698
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.25,
      "epoch": 0.613596830305373,
      "grad_norm": 0.0,
      "learning_rate": 1.932257441051411e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12699
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.40625,
      "epoch": 0.6136451488210282,
      "grad_norm": 0.0,
      "learning_rate": 1.9320158484731346e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12700
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 61.875,
      "epoch": 0.6136934673366834,
      "grad_norm": 0.0,
      "learning_rate": 1.9317742558948587e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 12701
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.46875,
      "epoch": 0.6137417858523386,
      "grad_norm": 0.0,
      "learning_rate": 1.9315326633165829e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12702
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.40625,
      "epoch": 0.6137901043679939,
      "grad_norm": 0.0,
      "learning_rate": 1.9312910707383067e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12703
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.84375,
      "epoch": 0.613838422883649,
      "grad_norm": 0.0,
      "learning_rate": 1.931049478160031e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12704
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.4375,
      "epoch": 0.6138867413993042,
      "grad_norm": 0.0,
      "learning_rate": 1.9308078855817547e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12705
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 54.28125,
      "epoch": 0.6139350599149594,
      "grad_norm": 40.31638816405471,
      "learning_rate": 1.930566293003479e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.3006648123264313,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12706
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.5,
      "epoch": 0.6139833784306146,
      "grad_norm": 0.0,
      "learning_rate": 1.930324700425203e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12707
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.46875,
      "epoch": 0.6140316969462698,
      "grad_norm": 0.0,
      "learning_rate": 1.930083107846927e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12708
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.40625,
      "epoch": 0.614080015461925,
      "grad_norm": 0.0,
      "learning_rate": 1.929841515268651e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12709
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.125,
      "epoch": 0.6141283339775802,
      "grad_norm": 0.0,
      "learning_rate": 1.9295999226903746e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12710
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.6875,
      "epoch": 0.6141766524932354,
      "grad_norm": 0.0,
      "learning_rate": 1.9293583301120988e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12711
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.6875,
      "epoch": 0.6142249710088906,
      "grad_norm": 0.0,
      "learning_rate": 1.929116737533823e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12712
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 63.375,
      "epoch": 0.6142732895245459,
      "grad_norm": 0.0,
      "learning_rate": 1.9288751449555468e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 12713
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.84375,
      "epoch": 0.614321608040201,
      "grad_norm": 0.0,
      "learning_rate": 1.928633552377271e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12714
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.4375,
      "epoch": 0.6143699265558562,
      "grad_norm": 0.0,
      "learning_rate": 1.9283919597989948e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12715
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.15625,
      "epoch": 0.6144182450715114,
      "grad_norm": 0.0,
      "learning_rate": 1.928150367220719e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12716
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.46875,
      "epoch": 0.6144665635871666,
      "grad_norm": 0.0,
      "learning_rate": 1.927908774642443e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12717
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.4375,
      "epoch": 0.6145148821028218,
      "grad_norm": 0.0,
      "learning_rate": 1.927667182064167e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12718
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.65625,
      "epoch": 0.614563200618477,
      "grad_norm": 0.0,
      "learning_rate": 1.927425589485891e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12719
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.125,
      "epoch": 0.6146115191341321,
      "grad_norm": 0.0,
      "learning_rate": 1.9271839969076147e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12720
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.46875,
      "epoch": 0.6146598376497874,
      "grad_norm": 0.0,
      "learning_rate": 1.9269424043293388e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12721
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.4375,
      "epoch": 0.6147081561654426,
      "grad_norm": 0.0,
      "learning_rate": 1.926700811751063e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12722
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.8125,
      "epoch": 0.6147564746810978,
      "grad_norm": 0.0,
      "learning_rate": 1.9264592191727868e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12723
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.25,
      "epoch": 0.614804793196753,
      "grad_norm": 0.0,
      "learning_rate": 1.926217626594511e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12724
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.40625,
      "epoch": 0.6148531117124082,
      "grad_norm": 0.0,
      "learning_rate": 1.9259760340162348e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12725
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.3125,
      "epoch": 0.6149014302280634,
      "grad_norm": 0.0,
      "learning_rate": 1.925734441437959e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12726
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.65625,
      "epoch": 0.6149497487437185,
      "grad_norm": 0.0,
      "learning_rate": 1.925492848859683e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12727
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.5,
      "epoch": 0.6149980672593738,
      "grad_norm": 0.0,
      "learning_rate": 1.925251256281407e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.260896235704422,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12728
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.0625,
      "epoch": 0.615046385775029,
      "grad_norm": 0.0,
      "learning_rate": 1.925009663703131e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12729
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.46875,
      "epoch": 0.6150947042906841,
      "grad_norm": 0.0,
      "learning_rate": 1.924768071124855e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12730
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.46875,
      "epoch": 0.6151430228063394,
      "grad_norm": 0.0,
      "learning_rate": 1.9245264785465788e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12731
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.53125,
      "epoch": 0.6151913413219946,
      "grad_norm": 0.0,
      "learning_rate": 1.924284885968303e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12732
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.625,
      "epoch": 0.6152396598376498,
      "grad_norm": 0.0,
      "learning_rate": 1.9240432933900269e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12733
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 54.75,
      "epoch": 0.615287978353305,
      "grad_norm": 0.0,
      "learning_rate": 1.923801700811751e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 12734
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 61.25,
      "epoch": 0.6153362968689602,
      "grad_norm": 0.0,
      "learning_rate": 1.9235601082334749e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12735
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.21875,
      "epoch": 0.6153846153846154,
      "grad_norm": 0.0,
      "learning_rate": 1.923318515655199e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12736
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.1875,
      "epoch": 0.6154329339002705,
      "grad_norm": 0.0,
      "learning_rate": 1.9230769230769231e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12737
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.65625,
      "epoch": 0.6154812524159258,
      "grad_norm": 0.0,
      "learning_rate": 1.922835330498647e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12738
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.9375,
      "epoch": 0.615529570931581,
      "grad_norm": 0.0,
      "learning_rate": 1.9225937379203711e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12739
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.125,
      "epoch": 0.6155778894472361,
      "grad_norm": 0.0,
      "learning_rate": 1.922352145342095e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12740
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.25,
      "epoch": 0.6156262079628914,
      "grad_norm": 0.0,
      "learning_rate": 1.922110552763819e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12741
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.25,
      "epoch": 0.6156745264785466,
      "grad_norm": 0.0,
      "learning_rate": 1.921868960185543e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12742
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.875,
      "epoch": 0.6157228449942018,
      "grad_norm": 0.0,
      "learning_rate": 1.921627367607267e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12743
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 61.90625,
      "epoch": 0.615771163509857,
      "grad_norm": 0.0,
      "learning_rate": 1.921385775028991e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 12744
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 52.8125,
      "epoch": 0.6158194820255122,
      "grad_norm": 16.060016871315838,
      "learning_rate": 1.921144182450715e-07,
      "loss": -0.0,
      "reward": 2.3125,
      "reward_std": 0.23047873377799988,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12745
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.5625,
      "epoch": 0.6158678005411674,
      "grad_norm": 0.0,
      "learning_rate": 1.920902589872439e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12746
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.65625,
      "epoch": 0.6159161190568225,
      "grad_norm": 0.0,
      "learning_rate": 1.9206609972941632e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12747
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.28125,
      "epoch": 0.6159644375724778,
      "grad_norm": 0.0,
      "learning_rate": 1.920419404715887e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12748
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.96875,
      "epoch": 0.616012756088133,
      "grad_norm": 0.0,
      "learning_rate": 1.9201778121376112e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12749
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.03125,
      "epoch": 0.6160610746037881,
      "grad_norm": 0.0,
      "learning_rate": 1.919936219559335e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12750
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.46875,
      "epoch": 0.6161093931194433,
      "grad_norm": 0.0,
      "learning_rate": 1.919694626981059e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12751
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.15625,
      "epoch": 0.6161577116350986,
      "grad_norm": 0.0,
      "learning_rate": 1.919453034402783e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12752
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.9375,
      "epoch": 0.6162060301507538,
      "grad_norm": 0.0,
      "learning_rate": 1.919211441824507e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12753
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.875,
      "epoch": 0.6162543486664089,
      "grad_norm": 0.0,
      "learning_rate": 1.918969849246231e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12754
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.6875,
      "epoch": 0.6163026671820642,
      "grad_norm": 0.0,
      "learning_rate": 1.918728256667955e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12755
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.28125,
      "epoch": 0.6163509856977194,
      "grad_norm": 0.0,
      "learning_rate": 1.918486664089679e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12756
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.4375,
      "epoch": 0.6163993042133745,
      "grad_norm": 0.0,
      "learning_rate": 1.9182450715114032e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12757
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 61.78125,
      "epoch": 0.6164476227290298,
      "grad_norm": 24.658603003777493,
      "learning_rate": 1.918003478933127e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.20044593513011932,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12758
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.34375,
      "epoch": 0.616495941244685,
      "grad_norm": 0.0,
      "learning_rate": 1.9177618863548512e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12759
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.125,
      "epoch": 0.6165442597603401,
      "grad_norm": 0.0,
      "learning_rate": 1.917520293776575e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 12760
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.21875,
      "epoch": 0.6165925782759953,
      "grad_norm": 0.0,
      "learning_rate": 1.9172787011982992e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12761
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.1875,
      "epoch": 0.6166408967916506,
      "grad_norm": 0.0,
      "learning_rate": 1.917037108620023e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12762
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.8125,
      "epoch": 0.6166892153073058,
      "grad_norm": 0.0,
      "learning_rate": 1.916795516041747e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12763
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.625,
      "epoch": 0.6167375338229609,
      "grad_norm": 0.0,
      "learning_rate": 1.916553923463471e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12764
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.25,
      "epoch": 0.6167858523386162,
      "grad_norm": 0.0,
      "learning_rate": 1.916312330885195e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12765
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.59375,
      "epoch": 0.6168341708542714,
      "grad_norm": 0.0,
      "learning_rate": 1.916070738306919e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12766
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.0,
      "epoch": 0.6168824893699265,
      "grad_norm": 0.0,
      "learning_rate": 1.9158291457286433e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12767
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 57.9375,
      "epoch": 0.6169308078855817,
      "grad_norm": 0.0,
      "learning_rate": 1.915587553150367e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 12768
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.46875,
      "epoch": 0.616979126401237,
      "grad_norm": 0.0,
      "learning_rate": 1.9153459605720913e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12769
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.5,
      "epoch": 0.6170274449168921,
      "grad_norm": 0.0,
      "learning_rate": 1.915104367993815e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12770
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.625,
      "epoch": 0.6170757634325473,
      "grad_norm": 0.0,
      "learning_rate": 1.9148627754155393e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12771
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.125,
      "epoch": 0.6171240819482026,
      "grad_norm": 0.0,
      "learning_rate": 1.9146211828372631e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12772
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.875,
      "epoch": 0.6171724004638578,
      "grad_norm": 0.0,
      "learning_rate": 1.914379590258987e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12773
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.28125,
      "epoch": 0.6172207189795129,
      "grad_norm": 0.0,
      "learning_rate": 1.9141379976807111e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12774
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.78125,
      "epoch": 0.6172690374951681,
      "grad_norm": 0.0,
      "learning_rate": 1.913896405102435e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12775
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.0,
      "epoch": 0.6173173560108234,
      "grad_norm": 0.0,
      "learning_rate": 1.9136548125241592e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12776
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.78125,
      "epoch": 0.6173656745264785,
      "grad_norm": 0.0,
      "learning_rate": 1.9134132199458833e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12777
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 45.34375,
      "epoch": 0.6174139930421337,
      "grad_norm": 25.757263531772026,
      "learning_rate": 1.9131716273676072e-07,
      "loss": -0.0,
      "reward": 2.21875,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12778
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.84375,
      "epoch": 0.617462311557789,
      "grad_norm": 0.0,
      "learning_rate": 1.9129300347893313e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12779
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.875,
      "epoch": 0.6175106300734441,
      "grad_norm": 0.0,
      "learning_rate": 1.9126884422110552e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12780
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.40625,
      "epoch": 0.6175589485890993,
      "grad_norm": 0.0,
      "learning_rate": 1.9124468496327793e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12781
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.1875,
      "epoch": 0.6176072671047546,
      "grad_norm": 0.0,
      "learning_rate": 1.9122052570545032e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12782
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.625,
      "epoch": 0.6176555856204098,
      "grad_norm": 0.0,
      "learning_rate": 1.911963664476227e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12783
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.3125,
      "epoch": 0.6177039041360649,
      "grad_norm": 0.0,
      "learning_rate": 1.9117220718979512e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12784
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.6875,
      "epoch": 0.6177522226517201,
      "grad_norm": 0.0,
      "learning_rate": 1.911480479319675e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12785
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.625,
      "epoch": 0.6178005411673754,
      "grad_norm": 0.0,
      "learning_rate": 1.9112388867413992e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12786
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.15625,
      "epoch": 0.6178488596830305,
      "grad_norm": 0.0,
      "learning_rate": 1.9109972941631233e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12787
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.65625,
      "epoch": 0.6178971781986857,
      "grad_norm": 0.0,
      "learning_rate": 1.9107557015848472e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12788
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.75,
      "epoch": 0.617945496714341,
      "grad_norm": 0.0,
      "learning_rate": 1.9105141090065713e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12789
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.1875,
      "epoch": 0.6179938152299961,
      "grad_norm": 0.0,
      "learning_rate": 1.9102725164282952e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12790
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.40625,
      "epoch": 0.6180421337456513,
      "grad_norm": 0.0,
      "learning_rate": 1.9100309238500193e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12791
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.9375,
      "epoch": 0.6180904522613065,
      "grad_norm": 0.0,
      "learning_rate": 1.9097893312717435e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12792
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.25,
      "epoch": 0.6181387707769618,
      "grad_norm": 0.0,
      "learning_rate": 1.909547738693467e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12793
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.875,
      "epoch": 0.6181870892926169,
      "grad_norm": 0.0,
      "learning_rate": 1.9093061461151912e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12794
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.40625,
      "epoch": 0.6182354078082721,
      "grad_norm": 0.0,
      "learning_rate": 1.909064553536915e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12795
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.96875,
      "epoch": 0.6182837263239274,
      "grad_norm": 0.0,
      "learning_rate": 1.9088229609586392e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12796
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.46875,
      "epoch": 0.6183320448395825,
      "grad_norm": 0.0,
      "learning_rate": 1.9085813683803634e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12797
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.59375,
      "epoch": 0.6183803633552377,
      "grad_norm": 0.0,
      "learning_rate": 1.9083397758020872e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12798
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.21875,
      "epoch": 0.618428681870893,
      "grad_norm": 0.0,
      "learning_rate": 1.9080981832238114e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12799
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.5,
      "epoch": 0.6184770003865481,
      "grad_norm": 0.0,
      "learning_rate": 1.9078565906455352e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12800
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.0,
      "epoch": 0.6185253189022033,
      "grad_norm": 0.0,
      "learning_rate": 1.9076149980672594e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12801
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 49.59375,
      "epoch": 0.6185736374178585,
      "grad_norm": 0.0,
      "learning_rate": 1.9073734054889835e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12802
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 54.875,
      "epoch": 0.6186219559335138,
      "grad_norm": 0.0,
      "learning_rate": 1.907131812910707e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 12803
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.59375,
      "epoch": 0.6186702744491689,
      "grad_norm": 0.0,
      "learning_rate": 1.9068902203324313e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12804
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.96875,
      "epoch": 0.6187185929648241,
      "grad_norm": 0.0,
      "learning_rate": 1.9066486277541551e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12805
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.84375,
      "epoch": 0.6187669114804794,
      "grad_norm": 0.0,
      "learning_rate": 1.9064070351758793e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12806
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.3125,
      "epoch": 0.6188152299961345,
      "grad_norm": 0.0,
      "learning_rate": 1.9061654425976034e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12807
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.4375,
      "epoch": 0.6188635485117897,
      "grad_norm": 0.0,
      "learning_rate": 1.9059238500193273e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12808
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.5625,
      "epoch": 0.6189118670274449,
      "grad_norm": 0.0,
      "learning_rate": 1.9056822574410514e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12809
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.78125,
      "epoch": 0.6189601855431001,
      "grad_norm": 0.0,
      "learning_rate": 1.9054406648627753e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12810
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.4375,
      "epoch": 0.6190085040587553,
      "grad_norm": 0.0,
      "learning_rate": 1.9051990722844994e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12811
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.71875,
      "epoch": 0.6190568225744105,
      "grad_norm": 0.0,
      "learning_rate": 1.9049574797062236e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12812
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.46875,
      "epoch": 0.6191051410900658,
      "grad_norm": 0.0,
      "learning_rate": 1.9047158871279472e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12813
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.65625,
      "epoch": 0.6191534596057209,
      "grad_norm": 0.0,
      "learning_rate": 1.9044742945496713e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12814
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.65625,
      "epoch": 0.6192017781213761,
      "grad_norm": 0.0,
      "learning_rate": 1.9042327019713952e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12815
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.5625,
      "epoch": 0.6192500966370313,
      "grad_norm": 0.0,
      "learning_rate": 1.9039911093931193e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12816
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.65625,
      "epoch": 0.6192984151526865,
      "grad_norm": 0.0,
      "learning_rate": 1.9037495168148434e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12817
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.6875,
      "epoch": 0.6193467336683417,
      "grad_norm": 0.0,
      "learning_rate": 1.9035079242365673e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12818
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.03125,
      "epoch": 0.6193950521839969,
      "grad_norm": 0.0,
      "learning_rate": 1.9032663316582915e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12819
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.34375,
      "epoch": 0.619443370699652,
      "grad_norm": 0.0,
      "learning_rate": 1.9030247390800153e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12820
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.28125,
      "epoch": 0.6194916892153073,
      "grad_norm": 0.0,
      "learning_rate": 1.9027831465017395e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12821
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.4375,
      "epoch": 0.6195400077309625,
      "grad_norm": 0.0,
      "learning_rate": 1.9025415539234636e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12822
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.71875,
      "epoch": 0.6195883262466177,
      "grad_norm": 0.0,
      "learning_rate": 1.9022999613451872e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12823
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.0625,
      "epoch": 0.6196366447622729,
      "grad_norm": 0.0,
      "learning_rate": 1.9020583687669113e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 12824
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.375,
      "epoch": 0.6196849632779281,
      "grad_norm": 0.0,
      "learning_rate": 1.9018167761886352e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12825
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 59.59375,
      "epoch": 0.6197332817935833,
      "grad_norm": 0.0,
      "learning_rate": 1.9015751836103594e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12826
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.6875,
      "epoch": 0.6197816003092385,
      "grad_norm": 0.0,
      "learning_rate": 1.9013335910320835e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12827
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.65625,
      "epoch": 0.6198299188248937,
      "grad_norm": 0.0,
      "learning_rate": 1.9010919984538074e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12828
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 62.84375,
      "epoch": 0.6198782373405489,
      "grad_norm": 0.0,
      "learning_rate": 1.9008504058755315e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 12829
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.5,
      "epoch": 0.619926555856204,
      "grad_norm": 0.0,
      "learning_rate": 1.9006088132972554e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12830
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.9375,
      "epoch": 0.6199748743718593,
      "grad_norm": 0.0,
      "learning_rate": 1.9003672207189795e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12831
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 48.34375,
      "epoch": 0.6200231928875145,
      "grad_norm": 0.0,
      "learning_rate": 1.9001256281407036e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12832
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.75,
      "epoch": 0.6200715114031697,
      "grad_norm": 0.0,
      "learning_rate": 1.8998840355624275e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12833
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.21875,
      "epoch": 0.6201198299188249,
      "grad_norm": 0.0,
      "learning_rate": 1.8996424429841514e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12834
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 55.09375,
      "epoch": 0.6201681484344801,
      "grad_norm": 44.322565503235786,
      "learning_rate": 1.8994008504058753e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.25947636365890503,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12835
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.3125,
      "epoch": 0.6202164669501353,
      "grad_norm": 0.0,
      "learning_rate": 1.8991592578275994e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12836
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 60.90625,
      "epoch": 0.6202647854657904,
      "grad_norm": 0.0,
      "learning_rate": 1.8989176652493235e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 12837
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.84375,
      "epoch": 0.6203131039814457,
      "grad_norm": 0.0,
      "learning_rate": 1.8986760726710474e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12838
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.8125,
      "epoch": 0.6203614224971009,
      "grad_norm": 0.0,
      "learning_rate": 1.8984344800927715e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12839
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.9375,
      "epoch": 0.620409741012756,
      "grad_norm": 0.0,
      "learning_rate": 1.8981928875144954e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12840
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.8125,
      "epoch": 0.6204580595284113,
      "grad_norm": 0.0,
      "learning_rate": 1.8979512949362195e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12841
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.75,
      "epoch": 0.6205063780440665,
      "grad_norm": 0.0,
      "learning_rate": 1.8977097023579437e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12842
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.5625,
      "epoch": 0.6205546965597217,
      "grad_norm": 0.0,
      "learning_rate": 1.8974681097796675e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12843
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.8125,
      "epoch": 0.6206030150753769,
      "grad_norm": 0.0,
      "learning_rate": 1.8972265172013914e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12844
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.71875,
      "epoch": 0.6206513335910321,
      "grad_norm": 0.0,
      "learning_rate": 1.8969849246231153e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12845
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.59375,
      "epoch": 0.6206996521066873,
      "grad_norm": 0.0,
      "learning_rate": 1.8967433320448394e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12846
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.90625,
      "epoch": 0.6207479706223424,
      "grad_norm": 0.0,
      "learning_rate": 1.8965017394665636e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12847
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.34375,
      "epoch": 0.6207962891379977,
      "grad_norm": 0.0,
      "learning_rate": 1.8962601468882874e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12848
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 59.25,
      "epoch": 0.6208446076536529,
      "grad_norm": 0.0,
      "learning_rate": 1.8960185543100116e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12849
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.03125,
      "epoch": 0.620892926169308,
      "grad_norm": 0.0,
      "learning_rate": 1.8957769617317354e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12850
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.6875,
      "epoch": 0.6209412446849633,
      "grad_norm": 0.0,
      "learning_rate": 1.8955353691534596e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12851
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.5625,
      "epoch": 0.6209895632006185,
      "grad_norm": 0.0,
      "learning_rate": 1.8952937765751837e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12852
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 48.1875,
      "epoch": 0.6210378817162737,
      "grad_norm": 0.0,
      "learning_rate": 1.8950521839969076e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12853
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.0625,
      "epoch": 0.6210862002319288,
      "grad_norm": 0.0,
      "learning_rate": 1.8948105914186315e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12854
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.25,
      "epoch": 0.6211345187475841,
      "grad_norm": 0.0,
      "learning_rate": 1.8945689988403553e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12855
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.09375,
      "epoch": 0.6211828372632393,
      "grad_norm": 0.0,
      "learning_rate": 1.8943274062620795e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12856
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.78125,
      "epoch": 0.6212311557788944,
      "grad_norm": 0.0,
      "learning_rate": 1.8940858136838036e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12857
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.4375,
      "epoch": 0.6212794742945497,
      "grad_norm": 0.0,
      "learning_rate": 1.8938442211055275e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12858
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.25,
      "epoch": 0.6213277928102049,
      "grad_norm": 0.0,
      "learning_rate": 1.8936026285272516e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12859
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.03125,
      "epoch": 0.62137611132586,
      "grad_norm": 0.0,
      "learning_rate": 1.8933610359489755e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12860
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 49.375,
      "epoch": 0.6214244298415152,
      "grad_norm": 0.0,
      "learning_rate": 1.8931194433706996e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12861
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.59375,
      "epoch": 0.6214727483571705,
      "grad_norm": 0.0,
      "learning_rate": 1.8928778507924238e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12862
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.15625,
      "epoch": 0.6215210668728257,
      "grad_norm": 0.0,
      "learning_rate": 1.8926362582141476e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12863
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.96875,
      "epoch": 0.6215693853884808,
      "grad_norm": 0.0,
      "learning_rate": 1.8923946656358718e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12864
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.3125,
      "epoch": 0.6216177039041361,
      "grad_norm": 0.0,
      "learning_rate": 1.8921530730575954e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12865
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.53125,
      "epoch": 0.6216660224197913,
      "grad_norm": 0.0,
      "learning_rate": 1.8919114804793195e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12866
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.46875,
      "epoch": 0.6217143409354464,
      "grad_norm": 0.0,
      "learning_rate": 1.8916698879010436e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12867
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.21875,
      "epoch": 0.6217626594511017,
      "grad_norm": 0.0,
      "learning_rate": 1.8914282953227675e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 12868
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.875,
      "epoch": 0.6218109779667569,
      "grad_norm": 0.0,
      "learning_rate": 1.8911867027444917e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12869
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.4375,
      "epoch": 0.621859296482412,
      "grad_norm": 0.0,
      "learning_rate": 1.8909451101662155e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12870
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.53125,
      "epoch": 0.6219076149980672,
      "grad_norm": 0.0,
      "learning_rate": 1.8907035175879397e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12871
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.5625,
      "epoch": 0.6219559335137225,
      "grad_norm": 0.0,
      "learning_rate": 1.8904619250096638e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12872
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.125,
      "epoch": 0.6220042520293777,
      "grad_norm": 0.0,
      "learning_rate": 1.8902203324313877e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12873
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.71875,
      "epoch": 0.6220525705450328,
      "grad_norm": 0.0,
      "learning_rate": 1.8899787398531118e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12874
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.46875,
      "epoch": 0.6221008890606881,
      "grad_norm": 0.0,
      "learning_rate": 1.8897371472748354e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12875
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.125,
      "epoch": 0.6221492075763433,
      "grad_norm": 0.0,
      "learning_rate": 1.8894955546965595e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12876
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.78125,
      "epoch": 0.6221975260919984,
      "grad_norm": 0.0,
      "learning_rate": 1.8892539621182837e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12877
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.03125,
      "epoch": 0.6222458446076536,
      "grad_norm": 0.0,
      "learning_rate": 1.8890123695400076e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12878
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 63.96875,
      "epoch": 0.6222941631233089,
      "grad_norm": 0.0,
      "learning_rate": 1.8887707769617317e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 12879
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 55.125,
      "epoch": 0.622342481638964,
      "grad_norm": 0.0,
      "learning_rate": 1.8885291843834556e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12880
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.75,
      "epoch": 0.6223908001546192,
      "grad_norm": 0.0,
      "learning_rate": 1.8882875918051797e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 12881
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.125,
      "epoch": 0.6224391186702745,
      "grad_norm": 0.0,
      "learning_rate": 1.8880459992269038e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12882
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 54.59375,
      "epoch": 0.6224874371859297,
      "grad_norm": 0.0,
      "learning_rate": 1.8878044066486277e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12883
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.875,
      "epoch": 0.6225357557015848,
      "grad_norm": 0.0,
      "learning_rate": 1.8875628140703518e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12884
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.0,
      "epoch": 0.62258407421724,
      "grad_norm": 0.0,
      "learning_rate": 1.8873212214920754e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12885
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.96875,
      "epoch": 0.6226323927328953,
      "grad_norm": 0.0,
      "learning_rate": 1.8870796289137996e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12886
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.65625,
      "epoch": 0.6226807112485504,
      "grad_norm": 0.0,
      "learning_rate": 1.8868380363355237e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12887
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.40625,
      "epoch": 0.6227290297642056,
      "grad_norm": 0.0,
      "learning_rate": 1.8865964437572476e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12888
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.46875,
      "epoch": 0.6227773482798609,
      "grad_norm": 0.0,
      "learning_rate": 1.8863548511789717e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12889
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.53125,
      "epoch": 0.622825666795516,
      "grad_norm": 0.0,
      "learning_rate": 1.8861132586006956e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12890
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.5,
      "epoch": 0.6228739853111712,
      "grad_norm": 0.0,
      "learning_rate": 1.8858716660224197e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12891
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.71875,
      "epoch": 0.6229223038268265,
      "grad_norm": 0.0,
      "learning_rate": 1.885630073444144e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12892
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.875,
      "epoch": 0.6229706223424817,
      "grad_norm": 0.0,
      "learning_rate": 1.8853884808658677e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12893
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.8125,
      "epoch": 0.6230189408581368,
      "grad_norm": 0.0,
      "learning_rate": 1.885146888287592e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12894
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 56.1875,
      "epoch": 0.623067259373792,
      "grad_norm": 0.0,
      "learning_rate": 1.8849052957093155e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12895
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.90625,
      "epoch": 0.6231155778894473,
      "grad_norm": 0.0,
      "learning_rate": 1.8846637031310396e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12896
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.25,
      "epoch": 0.6231638964051024,
      "grad_norm": 0.0,
      "learning_rate": 1.8844221105527638e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12897
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.78125,
      "epoch": 0.6232122149207576,
      "grad_norm": 0.0,
      "learning_rate": 1.8841805179744876e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12898
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.15625,
      "epoch": 0.6232605334364129,
      "grad_norm": 0.0,
      "learning_rate": 1.8839389253962118e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12899
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 45.125,
      "epoch": 0.623308851952068,
      "grad_norm": 0.0,
      "learning_rate": 1.8836973328179356e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12900
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.09375,
      "epoch": 0.6233571704677232,
      "grad_norm": 0.0,
      "learning_rate": 1.8834557402396598e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12901
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.1875,
      "epoch": 0.6234054889833784,
      "grad_norm": 0.0,
      "learning_rate": 1.883214147661384e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12902
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.8125,
      "epoch": 0.6234538074990337,
      "grad_norm": 0.0,
      "learning_rate": 1.8829725550831078e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12903
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.5,
      "epoch": 0.6235021260146888,
      "grad_norm": 0.0,
      "learning_rate": 1.882730962504832e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12904
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.90625,
      "epoch": 0.623550444530344,
      "grad_norm": 0.0,
      "learning_rate": 1.8824893699265555e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12905
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.71875,
      "epoch": 0.6235987630459993,
      "grad_norm": 0.0,
      "learning_rate": 1.8822477773482797e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12906
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.9375,
      "epoch": 0.6236470815616544,
      "grad_norm": 0.0,
      "learning_rate": 1.8820061847700038e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12907
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.65625,
      "epoch": 0.6236954000773096,
      "grad_norm": 0.0,
      "learning_rate": 1.8817645921917277e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12908
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.5625,
      "epoch": 0.6237437185929648,
      "grad_norm": 0.0,
      "learning_rate": 1.8815229996134518e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12909
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.0,
      "epoch": 0.62379203710862,
      "grad_norm": 0.0,
      "learning_rate": 1.8812814070351757e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12910
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.21875,
      "epoch": 0.6238403556242752,
      "grad_norm": 0.0,
      "learning_rate": 1.8810398144568998e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12911
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 46.34375,
      "epoch": 0.6238886741399304,
      "grad_norm": 0.0,
      "learning_rate": 1.880798221878624e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12912
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.25,
      "epoch": 0.6239369926555857,
      "grad_norm": 0.0,
      "learning_rate": 1.8805566293003478e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12913
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.0,
      "epoch": 0.6239853111712408,
      "grad_norm": 0.0,
      "learning_rate": 1.880315036722072e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12914
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.625,
      "epoch": 0.624033629686896,
      "grad_norm": 0.0,
      "learning_rate": 1.8800734441437958e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12915
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.125,
      "epoch": 0.6240819482025513,
      "grad_norm": 0.0,
      "learning_rate": 1.8798318515655197e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12916
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.25,
      "epoch": 0.6241302667182064,
      "grad_norm": 0.0,
      "learning_rate": 1.8795902589872438e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12917
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.0,
      "epoch": 0.6241785852338616,
      "grad_norm": 0.0,
      "learning_rate": 1.8793486664089677e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12918
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.21875,
      "epoch": 0.6242269037495168,
      "grad_norm": 0.0,
      "learning_rate": 1.8791070738306918e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12919
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.46875,
      "epoch": 0.624275222265172,
      "grad_norm": 0.0,
      "learning_rate": 1.8788654812524157e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12920
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.96875,
      "epoch": 0.6243235407808272,
      "grad_norm": 0.0,
      "learning_rate": 1.8786238886741399e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12921
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.53125,
      "epoch": 0.6243718592964824,
      "grad_norm": 0.0,
      "learning_rate": 1.878382296095864e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12922
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.875,
      "epoch": 0.6244201778121377,
      "grad_norm": 0.0,
      "learning_rate": 1.8781407035175879e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12923
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.09375,
      "epoch": 0.6244684963277928,
      "grad_norm": 0.0,
      "learning_rate": 1.877899110939312e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12924
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.03125,
      "epoch": 0.624516814843448,
      "grad_norm": 0.0,
      "learning_rate": 1.877657518361036e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12925
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.125,
      "epoch": 0.6245651333591032,
      "grad_norm": 0.0,
      "learning_rate": 1.8774159257827597e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12926
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.03125,
      "epoch": 0.6246134518747584,
      "grad_norm": 0.0,
      "learning_rate": 1.877174333204484e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12927
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.65625,
      "epoch": 0.6246617703904136,
      "grad_norm": 0.0,
      "learning_rate": 1.8769327406262078e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12928
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.28125,
      "epoch": 0.6247100889060688,
      "grad_norm": 0.0,
      "learning_rate": 1.876691148047932e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12929
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.71875,
      "epoch": 0.624758407421724,
      "grad_norm": 0.0,
      "learning_rate": 1.8764495554696558e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12930
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.84375,
      "epoch": 0.6248067259373792,
      "grad_norm": 0.0,
      "learning_rate": 1.87620796289138e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12931
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.5625,
      "epoch": 0.6248550444530344,
      "grad_norm": 0.0,
      "learning_rate": 1.875966370313104e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12932
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 43.40625,
      "epoch": 0.6249033629686896,
      "grad_norm": 0.0,
      "learning_rate": 1.875724777734828e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 12933
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.71875,
      "epoch": 0.6249516814843448,
      "grad_norm": 0.0,
      "learning_rate": 1.875483185156552e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12934
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.0625,
      "epoch": 0.625,
      "grad_norm": 0.0,
      "learning_rate": 1.875241592578276e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12935
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.34375,
      "epoch": 0.6250483185156552,
      "grad_norm": 0.0,
      "learning_rate": 1.875e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12936
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.15625,
      "epoch": 0.6250966370313104,
      "grad_norm": 0.0,
      "learning_rate": 1.874758407421724e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12937
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.53125,
      "epoch": 0.6251449555469656,
      "grad_norm": 0.0,
      "learning_rate": 1.8745168148434478e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12938
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.6875,
      "epoch": 0.6251932740626208,
      "grad_norm": 0.0,
      "learning_rate": 1.874275222265172e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12939
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.90625,
      "epoch": 0.625241592578276,
      "grad_norm": 0.0,
      "learning_rate": 1.8740336296868958e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12940
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.09375,
      "epoch": 0.6252899110939312,
      "grad_norm": 0.0,
      "learning_rate": 1.87379203710862e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12941
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.375,
      "epoch": 0.6253382296095864,
      "grad_norm": 0.0,
      "learning_rate": 1.873550444530344e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12942
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.59375,
      "epoch": 0.6253865481252416,
      "grad_norm": 0.0,
      "learning_rate": 1.873308851952068e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12943
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.46875,
      "epoch": 0.6254348666408968,
      "grad_norm": 0.0,
      "learning_rate": 1.873067259373792e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12944
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.3125,
      "epoch": 0.625483185156552,
      "grad_norm": 0.0,
      "learning_rate": 1.872825666795516e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 12945
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.0625,
      "epoch": 0.6255315036722072,
      "grad_norm": 0.0,
      "learning_rate": 1.87258407421724e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12946
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.375,
      "epoch": 0.6255798221878623,
      "grad_norm": 0.0,
      "learning_rate": 1.872342481638964e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12947
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.5,
      "epoch": 0.6256281407035176,
      "grad_norm": 0.0,
      "learning_rate": 1.8721008890606878e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12948
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.59375,
      "epoch": 0.6256764592191728,
      "grad_norm": 0.0,
      "learning_rate": 1.871859296482412e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12949
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.53125,
      "epoch": 0.625724777734828,
      "grad_norm": 0.0,
      "learning_rate": 1.8716177039041358e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12950
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.03125,
      "epoch": 0.6257730962504832,
      "grad_norm": 0.0,
      "learning_rate": 1.87137611132586e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12951
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.25,
      "epoch": 0.6258214147661384,
      "grad_norm": 0.0,
      "learning_rate": 1.871134518747584e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12952
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.65625,
      "epoch": 0.6258697332817936,
      "grad_norm": 0.0,
      "learning_rate": 1.870892926169308e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12953
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.3125,
      "epoch": 0.6259180517974487,
      "grad_norm": 0.0,
      "learning_rate": 1.870651333591032e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12954
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.0,
      "epoch": 0.625966370313104,
      "grad_norm": 0.0,
      "learning_rate": 1.870409741012756e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 12955
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.9375,
      "epoch": 0.6260146888287592,
      "grad_norm": 0.0,
      "learning_rate": 1.87016814843448e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12956
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.875,
      "epoch": 0.6260630073444143,
      "grad_norm": 0.0,
      "learning_rate": 1.869926555856204e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12957
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.375,
      "epoch": 0.6261113258600696,
      "grad_norm": 0.0,
      "learning_rate": 1.869684963277928e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12958
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.0,
      "epoch": 0.6261596443757248,
      "grad_norm": 0.0,
      "learning_rate": 1.869443370699652e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 12959
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.40625,
      "epoch": 0.62620796289138,
      "grad_norm": 0.0,
      "learning_rate": 1.869201778121376e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12960
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.9375,
      "epoch": 0.6262562814070352,
      "grad_norm": 0.0,
      "learning_rate": 1.8689601855431e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 12961
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.3125,
      "epoch": 0.6263045999226904,
      "grad_norm": 0.0,
      "learning_rate": 1.8687185929648241e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12962
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.84375,
      "epoch": 0.6263529184383456,
      "grad_norm": 0.0,
      "learning_rate": 1.868477000386548e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12963
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.40625,
      "epoch": 0.6264012369540007,
      "grad_norm": 0.0,
      "learning_rate": 1.8682354078082722e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 12964
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 48.53125,
      "epoch": 0.626449555469656,
      "grad_norm": 0.0,
      "learning_rate": 1.867993815229996e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12965
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.6875,
      "epoch": 0.6264978739853112,
      "grad_norm": 0.0,
      "learning_rate": 1.8677522226517202e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12966
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.78125,
      "epoch": 0.6265461925009663,
      "grad_norm": 0.0,
      "learning_rate": 1.8675106300734443e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 12967
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.53125,
      "epoch": 0.6265945110166216,
      "grad_norm": 0.0,
      "learning_rate": 1.867269037495168e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 12968
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.03125,
      "epoch": 0.6266428295322768,
      "grad_norm": 0.0,
      "learning_rate": 1.867027444916892e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12969
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.5625,
      "epoch": 0.626691148047932,
      "grad_norm": 0.0,
      "learning_rate": 1.866785852338616e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 12970
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.90625,
      "epoch": 0.6267394665635871,
      "grad_norm": 0.0,
      "learning_rate": 1.86654425976034e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12971
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.46875,
      "epoch": 0.6267877850792424,
      "grad_norm": 0.0,
      "learning_rate": 1.8663026671820642e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12972
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.5625,
      "epoch": 0.6268361035948976,
      "grad_norm": 0.0,
      "learning_rate": 1.866061074603788e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12973
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.1875,
      "epoch": 0.6268844221105527,
      "grad_norm": 0.0,
      "learning_rate": 1.8658194820255122e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 12974
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 53.71875,
      "epoch": 0.626932740626208,
      "grad_norm": 96.0242929677234,
      "learning_rate": 1.865577889447236e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 12975
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.5,
      "epoch": 0.6269810591418632,
      "grad_norm": 0.0,
      "learning_rate": 1.8653362968689602e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 12976
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.03125,
      "epoch": 0.6270293776575183,
      "grad_norm": 0.0,
      "learning_rate": 1.8650947042906843e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12977
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.625,
      "epoch": 0.6270776961731735,
      "grad_norm": 0.0,
      "learning_rate": 1.864853111712408e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12978
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.8125,
      "epoch": 0.6271260146888288,
      "grad_norm": 0.0,
      "learning_rate": 1.864611519134132e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12979
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.78125,
      "epoch": 0.627174333204484,
      "grad_norm": 0.0,
      "learning_rate": 1.864369926555856e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 12980
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.71875,
      "epoch": 0.6272226517201391,
      "grad_norm": 0.0,
      "learning_rate": 1.86412833397758e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12981
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.46875,
      "epoch": 0.6272709702357944,
      "grad_norm": 0.0,
      "learning_rate": 1.8638867413993042e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 12982
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.03125,
      "epoch": 0.6273192887514496,
      "grad_norm": 0.0,
      "learning_rate": 1.863645148821028e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12983
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.5625,
      "epoch": 0.6273676072671047,
      "grad_norm": 0.0,
      "learning_rate": 1.8634035562427522e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 12984
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.5,
      "epoch": 0.62741592578276,
      "grad_norm": 0.0,
      "learning_rate": 1.863161963664476e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12985
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.28125,
      "epoch": 0.6274642442984152,
      "grad_norm": 0.0,
      "learning_rate": 1.8629203710862002e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 12986
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.8125,
      "epoch": 0.6275125628140703,
      "grad_norm": 0.0,
      "learning_rate": 1.8626787785079244e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 12987
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.84375,
      "epoch": 0.6275608813297255,
      "grad_norm": 0.0,
      "learning_rate": 1.862437185929648e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 12988
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.03125,
      "epoch": 0.6276091998453808,
      "grad_norm": 0.0,
      "learning_rate": 1.862195593351372e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12989
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 60.9375,
      "epoch": 0.627657518361036,
      "grad_norm": 0.0,
      "learning_rate": 1.861954000773096e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 12990
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.90625,
      "epoch": 0.6277058368766911,
      "grad_norm": 0.0,
      "learning_rate": 1.86171240819482e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 12991
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.71875,
      "epoch": 0.6277541553923464,
      "grad_norm": 0.0,
      "learning_rate": 1.8614708156165443e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 12992
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.15625,
      "epoch": 0.6278024739080016,
      "grad_norm": 0.0,
      "learning_rate": 1.8612292230382681e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.19727617502212524,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12993
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.4375,
      "epoch": 0.6278507924236567,
      "grad_norm": 0.0,
      "learning_rate": 1.8609876304599923e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 12994
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.5,
      "epoch": 0.6278991109393119,
      "grad_norm": 0.0,
      "learning_rate": 1.8607460378817161e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12995
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 44.75,
      "epoch": 0.6279474294549672,
      "grad_norm": 0.0,
      "learning_rate": 1.8605044453034403e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 12996
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.15625,
      "epoch": 0.6279957479706223,
      "grad_norm": 0.0,
      "learning_rate": 1.8602628527251644e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 12997
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.6875,
      "epoch": 0.6280440664862775,
      "grad_norm": 0.0,
      "learning_rate": 1.860021260146888e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 12998
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.375,
      "epoch": 0.6280923850019328,
      "grad_norm": 0.0,
      "learning_rate": 1.8597796675686122e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 12999
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 56.09375,
      "epoch": 0.628140703517588,
      "grad_norm": 21.533344582800027,
      "learning_rate": 1.859538074990336e-07,
      "loss": -0.0,
      "reward": 2.328125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13000
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.0,
      "epoch": 0.6281890220332431,
      "grad_norm": 0.0,
      "learning_rate": 1.8592964824120602e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13001
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.71875,
      "epoch": 0.6282373405488983,
      "grad_norm": 0.0,
      "learning_rate": 1.8590548898337843e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13002
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.875,
      "epoch": 0.6282856590645536,
      "grad_norm": 0.0,
      "learning_rate": 1.8588132972555082e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13003
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.8125,
      "epoch": 0.6283339775802087,
      "grad_norm": 0.0,
      "learning_rate": 1.8585717046772323e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13004
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.03125,
      "epoch": 0.6283822960958639,
      "grad_norm": 0.0,
      "learning_rate": 1.8583301120989562e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13005
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.78125,
      "epoch": 0.6284306146115192,
      "grad_norm": 0.0,
      "learning_rate": 1.8580885195206803e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13006
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.15625,
      "epoch": 0.6284789331271743,
      "grad_norm": 0.0,
      "learning_rate": 1.8578469269424045e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13007
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.5,
      "epoch": 0.6285272516428295,
      "grad_norm": 0.0,
      "learning_rate": 1.857605334364128e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13008
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.34375,
      "epoch": 0.6285755701584848,
      "grad_norm": 0.0,
      "learning_rate": 1.8573637417858522e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13009
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.28125,
      "epoch": 0.62862388867414,
      "grad_norm": 0.0,
      "learning_rate": 1.857122149207576e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13010
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 53.75,
      "epoch": 0.6286722071897951,
      "grad_norm": 0.0,
      "learning_rate": 1.8568805566293002e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13011
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.65625,
      "epoch": 0.6287205257054503,
      "grad_norm": 0.0,
      "learning_rate": 1.8566389640510243e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13012
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 43.9375,
      "epoch": 0.6287688442211056,
      "grad_norm": 0.0,
      "learning_rate": 1.8563973714727482e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13013
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.78125,
      "epoch": 0.6288171627367607,
      "grad_norm": 0.0,
      "learning_rate": 1.8561557788944724e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13014
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.03125,
      "epoch": 0.6288654812524159,
      "grad_norm": 0.0,
      "learning_rate": 1.8559141863161962e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13015
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.0,
      "epoch": 0.6289137997680712,
      "grad_norm": 0.0,
      "learning_rate": 1.8556725937379204e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13016
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.6875,
      "epoch": 0.6289621182837263,
      "grad_norm": 0.0,
      "learning_rate": 1.8554310011596445e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13017
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.90625,
      "epoch": 0.6290104367993815,
      "grad_norm": 0.0,
      "learning_rate": 1.8551894085813684e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13018
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.3125,
      "epoch": 0.6290587553150367,
      "grad_norm": 0.0,
      "learning_rate": 1.8549478160030922e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13019
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.5625,
      "epoch": 0.629107073830692,
      "grad_norm": 0.0,
      "learning_rate": 1.854706223424816e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13020
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.375,
      "epoch": 0.6291553923463471,
      "grad_norm": 0.0,
      "learning_rate": 1.8544646308465402e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13021
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.40625,
      "epoch": 0.6292037108620023,
      "grad_norm": 0.0,
      "learning_rate": 1.8542230382682644e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13022
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 53.15625,
      "epoch": 0.6292520293776576,
      "grad_norm": 0.0,
      "learning_rate": 1.8539814456899883e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 13023
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.4375,
      "epoch": 0.6293003478933127,
      "grad_norm": 0.0,
      "learning_rate": 1.8537398531117124e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13024
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.6875,
      "epoch": 0.6293486664089679,
      "grad_norm": 0.0,
      "learning_rate": 1.8534982605334363e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13025
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.0,
      "epoch": 0.6293969849246231,
      "grad_norm": 0.0,
      "learning_rate": 1.8532566679551604e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13026
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.46875,
      "epoch": 0.6294453034402783,
      "grad_norm": 0.0,
      "learning_rate": 1.8530150753768845e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13027
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 59.46875,
      "epoch": 0.6294936219559335,
      "grad_norm": 218.30964345613572,
      "learning_rate": 1.8527734827986084e-07,
      "loss": -0.0,
      "reward": 2.359375,
      "reward_std": 0.2561880350112915,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13028
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.375,
      "epoch": 0.6295419404715887,
      "grad_norm": 0.0,
      "learning_rate": 1.8525318902203323e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13029
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 54.75,
      "epoch": 0.629590258987244,
      "grad_norm": 26.69459510915852,
      "learning_rate": 1.8522902976420562e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.22049658000469208,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13030
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.15625,
      "epoch": 0.6296385775028991,
      "grad_norm": 0.0,
      "learning_rate": 1.8520487050637803e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13031
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.53125,
      "epoch": 0.6296868960185543,
      "grad_norm": 0.0,
      "learning_rate": 1.8518071124855044e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13032
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 60.3125,
      "epoch": 0.6297352145342096,
      "grad_norm": 78.23853908502615,
      "learning_rate": 1.8515655199072283e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2725512385368347,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13033
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.21875,
      "epoch": 0.6297835330498647,
      "grad_norm": 0.0,
      "learning_rate": 1.8513239273289524e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13034
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.28125,
      "epoch": 0.6298318515655199,
      "grad_norm": 0.0,
      "learning_rate": 1.8510823347506763e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13035
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 49.71875,
      "epoch": 0.6298801700811751,
      "grad_norm": 0.0,
      "learning_rate": 1.8508407421724004e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13036
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.21875,
      "epoch": 0.6299284885968303,
      "grad_norm": 0.0,
      "learning_rate": 1.8505991495941246e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 13037
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 64.125,
      "epoch": 0.6299768071124855,
      "grad_norm": 0.0,
      "learning_rate": 1.8503575570158484e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 13038
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.0,
      "epoch": 0.6300251256281407,
      "grad_norm": 0.0,
      "learning_rate": 1.8501159644375723e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13039
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.875,
      "epoch": 0.630073444143796,
      "grad_norm": 0.0,
      "learning_rate": 1.8498743718592962e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13040
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 57.28125,
      "epoch": 0.6301217626594511,
      "grad_norm": 0.0,
      "learning_rate": 1.8496327792810203e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 13041
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.90625,
      "epoch": 0.6301700811751063,
      "grad_norm": 0.0,
      "learning_rate": 1.8493911867027445e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13042
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.21875,
      "epoch": 0.6302183996907615,
      "grad_norm": 0.0,
      "learning_rate": 1.8491495941244683e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13043
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.3125,
      "epoch": 0.6302667182064167,
      "grad_norm": 0.0,
      "learning_rate": 1.8489080015461925e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13044
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.5,
      "epoch": 0.6303150367220719,
      "grad_norm": 0.0,
      "learning_rate": 1.8486664089679163e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13045
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.75,
      "epoch": 0.6303633552377271,
      "grad_norm": 0.0,
      "learning_rate": 1.8484248163896405e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13046
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.5625,
      "epoch": 0.6304116737533823,
      "grad_norm": 0.0,
      "learning_rate": 1.8481832238113646e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13047
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.96875,
      "epoch": 0.6304599922690375,
      "grad_norm": 0.0,
      "learning_rate": 1.8479416312330885e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13048
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.15625,
      "epoch": 0.6305083107846927,
      "grad_norm": 0.0,
      "learning_rate": 1.8477000386548126e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13049
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.9375,
      "epoch": 0.630556629300348,
      "grad_norm": 0.0,
      "learning_rate": 1.8474584460765362e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13050
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.34375,
      "epoch": 0.6306049478160031,
      "grad_norm": 0.0,
      "learning_rate": 1.8472168534982604e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 13051
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.03125,
      "epoch": 0.6306532663316583,
      "grad_norm": 0.0,
      "learning_rate": 1.8469752609199845e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13052
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.9375,
      "epoch": 0.6307015848473135,
      "grad_norm": 0.0,
      "learning_rate": 1.8467336683417084e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13053
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.375,
      "epoch": 0.6307499033629687,
      "grad_norm": 0.0,
      "learning_rate": 1.8464920757634325e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13054
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.6875,
      "epoch": 0.6307982218786239,
      "grad_norm": 0.0,
      "learning_rate": 1.8462504831851564e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13055
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.78125,
      "epoch": 0.6308465403942791,
      "grad_norm": 0.0,
      "learning_rate": 1.8460088906068805e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13056
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.4375,
      "epoch": 0.6308948589099342,
      "grad_norm": 0.0,
      "learning_rate": 1.8457672980286047e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13057
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.53125,
      "epoch": 0.6309431774255895,
      "grad_norm": 0.0,
      "learning_rate": 1.8455257054503285e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13058
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.21875,
      "epoch": 0.6309914959412447,
      "grad_norm": 0.0,
      "learning_rate": 1.8452841128720527e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13059
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.875,
      "epoch": 0.6310398144568999,
      "grad_norm": 0.0,
      "learning_rate": 1.8450425202937763e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13060
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.375,
      "epoch": 0.6310881329725551,
      "grad_norm": 0.0,
      "learning_rate": 1.8448009277155004e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13061
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.375,
      "epoch": 0.6311364514882103,
      "grad_norm": 0.0,
      "learning_rate": 1.8445593351372245e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13062
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 45.90625,
      "epoch": 0.6311847700038655,
      "grad_norm": 0.0,
      "learning_rate": 1.8443177425589484e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13063
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.96875,
      "epoch": 0.6312330885195206,
      "grad_norm": 0.0,
      "learning_rate": 1.8440761499806726e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13064
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.78125,
      "epoch": 0.6312814070351759,
      "grad_norm": 0.0,
      "learning_rate": 1.8438345574023964e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 13065
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.9375,
      "epoch": 0.6313297255508311,
      "grad_norm": 0.0,
      "learning_rate": 1.8435929648241206e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13066
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 57.8125,
      "epoch": 0.6313780440664862,
      "grad_norm": 0.0,
      "learning_rate": 1.8433513722458447e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 13067
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.59375,
      "epoch": 0.6314263625821415,
      "grad_norm": 0.0,
      "learning_rate": 1.8431097796675686e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13068
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.03125,
      "epoch": 0.6314746810977967,
      "grad_norm": 0.0,
      "learning_rate": 1.8428681870892927e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13069
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.125,
      "epoch": 0.6315229996134519,
      "grad_norm": 0.0,
      "learning_rate": 1.8426265945110163e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13070
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.6875,
      "epoch": 0.631571318129107,
      "grad_norm": 0.0,
      "learning_rate": 1.8423850019327404e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13071
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.625,
      "epoch": 0.6316196366447623,
      "grad_norm": 0.0,
      "learning_rate": 1.8421434093544646e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 13072
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.46875,
      "epoch": 0.6316679551604175,
      "grad_norm": 0.0,
      "learning_rate": 1.8419018167761885e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13073
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.125,
      "epoch": 0.6317162736760726,
      "grad_norm": 0.0,
      "learning_rate": 1.8416602241979126e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13074
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.96875,
      "epoch": 0.6317645921917279,
      "grad_norm": 0.0,
      "learning_rate": 1.8414186316196365e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13075
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 47.84375,
      "epoch": 0.6318129107073831,
      "grad_norm": 0.0,
      "learning_rate": 1.8411770390413606e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13076
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 48.9375,
      "epoch": 0.6318612292230382,
      "grad_norm": 0.0,
      "learning_rate": 1.8409354464630847e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13077
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.5625,
      "epoch": 0.6319095477386935,
      "grad_norm": 0.0,
      "learning_rate": 1.8406938538848086e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13078
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.25,
      "epoch": 0.6319578662543487,
      "grad_norm": 0.0,
      "learning_rate": 1.8404522613065327e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13079
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.8125,
      "epoch": 0.6320061847700039,
      "grad_norm": 0.0,
      "learning_rate": 1.8402106687282563e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2630178928375244,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13080
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.5,
      "epoch": 0.632054503285659,
      "grad_norm": 0.0,
      "learning_rate": 1.8399690761499805e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13081
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.3125,
      "epoch": 0.6321028218013143,
      "grad_norm": 0.0,
      "learning_rate": 1.8397274835717046e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13082
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.59375,
      "epoch": 0.6321511403169695,
      "grad_norm": 0.0,
      "learning_rate": 1.8394858909934285e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13083
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 45.3125,
      "epoch": 0.6321994588326246,
      "grad_norm": 0.0,
      "learning_rate": 1.8392442984151526e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 13084
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.90625,
      "epoch": 0.6322477773482799,
      "grad_norm": 0.0,
      "learning_rate": 1.8390027058368765e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13085
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.8125,
      "epoch": 0.6322960958639351,
      "grad_norm": 0.0,
      "learning_rate": 1.8387611132586006e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13086
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.59375,
      "epoch": 0.6323444143795902,
      "grad_norm": 0.0,
      "learning_rate": 1.8385195206803248e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13087
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 56.0,
      "epoch": 0.6323927328952454,
      "grad_norm": 0.0,
      "learning_rate": 1.8382779281020486e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13088
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.84375,
      "epoch": 0.6324410514109007,
      "grad_norm": 0.0,
      "learning_rate": 1.8380363355237728e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13089
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.96875,
      "epoch": 0.6324893699265559,
      "grad_norm": 0.0,
      "learning_rate": 1.8377947429454967e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13090
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.78125,
      "epoch": 0.632537688442211,
      "grad_norm": 0.0,
      "learning_rate": 1.8375531503672205e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13091
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.4375,
      "epoch": 0.6325860069578663,
      "grad_norm": 0.0,
      "learning_rate": 1.8373115577889447e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13092
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.40625,
      "epoch": 0.6326343254735215,
      "grad_norm": 0.0,
      "learning_rate": 1.8370699652106685e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13093
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.5625,
      "epoch": 0.6326826439891766,
      "grad_norm": 0.0,
      "learning_rate": 1.8368283726323927e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13094
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.71875,
      "epoch": 0.6327309625048319,
      "grad_norm": 0.0,
      "learning_rate": 1.8365867800541165e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13095
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.40625,
      "epoch": 0.6327792810204871,
      "grad_norm": 0.0,
      "learning_rate": 1.8363451874758407e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13096
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.5625,
      "epoch": 0.6328275995361422,
      "grad_norm": 0.0,
      "learning_rate": 1.8361035948975648e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13097
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.8125,
      "epoch": 0.6328759180517974,
      "grad_norm": 0.0,
      "learning_rate": 1.8358620023192887e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13098
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.78125,
      "epoch": 0.6329242365674527,
      "grad_norm": 0.0,
      "learning_rate": 1.8356204097410128e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13099
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.15625,
      "epoch": 0.6329725550831079,
      "grad_norm": 0.0,
      "learning_rate": 1.8353788171627367e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13100
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.5625,
      "epoch": 0.633020873598763,
      "grad_norm": 0.0,
      "learning_rate": 1.8351372245844606e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13101
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.875,
      "epoch": 0.6330691921144183,
      "grad_norm": 0.0,
      "learning_rate": 1.8348956320061847e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13102
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.5625,
      "epoch": 0.6331175106300735,
      "grad_norm": 0.0,
      "learning_rate": 1.8346540394279086e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13103
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.71875,
      "epoch": 0.6331658291457286,
      "grad_norm": 0.0,
      "learning_rate": 1.8344124468496327e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13104
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.53125,
      "epoch": 0.6332141476613838,
      "grad_norm": 0.0,
      "learning_rate": 1.8341708542713566e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13105
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 51.71875,
      "epoch": 0.6332624661770391,
      "grad_norm": 80.05728965571718,
      "learning_rate": 1.8339292616930807e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13106
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.78125,
      "epoch": 0.6333107846926942,
      "grad_norm": 0.0,
      "learning_rate": 1.8336876691148049e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13107
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 60.25,
      "epoch": 0.6333591032083494,
      "grad_norm": 13.061744059066717,
      "learning_rate": 1.8334460765365287e-07,
      "loss": -0.0,
      "reward": 1.453125,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 13108
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.625,
      "epoch": 0.6334074217240047,
      "grad_norm": 0.0,
      "learning_rate": 1.8332044839582529e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13109
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.84375,
      "epoch": 0.6334557402396599,
      "grad_norm": 0.0,
      "learning_rate": 1.8329628913799767e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13110
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.6875,
      "epoch": 0.633504058755315,
      "grad_norm": 0.0,
      "learning_rate": 1.8327212988017006e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13111
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.90625,
      "epoch": 0.6335523772709702,
      "grad_norm": 0.0,
      "learning_rate": 1.8324797062234247e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13112
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.5625,
      "epoch": 0.6336006957866255,
      "grad_norm": 0.0,
      "learning_rate": 1.8322381136451486e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13113
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.5,
      "epoch": 0.6336490143022806,
      "grad_norm": 0.0,
      "learning_rate": 1.8319965210668727e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13114
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 46.3125,
      "epoch": 0.6336973328179358,
      "grad_norm": 0.0,
      "learning_rate": 1.8317549284885966e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 13115
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.375,
      "epoch": 0.6337456513335911,
      "grad_norm": 0.0,
      "learning_rate": 1.8315133359103208e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13116
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.0,
      "epoch": 0.6337939698492462,
      "grad_norm": 0.0,
      "learning_rate": 1.831271743332045e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13117
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.40625,
      "epoch": 0.6338422883649014,
      "grad_norm": 0.0,
      "learning_rate": 1.8310301507537688e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13118
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.25,
      "epoch": 0.6338906068805567,
      "grad_norm": 0.0,
      "learning_rate": 1.830788558175493e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13119
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.34375,
      "epoch": 0.6339389253962119,
      "grad_norm": 0.0,
      "learning_rate": 1.8305469655972168e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13120
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.8125,
      "epoch": 0.633987243911867,
      "grad_norm": 0.0,
      "learning_rate": 1.830305373018941e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13121
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.6875,
      "epoch": 0.6340355624275222,
      "grad_norm": 0.0,
      "learning_rate": 1.8300637804406648e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13122
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.40625,
      "epoch": 0.6340838809431775,
      "grad_norm": 0.0,
      "learning_rate": 1.8298221878623887e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13123
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.1875,
      "epoch": 0.6341321994588326,
      "grad_norm": 0.0,
      "learning_rate": 1.8295805952841128e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13124
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.9375,
      "epoch": 0.6341805179744878,
      "grad_norm": 0.0,
      "learning_rate": 1.8293390027058367e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13125
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.21875,
      "epoch": 0.634228836490143,
      "grad_norm": 0.0,
      "learning_rate": 1.8290974101275608e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13126
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.28125,
      "epoch": 0.6342771550057982,
      "grad_norm": 0.0,
      "learning_rate": 1.828855817549285e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13127
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.84375,
      "epoch": 0.6343254735214534,
      "grad_norm": 0.0,
      "learning_rate": 1.8286142249710088e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13128
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.46875,
      "epoch": 0.6343737920371086,
      "grad_norm": 0.0,
      "learning_rate": 1.828372632392733e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13129
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.40625,
      "epoch": 0.6344221105527639,
      "grad_norm": 0.0,
      "learning_rate": 1.8281310398144568e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 13130
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.8125,
      "epoch": 0.634470429068419,
      "grad_norm": 0.0,
      "learning_rate": 1.827889447236181e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13131
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.59375,
      "epoch": 0.6345187475840742,
      "grad_norm": 0.0,
      "learning_rate": 1.8276478546579048e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13132
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.40625,
      "epoch": 0.6345670660997295,
      "grad_norm": 0.0,
      "learning_rate": 1.8274062620796287e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13133
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.9375,
      "epoch": 0.6346153846153846,
      "grad_norm": 0.0,
      "learning_rate": 1.8271646695013528e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13134
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.53125,
      "epoch": 0.6346637031310398,
      "grad_norm": 0.0,
      "learning_rate": 1.8269230769230767e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13135
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 46.71875,
      "epoch": 0.634712021646695,
      "grad_norm": 0.0,
      "learning_rate": 1.8266814843448008e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13136
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.40625,
      "epoch": 0.6347603401623502,
      "grad_norm": 0.0,
      "learning_rate": 1.826439891766525e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13137
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.65625,
      "epoch": 0.6348086586780054,
      "grad_norm": 0.0,
      "learning_rate": 1.8261982991882488e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13138
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.21875,
      "epoch": 0.6348569771936606,
      "grad_norm": 0.0,
      "learning_rate": 1.825956706609973e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13139
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.9375,
      "epoch": 0.6349052957093159,
      "grad_norm": 0.0,
      "learning_rate": 1.8257151140316968e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13140
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.625,
      "epoch": 0.634953614224971,
      "grad_norm": 0.0,
      "learning_rate": 1.825473521453421e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13141
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.71875,
      "epoch": 0.6350019327406262,
      "grad_norm": 0.0,
      "learning_rate": 1.8252319288751449e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13142
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 45.0,
      "epoch": 0.6350502512562815,
      "grad_norm": 0.0,
      "learning_rate": 1.8249903362968687e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13143
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.875,
      "epoch": 0.6350985697719366,
      "grad_norm": 0.0,
      "learning_rate": 1.8247487437185929e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13144
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 55.25,
      "epoch": 0.6351468882875918,
      "grad_norm": 1646.5151080351964,
      "learning_rate": 1.8245071511403167e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13145
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 59.5625,
      "epoch": 0.635195206803247,
      "grad_norm": 0.0,
      "learning_rate": 1.824265558562041e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13146
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.3125,
      "epoch": 0.6352435253189022,
      "grad_norm": 0.0,
      "learning_rate": 1.824023965983765e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13147
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 57.0,
      "epoch": 0.6352918438345574,
      "grad_norm": 0.0,
      "learning_rate": 1.823782373405489e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 13148
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.4375,
      "epoch": 0.6353401623502126,
      "grad_norm": 0.0,
      "learning_rate": 1.823540780827213e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13149
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.0,
      "epoch": 0.6353884808658679,
      "grad_norm": 0.0,
      "learning_rate": 1.823299188248937e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13150
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.5625,
      "epoch": 0.635436799381523,
      "grad_norm": 0.0,
      "learning_rate": 1.823057595670661e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13151
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.0,
      "epoch": 0.6354851178971782,
      "grad_norm": 0.0,
      "learning_rate": 1.8228160030923852e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13152
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.46875,
      "epoch": 0.6355334364128334,
      "grad_norm": 0.0,
      "learning_rate": 1.8225744105141088e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13153
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.09375,
      "epoch": 0.6355817549284886,
      "grad_norm": 0.0,
      "learning_rate": 1.822332817935833e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13154
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.21875,
      "epoch": 0.6356300734441438,
      "grad_norm": 0.0,
      "learning_rate": 1.8220912253575568e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13155
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.4375,
      "epoch": 0.635678391959799,
      "grad_norm": 0.0,
      "learning_rate": 1.821849632779281e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13156
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.5625,
      "epoch": 0.6357267104754541,
      "grad_norm": 0.0,
      "learning_rate": 1.821608040201005e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13157
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.625,
      "epoch": 0.6357750289911094,
      "grad_norm": 0.0,
      "learning_rate": 1.821366447622729e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13158
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.09375,
      "epoch": 0.6358233475067646,
      "grad_norm": 0.0,
      "learning_rate": 1.821124855044453e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2651395797729492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13159
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 60.6875,
      "epoch": 0.6358716660224198,
      "grad_norm": 0.0,
      "learning_rate": 1.820883262466177e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13160
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.5,
      "epoch": 0.635919984538075,
      "grad_norm": 0.0,
      "learning_rate": 1.820641669887901e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13161
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 54.9375,
      "epoch": 0.6359683030537302,
      "grad_norm": 0.0,
      "learning_rate": 1.8204000773096252e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 13162
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.0625,
      "epoch": 0.6360166215693854,
      "grad_norm": 0.0,
      "learning_rate": 1.8201584847313488e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13163
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.125,
      "epoch": 0.6360649400850406,
      "grad_norm": 0.0,
      "learning_rate": 1.819916892153073e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 13164
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.53125,
      "epoch": 0.6361132586006958,
      "grad_norm": 0.0,
      "learning_rate": 1.8196752995747968e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13165
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.96875,
      "epoch": 0.636161577116351,
      "grad_norm": 0.0,
      "learning_rate": 1.819433706996521e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13166
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 44.03125,
      "epoch": 0.6362098956320061,
      "grad_norm": 0.0,
      "learning_rate": 1.819192114418245e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13167
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.09375,
      "epoch": 0.6362582141476614,
      "grad_norm": 0.0,
      "learning_rate": 1.818950521839969e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13168
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.375,
      "epoch": 0.6363065326633166,
      "grad_norm": 0.0,
      "learning_rate": 1.818708929261693e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13169
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.03125,
      "epoch": 0.6363548511789718,
      "grad_norm": 0.0,
      "learning_rate": 1.818467336683417e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13170
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.125,
      "epoch": 0.636403169694627,
      "grad_norm": 0.0,
      "learning_rate": 1.818225744105141e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13171
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.96875,
      "epoch": 0.6364514882102822,
      "grad_norm": 0.0,
      "learning_rate": 1.8179841515268652e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13172
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.125,
      "epoch": 0.6364998067259374,
      "grad_norm": 0.0,
      "learning_rate": 1.8177425589485888e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13173
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.125,
      "epoch": 0.6365481252415925,
      "grad_norm": 0.0,
      "learning_rate": 1.817500966370313e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13174
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.3125,
      "epoch": 0.6365964437572478,
      "grad_norm": 0.0,
      "learning_rate": 1.8172593737920369e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13175
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.6875,
      "epoch": 0.636644762272903,
      "grad_norm": 0.0,
      "learning_rate": 1.817017781213761e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13176
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.0625,
      "epoch": 0.6366930807885581,
      "grad_norm": 0.0,
      "learning_rate": 1.816776188635485e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13177
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.15625,
      "epoch": 0.6367413993042134,
      "grad_norm": 0.0,
      "learning_rate": 1.816534596057209e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13178
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.9375,
      "epoch": 0.6367897178198686,
      "grad_norm": 0.0,
      "learning_rate": 1.8162930034789331e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13179
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.5,
      "epoch": 0.6368380363355238,
      "grad_norm": 0.0,
      "learning_rate": 1.816051410900657e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13180
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.53125,
      "epoch": 0.636886354851179,
      "grad_norm": 0.0,
      "learning_rate": 1.8158098183223811e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13181
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.71875,
      "epoch": 0.6369346733668342,
      "grad_norm": 0.0,
      "learning_rate": 1.8155682257441053e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13182
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.09375,
      "epoch": 0.6369829918824894,
      "grad_norm": 0.0,
      "learning_rate": 1.815326633165829e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13183
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.03125,
      "epoch": 0.6370313103981445,
      "grad_norm": 0.0,
      "learning_rate": 1.815085040587553e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13184
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.6875,
      "epoch": 0.6370796289137998,
      "grad_norm": 0.0,
      "learning_rate": 1.814843448009277e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13185
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 52.96875,
      "epoch": 0.637127947429455,
      "grad_norm": 33.590252862268095,
      "learning_rate": 1.814601855431001e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13186
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.53125,
      "epoch": 0.6371762659451101,
      "grad_norm": 0.0,
      "learning_rate": 1.8143602628527252e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 13187
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.5625,
      "epoch": 0.6372245844607654,
      "grad_norm": 0.0,
      "learning_rate": 1.814118670274449e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13188
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 52.875,
      "epoch": 0.6372729029764206,
      "grad_norm": 16.31206361131054,
      "learning_rate": 1.8138770776961732e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.18857255578041077,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13189
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.28125,
      "epoch": 0.6373212214920758,
      "grad_norm": 0.0,
      "learning_rate": 1.813635485117897e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13190
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.46875,
      "epoch": 0.6373695400077309,
      "grad_norm": 0.0,
      "learning_rate": 1.8133938925396212e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13191
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.75,
      "epoch": 0.6374178585233862,
      "grad_norm": 0.0,
      "learning_rate": 1.8131522999613453e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13192
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.0,
      "epoch": 0.6374661770390414,
      "grad_norm": 0.0,
      "learning_rate": 1.8129107073830692e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13193
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.5,
      "epoch": 0.6375144955546965,
      "grad_norm": 0.0,
      "learning_rate": 1.812669114804793e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13194
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.1875,
      "epoch": 0.6375628140703518,
      "grad_norm": 0.0,
      "learning_rate": 1.812427522226517e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13195
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.65625,
      "epoch": 0.637611132586007,
      "grad_norm": 0.0,
      "learning_rate": 1.812185929648241e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13196
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.78125,
      "epoch": 0.6376594511016621,
      "grad_norm": 0.0,
      "learning_rate": 1.8119443370699652e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13197
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.125,
      "epoch": 0.6377077696173173,
      "grad_norm": 0.0,
      "learning_rate": 1.811702744491689e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13198
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 65.8125,
      "epoch": 0.6377560881329726,
      "grad_norm": 0.0,
      "learning_rate": 1.8114611519134132e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13199
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.8125,
      "epoch": 0.6378044066486278,
      "grad_norm": 0.0,
      "learning_rate": 1.811219559335137e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13200
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.875,
      "epoch": 0.6378527251642829,
      "grad_norm": 0.0,
      "learning_rate": 1.8109779667568612e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13201
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.59375,
      "epoch": 0.6379010436799382,
      "grad_norm": 0.0,
      "learning_rate": 1.8107363741785854e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13202
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.625,
      "epoch": 0.6379493621955934,
      "grad_norm": 0.0,
      "learning_rate": 1.8104947816003092e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13203
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.34375,
      "epoch": 0.6379976807112485,
      "grad_norm": 0.0,
      "learning_rate": 1.810253189022033e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13204
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.15625,
      "epoch": 0.6380459992269037,
      "grad_norm": 0.0,
      "learning_rate": 1.810011596443757e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13205
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 62.78125,
      "epoch": 0.638094317742559,
      "grad_norm": 0.0,
      "learning_rate": 1.809770003865481e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13206
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.96875,
      "epoch": 0.6381426362582141,
      "grad_norm": 0.0,
      "learning_rate": 1.8095284112872052e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13207
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 57.125,
      "epoch": 0.6381909547738693,
      "grad_norm": 0.0,
      "learning_rate": 1.809286818708929e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13208
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.5625,
      "epoch": 0.6382392732895246,
      "grad_norm": 0.0,
      "learning_rate": 1.8090452261306533e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13209
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.3125,
      "epoch": 0.6382875918051798,
      "grad_norm": 0.0,
      "learning_rate": 1.808803633552377e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13210
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.4375,
      "epoch": 0.6383359103208349,
      "grad_norm": 0.0,
      "learning_rate": 1.8085620409741013e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13211
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.03125,
      "epoch": 0.6383842288364902,
      "grad_norm": 0.0,
      "learning_rate": 1.8083204483958254e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13212
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.4375,
      "epoch": 0.6384325473521454,
      "grad_norm": 0.0,
      "learning_rate": 1.8080788558175493e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13213
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.46875,
      "epoch": 0.6384808658678005,
      "grad_norm": 0.0,
      "learning_rate": 1.8078372632392731e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13214
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.28125,
      "epoch": 0.6385291843834557,
      "grad_norm": 0.0,
      "learning_rate": 1.807595670660997e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13215
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.3125,
      "epoch": 0.638577502899111,
      "grad_norm": 0.0,
      "learning_rate": 1.8073540780827211e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 13216
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 61.84375,
      "epoch": 0.6386258214147661,
      "grad_norm": 44.704610331098735,
      "learning_rate": 1.8071124855044453e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13217
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.3125,
      "epoch": 0.6386741399304213,
      "grad_norm": 0.0,
      "learning_rate": 1.8068708929261692e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13218
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.96875,
      "epoch": 0.6387224584460766,
      "grad_norm": 0.0,
      "learning_rate": 1.8066293003478933e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13219
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.375,
      "epoch": 0.6387707769617318,
      "grad_norm": 0.0,
      "learning_rate": 1.8063877077696172e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13220
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.40625,
      "epoch": 0.6388190954773869,
      "grad_norm": 0.0,
      "learning_rate": 1.8061461151913413e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 13221
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.5,
      "epoch": 0.6388674139930421,
      "grad_norm": 0.0,
      "learning_rate": 1.8059045226130654e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13222
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.15625,
      "epoch": 0.6389157325086974,
      "grad_norm": 0.0,
      "learning_rate": 1.8056629300347893e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13223
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.875,
      "epoch": 0.6389640510243525,
      "grad_norm": 0.0,
      "learning_rate": 1.8054213374565134e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13224
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.25,
      "epoch": 0.6390123695400077,
      "grad_norm": 0.0,
      "learning_rate": 1.805179744878237e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13225
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.09375,
      "epoch": 0.639060688055663,
      "grad_norm": 0.0,
      "learning_rate": 1.8049381522999612e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13226
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.25,
      "epoch": 0.6391090065713181,
      "grad_norm": 0.0,
      "learning_rate": 1.8046965597216853e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13227
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.46875,
      "epoch": 0.6391573250869733,
      "grad_norm": 0.0,
      "learning_rate": 1.8044549671434092e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13228
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.625,
      "epoch": 0.6392056436026285,
      "grad_norm": 0.0,
      "learning_rate": 1.8042133745651333e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13229
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.625,
      "epoch": 0.6392539621182838,
      "grad_norm": 0.0,
      "learning_rate": 1.8039717819868572e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13230
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.625,
      "epoch": 0.6393022806339389,
      "grad_norm": 0.0,
      "learning_rate": 1.8037301894085813e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13231
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.625,
      "epoch": 0.6393505991495941,
      "grad_norm": 0.0,
      "learning_rate": 1.8034885968303055e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13232
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.46875,
      "epoch": 0.6393989176652494,
      "grad_norm": 0.0,
      "learning_rate": 1.8032470042520293e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13233
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.53125,
      "epoch": 0.6394472361809045,
      "grad_norm": 0.0,
      "learning_rate": 1.8030054116737535e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13234
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.125,
      "epoch": 0.6394955546965597,
      "grad_norm": 0.0,
      "learning_rate": 1.802763819095477e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13235
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 54.59375,
      "epoch": 0.639543873212215,
      "grad_norm": 19.399552555002412,
      "learning_rate": 1.8025222265172012e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.21128857135772705,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 0.96875,
      "step": 13236
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 55.03125,
      "epoch": 0.6395921917278701,
      "grad_norm": 0.0,
      "learning_rate": 1.8022806339389254e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13237
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 54.8125,
      "epoch": 0.6396405102435253,
      "grad_norm": 0.0,
      "learning_rate": 1.8020390413606492e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13238
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.34375,
      "epoch": 0.6396888287591805,
      "grad_norm": 0.0,
      "learning_rate": 1.8017974487823734e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13239
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 50.3125,
      "epoch": 0.6397371472748358,
      "grad_norm": 36.46492099206997,
      "learning_rate": 1.8015558562040972e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.22738909721374512,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13240
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.5,
      "epoch": 0.6397854657904909,
      "grad_norm": 0.0,
      "learning_rate": 1.8013142636258214e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13241
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.03125,
      "epoch": 0.6398337843061461,
      "grad_norm": 0.0,
      "learning_rate": 1.8010726710475455e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13242
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.96875,
      "epoch": 0.6398821028218014,
      "grad_norm": 0.0,
      "learning_rate": 1.8008310784692694e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13243
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.40625,
      "epoch": 0.6399304213374565,
      "grad_norm": 0.0,
      "learning_rate": 1.8005894858909935e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13244
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.625,
      "epoch": 0.6399787398531117,
      "grad_norm": 0.0,
      "learning_rate": 1.800347893312717e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13245
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 55.375,
      "epoch": 0.6400270583687669,
      "grad_norm": 0.0,
      "learning_rate": 1.8001063007344413e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 13246
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.3125,
      "epoch": 0.6400753768844221,
      "grad_norm": 0.0,
      "learning_rate": 1.7998647081561654e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13247
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.15625,
      "epoch": 0.6401236954000773,
      "grad_norm": 0.0,
      "learning_rate": 1.7996231155778893e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13248
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 43.84375,
      "epoch": 0.6401720139157325,
      "grad_norm": 0.0,
      "learning_rate": 1.7993815229996134e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 13249
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 55.6875,
      "epoch": 0.6402203324313878,
      "grad_norm": 0.0,
      "learning_rate": 1.7991399304213373e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13250
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 55.65625,
      "epoch": 0.6402686509470429,
      "grad_norm": 0.0,
      "learning_rate": 1.7988983378430614e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 13251
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.1875,
      "epoch": 0.6403169694626981,
      "grad_norm": 0.0,
      "learning_rate": 1.7986567452647856e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13252
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.5625,
      "epoch": 0.6403652879783533,
      "grad_norm": 0.0,
      "learning_rate": 1.7984151526865094e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13253
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.59375,
      "epoch": 0.6404136064940085,
      "grad_norm": 0.0,
      "learning_rate": 1.7981735601082336e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13254
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.75,
      "epoch": 0.6404619250096637,
      "grad_norm": 0.0,
      "learning_rate": 1.7979319675299572e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13255
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.65625,
      "epoch": 0.6405102435253189,
      "grad_norm": 0.0,
      "learning_rate": 1.7976903749516813e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13256
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.53125,
      "epoch": 0.640558562040974,
      "grad_norm": 0.0,
      "learning_rate": 1.7974487823734054e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13257
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.78125,
      "epoch": 0.6406068805566293,
      "grad_norm": 0.0,
      "learning_rate": 1.7972071897951293e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13258
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.625,
      "epoch": 0.6406551990722845,
      "grad_norm": 0.0,
      "learning_rate": 1.7969655972168535e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13259
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.34375,
      "epoch": 0.6407035175879398,
      "grad_norm": 0.0,
      "learning_rate": 1.7967240046385773e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13260
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.6875,
      "epoch": 0.6407518361035949,
      "grad_norm": 0.0,
      "learning_rate": 1.7964824120603015e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13261
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.1875,
      "epoch": 0.6408001546192501,
      "grad_norm": 0.0,
      "learning_rate": 1.7962408194820256e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13262
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.4375,
      "epoch": 0.6408484731349053,
      "grad_norm": 0.0,
      "learning_rate": 1.7959992269037495e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13263
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.65625,
      "epoch": 0.6408967916505605,
      "grad_norm": 0.0,
      "learning_rate": 1.7957576343254736e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13264
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.6875,
      "epoch": 0.6409451101662157,
      "grad_norm": 0.0,
      "learning_rate": 1.7955160417471972e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13265
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.21875,
      "epoch": 0.6409934286818709,
      "grad_norm": 0.0,
      "learning_rate": 1.7952744491689213e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13266
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.3125,
      "epoch": 0.641041747197526,
      "grad_norm": 0.0,
      "learning_rate": 1.7950328565906455e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13267
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.09375,
      "epoch": 0.6410900657131813,
      "grad_norm": 0.0,
      "learning_rate": 1.7947912640123694e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13268
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.5,
      "epoch": 0.6411383842288365,
      "grad_norm": 0.0,
      "learning_rate": 1.7945496714340935e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13269
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.46875,
      "epoch": 0.6411867027444917,
      "grad_norm": 0.0,
      "learning_rate": 1.7943080788558174e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13270
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.0625,
      "epoch": 0.6412350212601469,
      "grad_norm": 0.0,
      "learning_rate": 1.7940664862775415e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 13271
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.4375,
      "epoch": 0.6412833397758021,
      "grad_norm": 0.0,
      "learning_rate": 1.7938248936992656e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13272
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.9375,
      "epoch": 0.6413316582914573,
      "grad_norm": 0.0,
      "learning_rate": 1.7935833011209895e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13273
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.5,
      "epoch": 0.6413799768071125,
      "grad_norm": 0.0,
      "learning_rate": 1.7933417085427136e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13274
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.71875,
      "epoch": 0.6414282953227677,
      "grad_norm": 0.0,
      "learning_rate": 1.7931001159644375e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13275
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.75,
      "epoch": 0.6414766138384229,
      "grad_norm": 0.0,
      "learning_rate": 1.7928585233861614e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13276
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.84375,
      "epoch": 0.641524932354078,
      "grad_norm": 0.0,
      "learning_rate": 1.7926169308078855e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13277
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 41.25,
      "epoch": 0.6415732508697333,
      "grad_norm": 0.0,
      "learning_rate": 1.7923753382296094e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13278
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.125,
      "epoch": 0.6416215693853885,
      "grad_norm": 0.0,
      "learning_rate": 1.7921337456513335e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13279
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.96875,
      "epoch": 0.6416698879010437,
      "grad_norm": 0.0,
      "learning_rate": 1.7918921530730574e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13280
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 44.46875,
      "epoch": 0.6417182064166989,
      "grad_norm": 0.0,
      "learning_rate": 1.7916505604947815e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13281
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.78125,
      "epoch": 0.6417665249323541,
      "grad_norm": 0.0,
      "learning_rate": 1.7914089679165057e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13282
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.8125,
      "epoch": 0.6418148434480093,
      "grad_norm": 0.0,
      "learning_rate": 1.7911673753382295e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13283
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.0,
      "epoch": 0.6418631619636644,
      "grad_norm": 0.0,
      "learning_rate": 1.7909257827599537e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13284
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.9375,
      "epoch": 0.6419114804793197,
      "grad_norm": 0.0,
      "learning_rate": 1.7906841901816776e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13285
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.3125,
      "epoch": 0.6419597989949749,
      "grad_norm": 0.0,
      "learning_rate": 1.7904425976034014e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13286
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.46875,
      "epoch": 0.64200811751063,
      "grad_norm": 0.0,
      "learning_rate": 1.7902010050251256e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13287
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.15625,
      "epoch": 0.6420564360262853,
      "grad_norm": 0.0,
      "learning_rate": 1.7899594124468494e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13288
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.4375,
      "epoch": 0.6421047545419405,
      "grad_norm": 0.0,
      "learning_rate": 1.7897178198685736e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13289
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.625,
      "epoch": 0.6421530730575957,
      "grad_norm": 0.0,
      "learning_rate": 1.7894762272902974e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13290
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.0,
      "epoch": 0.6422013915732508,
      "grad_norm": 0.0,
      "learning_rate": 1.7892346347120216e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13291
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.53125,
      "epoch": 0.6422497100889061,
      "grad_norm": 0.0,
      "learning_rate": 1.7889930421337457e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13292
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.78125,
      "epoch": 0.6422980286045613,
      "grad_norm": 0.0,
      "learning_rate": 1.7887514495554696e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13293
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.4375,
      "epoch": 0.6423463471202164,
      "grad_norm": 0.0,
      "learning_rate": 1.7885098569771937e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13294
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.5,
      "epoch": 0.6423946656358717,
      "grad_norm": 0.0,
      "learning_rate": 1.7882682643989176e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13295
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 53.3125,
      "epoch": 0.6424429841515269,
      "grad_norm": 0.0,
      "learning_rate": 1.7880266718206417e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13296
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 52.78125,
      "epoch": 0.642491302667182,
      "grad_norm": 0.0,
      "learning_rate": 1.7877850792423656e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13297
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.09375,
      "epoch": 0.6425396211828373,
      "grad_norm": 0.0,
      "learning_rate": 1.7875434866640895e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13298
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 46.5625,
      "epoch": 0.6425879396984925,
      "grad_norm": 0.0,
      "learning_rate": 1.7873018940858136e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13299
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.34375,
      "epoch": 0.6426362582141477,
      "grad_norm": 0.0,
      "learning_rate": 1.7870603015075375e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13300
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 49.4375,
      "epoch": 0.6426845767298028,
      "grad_norm": 0.0,
      "learning_rate": 1.7868187089292616e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13301
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.3125,
      "epoch": 0.6427328952454581,
      "grad_norm": 0.0,
      "learning_rate": 1.7865771163509858e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13302
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.3125,
      "epoch": 0.6427812137611133,
      "grad_norm": 0.0,
      "learning_rate": 1.7863355237727096e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13303
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.71875,
      "epoch": 0.6428295322767684,
      "grad_norm": 0.0,
      "learning_rate": 1.7860939311944338e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13304
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.125,
      "epoch": 0.6428778507924237,
      "grad_norm": 0.0,
      "learning_rate": 1.7858523386161576e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13305
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.0,
      "epoch": 0.6429261693080789,
      "grad_norm": 0.0,
      "learning_rate": 1.7856107460378818e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13306
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.34375,
      "epoch": 0.642974487823734,
      "grad_norm": 0.0,
      "learning_rate": 1.7853691534596056e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13307
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.28125,
      "epoch": 0.6430228063393892,
      "grad_norm": 0.0,
      "learning_rate": 1.7851275608813295e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13308
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.28125,
      "epoch": 0.6430711248550445,
      "grad_norm": 0.0,
      "learning_rate": 1.7848859683030536e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13309
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.84375,
      "epoch": 0.6431194433706997,
      "grad_norm": 0.0,
      "learning_rate": 1.7846443757247775e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13310
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.53125,
      "epoch": 0.6431677618863548,
      "grad_norm": 0.0,
      "learning_rate": 1.7844027831465017e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13311
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.9375,
      "epoch": 0.6432160804020101,
      "grad_norm": 0.0,
      "learning_rate": 1.7841611905682258e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13312
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.8125,
      "epoch": 0.6432643989176653,
      "grad_norm": 0.0,
      "learning_rate": 1.7839195979899497e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13313
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.0,
      "epoch": 0.6433127174333204,
      "grad_norm": 0.0,
      "learning_rate": 1.7836780054116738e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13314
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.46875,
      "epoch": 0.6433610359489756,
      "grad_norm": 0.0,
      "learning_rate": 1.7834364128333977e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13315
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.96875,
      "epoch": 0.6434093544646309,
      "grad_norm": 0.0,
      "learning_rate": 1.7831948202551218e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13316
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.75,
      "epoch": 0.643457672980286,
      "grad_norm": 0.0,
      "learning_rate": 1.7829532276768457e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13317
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.5,
      "epoch": 0.6435059914959412,
      "grad_norm": 0.0,
      "learning_rate": 1.7827116350985696e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13318
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.96875,
      "epoch": 0.6435543100115965,
      "grad_norm": 0.0,
      "learning_rate": 1.7824700425202937e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13319
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.84375,
      "epoch": 0.6436026285272517,
      "grad_norm": 0.0,
      "learning_rate": 1.7822284499420176e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13320
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.5625,
      "epoch": 0.6436509470429068,
      "grad_norm": 0.0,
      "learning_rate": 1.7819868573637417e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13321
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.875,
      "epoch": 0.643699265558562,
      "grad_norm": 0.0,
      "learning_rate": 1.7817452647854658e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13322
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.5,
      "epoch": 0.6437475840742173,
      "grad_norm": 0.0,
      "learning_rate": 1.7815036722071897e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13323
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.40625,
      "epoch": 0.6437959025898724,
      "grad_norm": 0.0,
      "learning_rate": 1.7812620796289138e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13324
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.25,
      "epoch": 0.6438442211055276,
      "grad_norm": 0.0,
      "learning_rate": 1.7810204870506377e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13325
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.53125,
      "epoch": 0.6438925396211829,
      "grad_norm": 0.0,
      "learning_rate": 1.7807788944723618e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13326
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.53125,
      "epoch": 0.643940858136838,
      "grad_norm": 0.0,
      "learning_rate": 1.780537301894086e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13327
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.625,
      "epoch": 0.6439891766524932,
      "grad_norm": 0.0,
      "learning_rate": 1.7802957093158096e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13328
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.90625,
      "epoch": 0.6440374951681485,
      "grad_norm": 0.0,
      "learning_rate": 1.7800541167375337e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13329
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.78125,
      "epoch": 0.6440858136838037,
      "grad_norm": 0.0,
      "learning_rate": 1.7798125241592576e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13330
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.4375,
      "epoch": 0.6441341321994588,
      "grad_norm": 0.0,
      "learning_rate": 1.7795709315809817e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13331
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 47.09375,
      "epoch": 0.644182450715114,
      "grad_norm": 0.0,
      "learning_rate": 1.779329339002706e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13332
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.40625,
      "epoch": 0.6442307692307693,
      "grad_norm": 0.0,
      "learning_rate": 1.7790877464244297e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13333
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.4375,
      "epoch": 0.6442790877464244,
      "grad_norm": 0.0,
      "learning_rate": 1.778846153846154e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13334
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.46875,
      "epoch": 0.6443274062620796,
      "grad_norm": 0.0,
      "learning_rate": 1.7786045612678777e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13335
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.3125,
      "epoch": 0.6443757247777349,
      "grad_norm": 0.0,
      "learning_rate": 1.778362968689602e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13336
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.96875,
      "epoch": 0.64442404329339,
      "grad_norm": 0.0,
      "learning_rate": 1.778121376111326e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13337
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.0625,
      "epoch": 0.6444723618090452,
      "grad_norm": 0.0,
      "learning_rate": 1.7778797835330496e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13338
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.4375,
      "epoch": 0.6445206803247004,
      "grad_norm": 0.0,
      "learning_rate": 1.7776381909547738e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13339
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.3125,
      "epoch": 0.6445689988403557,
      "grad_norm": 0.0,
      "learning_rate": 1.7773965983764976e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13340
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.28125,
      "epoch": 0.6446173173560108,
      "grad_norm": 0.0,
      "learning_rate": 1.7771550057982218e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13341
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.0,
      "epoch": 0.644665635871666,
      "grad_norm": 0.0,
      "learning_rate": 1.776913413219946e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13342
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.8125,
      "epoch": 0.6447139543873213,
      "grad_norm": 0.0,
      "learning_rate": 1.7766718206416698e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13343
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.34375,
      "epoch": 0.6447622729029764,
      "grad_norm": 0.0,
      "learning_rate": 1.776430228063394e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13344
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.4375,
      "epoch": 0.6448105914186316,
      "grad_norm": 0.0,
      "learning_rate": 1.7761886354851178e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13345
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.90625,
      "epoch": 0.6448589099342869,
      "grad_norm": 0.0,
      "learning_rate": 1.775947042906842e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13346
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.59375,
      "epoch": 0.644907228449942,
      "grad_norm": 0.0,
      "learning_rate": 1.775705450328566e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13347
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.8125,
      "epoch": 0.6449555469655972,
      "grad_norm": 0.0,
      "learning_rate": 1.7754638577502897e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13348
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.40625,
      "epoch": 0.6450038654812524,
      "grad_norm": 0.0,
      "learning_rate": 1.7752222651720138e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13349
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.15625,
      "epoch": 0.6450521839969077,
      "grad_norm": 0.0,
      "learning_rate": 1.7749806725937377e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 13350
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 43.5625,
      "epoch": 0.6451005025125628,
      "grad_norm": 157.6007039048536,
      "learning_rate": 1.7747390800154618e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.2404065728187561,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13351
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.21875,
      "epoch": 0.645148821028218,
      "grad_norm": 0.0,
      "learning_rate": 1.774497487437186e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13352
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 50.78125,
      "epoch": 0.6451971395438733,
      "grad_norm": 0.0,
      "learning_rate": 1.7742558948589098e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13353
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.25,
      "epoch": 0.6452454580595284,
      "grad_norm": 0.0,
      "learning_rate": 1.774014302280634e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13354
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.40625,
      "epoch": 0.6452937765751836,
      "grad_norm": 0.0,
      "learning_rate": 1.7737727097023578e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13355
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 51.21875,
      "epoch": 0.6453420950908388,
      "grad_norm": 93.26547886347376,
      "learning_rate": 1.773531117124082e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13356
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 60.78125,
      "epoch": 0.645390413606494,
      "grad_norm": 0.0,
      "learning_rate": 1.773289524545806e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13357
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 45.75,
      "epoch": 0.6454387321221492,
      "grad_norm": 0.0,
      "learning_rate": 1.7730479319675297e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13358
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.03125,
      "epoch": 0.6454870506378044,
      "grad_norm": 0.0,
      "learning_rate": 1.7728063393892538e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13359
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.375,
      "epoch": 0.6455353691534597,
      "grad_norm": 0.0,
      "learning_rate": 1.7725647468109777e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13360
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.59375,
      "epoch": 0.6455836876691148,
      "grad_norm": 0.0,
      "learning_rate": 1.7723231542327019e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13361
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.46875,
      "epoch": 0.64563200618477,
      "grad_norm": 0.0,
      "learning_rate": 1.772081561654426e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13362
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 46.65625,
      "epoch": 0.6456803247004252,
      "grad_norm": 0.0,
      "learning_rate": 1.7718399690761499e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 13363
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.65625,
      "epoch": 0.6457286432160804,
      "grad_norm": 0.0,
      "learning_rate": 1.771598376497874e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13364
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.03125,
      "epoch": 0.6457769617317356,
      "grad_norm": 0.0,
      "learning_rate": 1.7713567839195979e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13365
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 49.375,
      "epoch": 0.6458252802473908,
      "grad_norm": 0.0,
      "learning_rate": 1.771115191341322e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13366
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.1875,
      "epoch": 0.645873598763046,
      "grad_norm": 0.0,
      "learning_rate": 1.7708735987630461e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13367
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.25,
      "epoch": 0.6459219172787012,
      "grad_norm": 0.0,
      "learning_rate": 1.7706320061847697e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13368
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.4375,
      "epoch": 0.6459702357943564,
      "grad_norm": 0.0,
      "learning_rate": 1.770390413606494e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13369
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.4375,
      "epoch": 0.6460185543100117,
      "grad_norm": 0.0,
      "learning_rate": 1.7701488210282178e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13370
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 54.90625,
      "epoch": 0.6460668728256668,
      "grad_norm": 0.0,
      "learning_rate": 1.769907228449942e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13371
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.78125,
      "epoch": 0.646115191341322,
      "grad_norm": 0.0,
      "learning_rate": 1.769665635871666e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 13372
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 42.96875,
      "epoch": 0.6461635098569772,
      "grad_norm": 0.0,
      "learning_rate": 1.76942404329339e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13373
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 53.21875,
      "epoch": 0.6462118283726324,
      "grad_norm": 0.0,
      "learning_rate": 1.769182450715114e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13374
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.875,
      "epoch": 0.6462601468882876,
      "grad_norm": 0.0,
      "learning_rate": 1.768940858136838e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13375
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.625,
      "epoch": 0.6463084654039428,
      "grad_norm": 0.0,
      "learning_rate": 1.768699265558562e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13376
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.15625,
      "epoch": 0.6463567839195979,
      "grad_norm": 0.0,
      "learning_rate": 1.7684576729802862e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13377
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.28125,
      "epoch": 0.6464051024352532,
      "grad_norm": 0.0,
      "learning_rate": 1.76821608040201e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13378
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 45.71875,
      "epoch": 0.6464534209509084,
      "grad_norm": 0.0,
      "learning_rate": 1.767974487823734e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13379
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.875,
      "epoch": 0.6465017394665636,
      "grad_norm": 0.0,
      "learning_rate": 1.7677328952454578e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13380
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 42.96875,
      "epoch": 0.6465500579822188,
      "grad_norm": 0.0,
      "learning_rate": 1.767491302667182e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13381
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.21875,
      "epoch": 0.646598376497874,
      "grad_norm": 0.0,
      "learning_rate": 1.767249710088906e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13382
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.46875,
      "epoch": 0.6466466950135292,
      "grad_norm": 0.0,
      "learning_rate": 1.76700811751063e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13383
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.1875,
      "epoch": 0.6466950135291843,
      "grad_norm": 0.0,
      "learning_rate": 1.766766524932354e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13384
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.5625,
      "epoch": 0.6467433320448396,
      "grad_norm": 0.0,
      "learning_rate": 1.766524932354078e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13385
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 43.40625,
      "epoch": 0.6467916505604948,
      "grad_norm": 0.0,
      "learning_rate": 1.766283339775802e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13386
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.5625,
      "epoch": 0.6468399690761499,
      "grad_norm": 0.0,
      "learning_rate": 1.7660417471975262e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13387
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.75,
      "epoch": 0.6468882875918052,
      "grad_norm": 0.0,
      "learning_rate": 1.76580015461925e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13388
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.84375,
      "epoch": 0.6469366061074604,
      "grad_norm": 0.0,
      "learning_rate": 1.765558562040974e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13389
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.21875,
      "epoch": 0.6469849246231156,
      "grad_norm": 0.0,
      "learning_rate": 1.7653169694626978e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13390
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.71875,
      "epoch": 0.6470332431387708,
      "grad_norm": 0.0,
      "learning_rate": 1.765075376884422e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 13391
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 43.03125,
      "epoch": 0.647081561654426,
      "grad_norm": 0.0,
      "learning_rate": 1.764833784306146e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13392
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.0,
      "epoch": 0.6471298801700812,
      "grad_norm": 0.0,
      "learning_rate": 1.76459219172787e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13393
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.0,
      "epoch": 0.6471781986857363,
      "grad_norm": 0.0,
      "learning_rate": 1.764350599149594e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13394
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.65625,
      "epoch": 0.6472265172013916,
      "grad_norm": 0.0,
      "learning_rate": 1.764109006571318e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13395
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.03125,
      "epoch": 0.6472748357170468,
      "grad_norm": 0.0,
      "learning_rate": 1.763867413993042e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13396
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.78125,
      "epoch": 0.6473231542327019,
      "grad_norm": 0.0,
      "learning_rate": 1.7636258214147663e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13397
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.96875,
      "epoch": 0.6473714727483572,
      "grad_norm": 0.0,
      "learning_rate": 1.76338422883649e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13398
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.4375,
      "epoch": 0.6474197912640124,
      "grad_norm": 0.0,
      "learning_rate": 1.7631426362582143e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13399
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.5625,
      "epoch": 0.6474681097796676,
      "grad_norm": 0.0,
      "learning_rate": 1.762901043679938e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13400
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.5,
      "epoch": 0.6475164282953227,
      "grad_norm": 0.0,
      "learning_rate": 1.762659451101662e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13401
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.125,
      "epoch": 0.647564746810978,
      "grad_norm": 0.0,
      "learning_rate": 1.7624178585233861e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13402
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 49.40625,
      "epoch": 0.6476130653266332,
      "grad_norm": 0.0,
      "learning_rate": 1.76217626594511e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13403
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.65625,
      "epoch": 0.6476613838422883,
      "grad_norm": 0.0,
      "learning_rate": 1.7619346733668342e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13404
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 41.9375,
      "epoch": 0.6477097023579436,
      "grad_norm": 0.0,
      "learning_rate": 1.761693080788558e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13405
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.53125,
      "epoch": 0.6477580208735988,
      "grad_norm": 0.0,
      "learning_rate": 1.7614514882102822e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13406
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.96875,
      "epoch": 0.6478063393892539,
      "grad_norm": 0.0,
      "learning_rate": 1.7612098956320063e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13407
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.78125,
      "epoch": 0.6478546579049091,
      "grad_norm": 0.0,
      "learning_rate": 1.7609683030537302e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13408
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.125,
      "epoch": 0.6479029764205644,
      "grad_norm": 0.0,
      "learning_rate": 1.7607267104754543e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13409
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.15625,
      "epoch": 0.6479512949362196,
      "grad_norm": 0.0,
      "learning_rate": 1.760485117897178e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 13410
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.3125,
      "epoch": 0.6479996134518747,
      "grad_norm": 0.0,
      "learning_rate": 1.760243525318902e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13411
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.625,
      "epoch": 0.64804793196753,
      "grad_norm": 0.0,
      "learning_rate": 1.7600019327406262e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13412
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 44.46875,
      "epoch": 0.6480962504831852,
      "grad_norm": 0.0,
      "learning_rate": 1.75976034016235e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 13413
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.53125,
      "epoch": 0.6481445689988403,
      "grad_norm": 0.0,
      "learning_rate": 1.7595187475840742e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13414
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 55.875,
      "epoch": 0.6481928875144956,
      "grad_norm": 20.186697593259723,
      "learning_rate": 1.759277155005798e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13415
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 47.84375,
      "epoch": 0.6482412060301508,
      "grad_norm": 0.0,
      "learning_rate": 1.7590355624275222e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13416
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 46.6875,
      "epoch": 0.6482895245458059,
      "grad_norm": 0.0,
      "learning_rate": 1.7587939698492463e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13417
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 59.6875,
      "epoch": 0.6483378430614611,
      "grad_norm": 0.0,
      "learning_rate": 1.7585523772709702e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 13418
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.21875,
      "epoch": 0.6483861615771164,
      "grad_norm": 0.0,
      "learning_rate": 1.7583107846926943e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13419
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.40625,
      "epoch": 0.6484344800927716,
      "grad_norm": 0.0,
      "learning_rate": 1.758069192114418e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13420
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.15625,
      "epoch": 0.6484827986084267,
      "grad_norm": 0.0,
      "learning_rate": 1.757827599536142e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13421
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.75,
      "epoch": 0.648531117124082,
      "grad_norm": 0.0,
      "learning_rate": 1.7575860069578662e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13422
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.21875,
      "epoch": 0.6485794356397372,
      "grad_norm": 0.0,
      "learning_rate": 1.75734441437959e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13423
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.46875,
      "epoch": 0.6486277541553923,
      "grad_norm": 0.0,
      "learning_rate": 1.7571028218013142e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13424
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 43.4375,
      "epoch": 0.6486760726710475,
      "grad_norm": 0.0,
      "learning_rate": 1.756861229223038e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13425
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.8125,
      "epoch": 0.6487243911867028,
      "grad_norm": 0.0,
      "learning_rate": 1.7566196366447622e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13426
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 58.09375,
      "epoch": 0.6487727097023579,
      "grad_norm": 48.99924034074932,
      "learning_rate": 1.7563780440664864e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.3061639666557312,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13427
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.65625,
      "epoch": 0.6488210282180131,
      "grad_norm": 0.0,
      "learning_rate": 1.7561364514882102e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13428
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.46875,
      "epoch": 0.6488693467336684,
      "grad_norm": 0.0,
      "learning_rate": 1.7558948589099344e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13429
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.5,
      "epoch": 0.6489176652493236,
      "grad_norm": 0.0,
      "learning_rate": 1.755653266331658e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13430
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.78125,
      "epoch": 0.6489659837649787,
      "grad_norm": 0.0,
      "learning_rate": 1.755411673753382e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13431
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.9375,
      "epoch": 0.649014302280634,
      "grad_norm": 0.0,
      "learning_rate": 1.7551700811751063e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13432
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.125,
      "epoch": 0.6490626207962892,
      "grad_norm": 0.0,
      "learning_rate": 1.7549284885968301e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13433
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.78125,
      "epoch": 0.6491109393119443,
      "grad_norm": 0.0,
      "learning_rate": 1.7546868960185543e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13434
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 46.75,
      "epoch": 0.6491592578275995,
      "grad_norm": 0.0,
      "learning_rate": 1.7544453034402781e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13435
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.1875,
      "epoch": 0.6492075763432548,
      "grad_norm": 0.0,
      "learning_rate": 1.7542037108620023e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13436
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.75,
      "epoch": 0.6492558948589099,
      "grad_norm": 0.0,
      "learning_rate": 1.7539621182837264e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13437
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.28125,
      "epoch": 0.6493042133745651,
      "grad_norm": 0.0,
      "learning_rate": 1.7537205257054503e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13438
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.65625,
      "epoch": 0.6493525318902204,
      "grad_norm": 0.0,
      "learning_rate": 1.7534789331271744e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13439
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 53.0625,
      "epoch": 0.6494008504058756,
      "grad_norm": 59.16068945276529,
      "learning_rate": 1.753237340548898e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13440
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 52.9375,
      "epoch": 0.6494491689215307,
      "grad_norm": 0.0,
      "learning_rate": 1.7529957479706222e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13441
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.25,
      "epoch": 0.6494974874371859,
      "grad_norm": 0.0,
      "learning_rate": 1.7527541553923463e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13442
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.9375,
      "epoch": 0.6495458059528412,
      "grad_norm": 0.0,
      "learning_rate": 1.7525125628140702e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13443
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 45.65625,
      "epoch": 0.6495941244684963,
      "grad_norm": 0.0,
      "learning_rate": 1.7522709702357943e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13444
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.8125,
      "epoch": 0.6496424429841515,
      "grad_norm": 0.0,
      "learning_rate": 1.7520293776575182e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13445
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.34375,
      "epoch": 0.6496907614998068,
      "grad_norm": 0.0,
      "learning_rate": 1.7517877850792423e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 13446
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 60.34375,
      "epoch": 0.6497390800154619,
      "grad_norm": 0.0,
      "learning_rate": 1.7515461925009665e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13447
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.71875,
      "epoch": 0.6497873985311171,
      "grad_norm": 0.0,
      "learning_rate": 1.7513045999226903e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13448
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.34375,
      "epoch": 0.6498357170467723,
      "grad_norm": 0.0,
      "learning_rate": 1.7510630073444145e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13449
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 49.96875,
      "epoch": 0.6498840355624276,
      "grad_norm": 247.4567346561202,
      "learning_rate": 1.7508214147661383e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.28247910737991333,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13450
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.28125,
      "epoch": 0.6499323540780827,
      "grad_norm": 0.0,
      "learning_rate": 1.7505798221878622e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13451
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.875,
      "epoch": 0.6499806725937379,
      "grad_norm": 0.0,
      "learning_rate": 1.7503382296095863e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13452
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.75,
      "epoch": 0.6500289911093932,
      "grad_norm": 0.0,
      "learning_rate": 1.7500966370313102e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13453
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.90625,
      "epoch": 0.6500773096250483,
      "grad_norm": 0.0,
      "learning_rate": 1.7498550444530343e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13454
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 45.375,
      "epoch": 0.6501256281407035,
      "grad_norm": 0.0,
      "learning_rate": 1.7496134518747582e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13455
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.4375,
      "epoch": 0.6501739466563587,
      "grad_norm": 0.0,
      "learning_rate": 1.7493718592964824e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13456
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.03125,
      "epoch": 0.6502222651720139,
      "grad_norm": 0.0,
      "learning_rate": 1.7491302667182065e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13457
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.15625,
      "epoch": 0.6502705836876691,
      "grad_norm": 0.0,
      "learning_rate": 1.7488886741399304e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13458
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.75,
      "epoch": 0.6503189022033243,
      "grad_norm": 0.0,
      "learning_rate": 1.7486470815616545e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13459
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.53125,
      "epoch": 0.6503672207189796,
      "grad_norm": 0.0,
      "learning_rate": 1.7484054889833784e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13460
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 56.59375,
      "epoch": 0.6504155392346347,
      "grad_norm": 116.07481688029165,
      "learning_rate": 1.7481638964051022e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 13461
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.34375,
      "epoch": 0.6504638577502899,
      "grad_norm": 0.0,
      "learning_rate": 1.7479223038268264e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13462
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.8125,
      "epoch": 0.6505121762659452,
      "grad_norm": 0.0,
      "learning_rate": 1.7476807112485503e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13463
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.59375,
      "epoch": 0.6505604947816003,
      "grad_norm": 0.0,
      "learning_rate": 1.7474391186702744e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13464
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 60.5625,
      "epoch": 0.6506088132972555,
      "grad_norm": 0.0,
      "learning_rate": 1.7471975260919983e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13465
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.5,
      "epoch": 0.6506571318129107,
      "grad_norm": 0.0,
      "learning_rate": 1.7469559335137224e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13466
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.0,
      "epoch": 0.6507054503285659,
      "grad_norm": 0.0,
      "learning_rate": 1.7467143409354465e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13467
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.90625,
      "epoch": 0.6507537688442211,
      "grad_norm": 0.0,
      "learning_rate": 1.7464727483571704e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13468
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.5,
      "epoch": 0.6508020873598763,
      "grad_norm": 0.0,
      "learning_rate": 1.7462311557788945e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13469
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.875,
      "epoch": 0.6508504058755316,
      "grad_norm": 0.0,
      "learning_rate": 1.7459895632006184e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13470
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.375,
      "epoch": 0.6508987243911867,
      "grad_norm": 0.0,
      "learning_rate": 1.7457479706223423e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13471
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 45.96875,
      "epoch": 0.6509470429068419,
      "grad_norm": 0.0,
      "learning_rate": 1.7455063780440664e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13472
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 46.25,
      "epoch": 0.6509953614224971,
      "grad_norm": 0.0,
      "learning_rate": 1.7452647854657903e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13473
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.9375,
      "epoch": 0.6510436799381523,
      "grad_norm": 0.0,
      "learning_rate": 1.7450231928875144e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13474
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.53125,
      "epoch": 0.6510919984538075,
      "grad_norm": 0.0,
      "learning_rate": 1.7447816003092383e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13475
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 46.71875,
      "epoch": 0.6511403169694627,
      "grad_norm": 0.0,
      "learning_rate": 1.7445400077309624e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 13476
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 44.4375,
      "epoch": 0.6511886354851179,
      "grad_norm": 0.0,
      "learning_rate": 1.7442984151526866e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 13477
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.84375,
      "epoch": 0.6512369540007731,
      "grad_norm": 0.0,
      "learning_rate": 1.7440568225744104e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13478
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 49.65625,
      "epoch": 0.6512852725164283,
      "grad_norm": 0.0,
      "learning_rate": 1.7438152299961346e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13479
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 45.03125,
      "epoch": 0.6513335910320835,
      "grad_norm": 0.0,
      "learning_rate": 1.7435736374178585e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13480
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.28125,
      "epoch": 0.6513819095477387,
      "grad_norm": 0.0,
      "learning_rate": 1.7433320448395826e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13481
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 51.90625,
      "epoch": 0.6514302280633939,
      "grad_norm": 0.0,
      "learning_rate": 1.7430904522613065e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13482
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 45.375,
      "epoch": 0.6514785465790491,
      "grad_norm": 0.0,
      "learning_rate": 1.7428488596830303e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13483
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.15625,
      "epoch": 0.6515268650947043,
      "grad_norm": 0.0,
      "learning_rate": 1.7426072671047545e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13484
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 41.375,
      "epoch": 0.6515751836103595,
      "grad_norm": 0.0,
      "learning_rate": 1.7423656745264783e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13485
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.25,
      "epoch": 0.6516235021260147,
      "grad_norm": 0.0,
      "learning_rate": 1.7421240819482025e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13486
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.53125,
      "epoch": 0.6516718206416698,
      "grad_norm": 0.0,
      "learning_rate": 1.7418824893699266e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13487
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.53125,
      "epoch": 0.6517201391573251,
      "grad_norm": 0.0,
      "learning_rate": 1.7416408967916505e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13488
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.125,
      "epoch": 0.6517684576729803,
      "grad_norm": 0.0,
      "learning_rate": 1.7413993042133746e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13489
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.0,
      "epoch": 0.6518167761886355,
      "grad_norm": 0.0,
      "learning_rate": 1.7411577116350985e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13490
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.59375,
      "epoch": 0.6518650947042907,
      "grad_norm": 0.0,
      "learning_rate": 1.7409161190568226e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13491
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 51.78125,
      "epoch": 0.6519134132199459,
      "grad_norm": 56.38899090458687,
      "learning_rate": 1.7406745264785465e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.22152717411518097,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13492
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 54.375,
      "epoch": 0.6519617317356011,
      "grad_norm": 0.0,
      "learning_rate": 1.7404329339002704e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13493
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.25,
      "epoch": 0.6520100502512562,
      "grad_norm": 0.0,
      "learning_rate": 1.7401913413219945e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13494
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.40625,
      "epoch": 0.6520583687669115,
      "grad_norm": 0.0,
      "learning_rate": 1.7399497487437184e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13495
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.34375,
      "epoch": 0.6521066872825667,
      "grad_norm": 0.0,
      "learning_rate": 1.7397081561654425e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13496
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.96875,
      "epoch": 0.6521550057982218,
      "grad_norm": 0.0,
      "learning_rate": 1.7394665635871667e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13497
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 55.0,
      "epoch": 0.6522033243138771,
      "grad_norm": 0.0,
      "learning_rate": 1.7392249710088905e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13498
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.21875,
      "epoch": 0.6522516428295323,
      "grad_norm": 0.0,
      "learning_rate": 1.7389833784306147e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13499
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.875,
      "epoch": 0.6522999613451875,
      "grad_norm": 0.0,
      "learning_rate": 1.7387417858523385e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13500
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 44.3125,
      "epoch": 0.6523482798608427,
      "grad_norm": 0.0,
      "learning_rate": 1.7385001932740627e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13501
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.5625,
      "epoch": 0.6523965983764979,
      "grad_norm": 0.0,
      "learning_rate": 1.7382586006957865e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13502
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.6875,
      "epoch": 0.6524449168921531,
      "grad_norm": 0.0,
      "learning_rate": 1.7380170081175104e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 13503
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.6875,
      "epoch": 0.6524932354078082,
      "grad_norm": 0.0,
      "learning_rate": 1.7377754155392345e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13504
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.625,
      "epoch": 0.6525415539234635,
      "grad_norm": 0.0,
      "learning_rate": 1.7375338229609584e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13505
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.4375,
      "epoch": 0.6525898724391187,
      "grad_norm": 0.0,
      "learning_rate": 1.7372922303826826e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13506
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.71875,
      "epoch": 0.6526381909547738,
      "grad_norm": 0.0,
      "learning_rate": 1.7370506378044067e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13507
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.21875,
      "epoch": 0.652686509470429,
      "grad_norm": 0.0,
      "learning_rate": 1.7368090452261306e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13508
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.96875,
      "epoch": 0.6527348279860843,
      "grad_norm": 0.0,
      "learning_rate": 1.7365674526478547e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13509
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.375,
      "epoch": 0.6527831465017395,
      "grad_norm": 0.0,
      "learning_rate": 1.7363258600695786e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13510
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.75,
      "epoch": 0.6528314650173946,
      "grad_norm": 0.0,
      "learning_rate": 1.7360842674913027e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13511
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.5625,
      "epoch": 0.6528797835330499,
      "grad_norm": 0.0,
      "learning_rate": 1.7358426749130268e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13512
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.4375,
      "epoch": 0.6529281020487051,
      "grad_norm": 0.0,
      "learning_rate": 1.7356010823347504e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13513
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.96875,
      "epoch": 0.6529764205643602,
      "grad_norm": 0.0,
      "learning_rate": 1.7353594897564746e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13514
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.3125,
      "epoch": 0.6530247390800155,
      "grad_norm": 0.0,
      "learning_rate": 1.7351178971781985e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 13515
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.46875,
      "epoch": 0.6530730575956707,
      "grad_norm": 0.0,
      "learning_rate": 1.7348763045999226e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13516
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.8125,
      "epoch": 0.6531213761113258,
      "grad_norm": 0.0,
      "learning_rate": 1.7346347120216467e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13517
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.03125,
      "epoch": 0.653169694626981,
      "grad_norm": 0.0,
      "learning_rate": 1.7343931194433706e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13518
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.65625,
      "epoch": 0.6532180131426363,
      "grad_norm": 0.0,
      "learning_rate": 1.7341515268650947e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13519
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.3125,
      "epoch": 0.6532663316582915,
      "grad_norm": 0.0,
      "learning_rate": 1.7339099342868186e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13520
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.0,
      "epoch": 0.6533146501739466,
      "grad_norm": 0.0,
      "learning_rate": 1.7336683417085427e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13521
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.5625,
      "epoch": 0.6533629686896019,
      "grad_norm": 0.0,
      "learning_rate": 1.733426749130267e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13522
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.25,
      "epoch": 0.6534112872052571,
      "grad_norm": 0.0,
      "learning_rate": 1.7331851565519905e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13523
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.0625,
      "epoch": 0.6534596057209122,
      "grad_norm": 0.0,
      "learning_rate": 1.7329435639737146e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13524
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.1875,
      "epoch": 0.6535079242365675,
      "grad_norm": 0.0,
      "learning_rate": 1.7327019713954385e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13525
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.0,
      "epoch": 0.6535562427522227,
      "grad_norm": 0.0,
      "learning_rate": 1.7324603788171626e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13526
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 49.6875,
      "epoch": 0.6536045612678778,
      "grad_norm": 0.0,
      "learning_rate": 1.7322187862388868e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13527
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.21875,
      "epoch": 0.653652879783533,
      "grad_norm": 0.0,
      "learning_rate": 1.7319771936606106e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13528
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.5625,
      "epoch": 0.6537011982991883,
      "grad_norm": 0.0,
      "learning_rate": 1.7317356010823348e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13529
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.21875,
      "epoch": 0.6537495168148435,
      "grad_norm": 0.0,
      "learning_rate": 1.7314940085040586e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13530
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 57.78125,
      "epoch": 0.6537978353304986,
      "grad_norm": 20.717791053264406,
      "learning_rate": 1.7312524159257828e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13531
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.75,
      "epoch": 0.6538461538461539,
      "grad_norm": 0.0,
      "learning_rate": 1.731010823347507e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13532
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.46875,
      "epoch": 0.6538944723618091,
      "grad_norm": 0.0,
      "learning_rate": 1.7307692307692305e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13533
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.5,
      "epoch": 0.6539427908774642,
      "grad_norm": 0.0,
      "learning_rate": 1.7305276381909547e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13534
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.1875,
      "epoch": 0.6539911093931194,
      "grad_norm": 0.0,
      "learning_rate": 1.7302860456126785e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13535
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.5,
      "epoch": 0.6540394279087747,
      "grad_norm": 0.0,
      "learning_rate": 1.7300444530344027e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13536
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.6875,
      "epoch": 0.6540877464244298,
      "grad_norm": 0.0,
      "learning_rate": 1.7298028604561268e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13537
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.28125,
      "epoch": 0.654136064940085,
      "grad_norm": 0.0,
      "learning_rate": 1.7295612678778507e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13538
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.03125,
      "epoch": 0.6541843834557403,
      "grad_norm": 0.0,
      "learning_rate": 1.7293196752995748e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13539
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 57.125,
      "epoch": 0.6542327019713955,
      "grad_norm": 49.86165792739702,
      "learning_rate": 1.7290780827212987e-07,
      "loss": -0.0,
      "reward": 1.59375,
      "reward_std": 0.22152717411518097,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13540
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.84375,
      "epoch": 0.6542810204870506,
      "grad_norm": 0.0,
      "learning_rate": 1.7288364901430228e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13541
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.15625,
      "epoch": 0.6543293390027058,
      "grad_norm": 0.0,
      "learning_rate": 1.728594897564747e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13542
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.96875,
      "epoch": 0.6543776575183611,
      "grad_norm": 0.0,
      "learning_rate": 1.7283533049864706e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13543
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 60.96875,
      "epoch": 0.6544259760340162,
      "grad_norm": 0.0,
      "learning_rate": 1.7281117124081947e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13544
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.59375,
      "epoch": 0.6544742945496714,
      "grad_norm": 0.0,
      "learning_rate": 1.7278701198299186e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13545
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.5625,
      "epoch": 0.6545226130653267,
      "grad_norm": 0.0,
      "learning_rate": 1.7276285272516427e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13546
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.5625,
      "epoch": 0.6545709315809818,
      "grad_norm": 0.0,
      "learning_rate": 1.7273869346733668e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13547
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.65625,
      "epoch": 0.654619250096637,
      "grad_norm": 0.0,
      "learning_rate": 1.7271453420950907e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13548
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.21875,
      "epoch": 0.6546675686122922,
      "grad_norm": 0.0,
      "learning_rate": 1.7269037495168149e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13549
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.5,
      "epoch": 0.6547158871279475,
      "grad_norm": 0.0,
      "learning_rate": 1.7266621569385387e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13550
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 59.46875,
      "epoch": 0.6547642056436026,
      "grad_norm": 0.0,
      "learning_rate": 1.7264205643602629e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13551
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.9375,
      "epoch": 0.6548125241592578,
      "grad_norm": 0.0,
      "learning_rate": 1.726178971781987e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13552
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.40625,
      "epoch": 0.6548608426749131,
      "grad_norm": 0.0,
      "learning_rate": 1.725937379203711e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13553
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.78125,
      "epoch": 0.6549091611905682,
      "grad_norm": 0.0,
      "learning_rate": 1.7256957866254347e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13554
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.5,
      "epoch": 0.6549574797062234,
      "grad_norm": 0.0,
      "learning_rate": 1.7254541940471586e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13555
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.9375,
      "epoch": 0.6550057982218787,
      "grad_norm": 0.0,
      "learning_rate": 1.7252126014688828e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13556
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.0,
      "epoch": 0.6550541167375338,
      "grad_norm": 0.0,
      "learning_rate": 1.724971008890607e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13557
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.03125,
      "epoch": 0.655102435253189,
      "grad_norm": 0.0,
      "learning_rate": 1.7247294163123308e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13558
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.75,
      "epoch": 0.6551507537688442,
      "grad_norm": 0.0,
      "learning_rate": 1.724487823734055e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13559
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.34375,
      "epoch": 0.6551990722844995,
      "grad_norm": 0.0,
      "learning_rate": 1.7242462311557788e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13560
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 55.65625,
      "epoch": 0.6552473908001546,
      "grad_norm": 13.03349398292567,
      "learning_rate": 1.724004638577503e-07,
      "loss": -0.0,
      "reward": 1.25,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.1875,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 13561
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 42.09375,
      "epoch": 0.6552957093158098,
      "grad_norm": 0.0,
      "learning_rate": 1.723763045999227e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13562
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.96875,
      "epoch": 0.6553440278314651,
      "grad_norm": 0.0,
      "learning_rate": 1.723521453420951e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13563
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.59375,
      "epoch": 0.6553923463471202,
      "grad_norm": 0.0,
      "learning_rate": 1.7232798608426748e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13564
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.5625,
      "epoch": 0.6554406648627754,
      "grad_norm": 0.0,
      "learning_rate": 1.7230382682643987e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13565
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.0,
      "epoch": 0.6554889833784306,
      "grad_norm": 0.0,
      "learning_rate": 1.7227966756861228e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13566
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.90625,
      "epoch": 0.6555373018940858,
      "grad_norm": 0.0,
      "learning_rate": 1.722555083107847e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13567
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.625,
      "epoch": 0.655585620409741,
      "grad_norm": 0.0,
      "learning_rate": 1.7223134905295708e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13568
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.59375,
      "epoch": 0.6556339389253962,
      "grad_norm": 0.0,
      "learning_rate": 1.722071897951295e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13569
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.875,
      "epoch": 0.6556822574410515,
      "grad_norm": 0.0,
      "learning_rate": 1.7218303053730188e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13570
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.78125,
      "epoch": 0.6557305759567066,
      "grad_norm": 0.0,
      "learning_rate": 1.721588712794743e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13571
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.71875,
      "epoch": 0.6557788944723618,
      "grad_norm": 0.0,
      "learning_rate": 1.721347120216467e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13572
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.0625,
      "epoch": 0.655827212988017,
      "grad_norm": 0.0,
      "learning_rate": 1.721105527638191e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13573
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 57.0,
      "epoch": 0.6558755315036722,
      "grad_norm": 0.0,
      "learning_rate": 1.7208639350599148e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13574
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 62.78125,
      "epoch": 0.6559238500193274,
      "grad_norm": 0.0,
      "learning_rate": 1.7206223424816387e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 13575
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.96875,
      "epoch": 0.6559721685349826,
      "grad_norm": 0.0,
      "learning_rate": 1.7203807499033628e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13576
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.5,
      "epoch": 0.6560204870506378,
      "grad_norm": 0.0,
      "learning_rate": 1.720139157325087e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13577
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.5625,
      "epoch": 0.656068805566293,
      "grad_norm": 0.0,
      "learning_rate": 1.7198975647468108e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13578
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.28125,
      "epoch": 0.6561171240819482,
      "grad_norm": 0.0,
      "learning_rate": 1.719655972168535e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13579
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.21875,
      "epoch": 0.6561654425976035,
      "grad_norm": 0.0,
      "learning_rate": 1.7194143795902588e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13580
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.78125,
      "epoch": 0.6562137611132586,
      "grad_norm": 0.0,
      "learning_rate": 1.719172787011983e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13581
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.625,
      "epoch": 0.6562620796289138,
      "grad_norm": 0.0,
      "learning_rate": 1.718931194433707e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13582
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.0625,
      "epoch": 0.656310398144569,
      "grad_norm": 0.0,
      "learning_rate": 1.718689601855431e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13583
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.65625,
      "epoch": 0.6563587166602242,
      "grad_norm": 0.0,
      "learning_rate": 1.718448009277155e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13584
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.4375,
      "epoch": 0.6564070351758794,
      "grad_norm": 0.0,
      "learning_rate": 1.7182064166988787e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13585
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.0,
      "epoch": 0.6564553536915346,
      "grad_norm": 0.0,
      "learning_rate": 1.717964824120603e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13586
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 46.0625,
      "epoch": 0.6565036722071897,
      "grad_norm": 0.0,
      "learning_rate": 1.717723231542327e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13587
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.34375,
      "epoch": 0.656551990722845,
      "grad_norm": 0.0,
      "learning_rate": 1.717481638964051e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13588
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.03125,
      "epoch": 0.6566003092385002,
      "grad_norm": 0.0,
      "learning_rate": 1.717240046385775e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13589
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.46875,
      "epoch": 0.6566486277541554,
      "grad_norm": 0.0,
      "learning_rate": 1.716998453807499e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13590
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.5,
      "epoch": 0.6566969462698106,
      "grad_norm": 0.0,
      "learning_rate": 1.716756861229223e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13591
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.125,
      "epoch": 0.6567452647854658,
      "grad_norm": 0.0,
      "learning_rate": 1.7165152686509472e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13592
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.03125,
      "epoch": 0.656793583301121,
      "grad_norm": 0.0,
      "learning_rate": 1.716273676072671e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 13593
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.09375,
      "epoch": 0.6568419018167762,
      "grad_norm": 0.0,
      "learning_rate": 1.7160320834943952e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13594
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.34375,
      "epoch": 0.6568902203324314,
      "grad_norm": 0.0,
      "learning_rate": 1.7157904909161188e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13595
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.75,
      "epoch": 0.6569385388480866,
      "grad_norm": 0.0,
      "learning_rate": 1.715548898337843e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 13596
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 46.5625,
      "epoch": 0.6569868573637417,
      "grad_norm": 0.0,
      "learning_rate": 1.715307305759567e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13597
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.125,
      "epoch": 0.657035175879397,
      "grad_norm": 0.0,
      "learning_rate": 1.715065713181291e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 13598
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.21875,
      "epoch": 0.6570834943950522,
      "grad_norm": 0.0,
      "learning_rate": 1.714824120603015e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13599
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.96875,
      "epoch": 0.6571318129107074,
      "grad_norm": 0.0,
      "learning_rate": 1.714582528024739e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13600
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.8125,
      "epoch": 0.6571801314263626,
      "grad_norm": 0.0,
      "learning_rate": 1.714340935446463e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13601
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.25,
      "epoch": 0.6572284499420178,
      "grad_norm": 0.0,
      "learning_rate": 1.7140993428681872e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13602
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.9375,
      "epoch": 0.657276768457673,
      "grad_norm": 0.0,
      "learning_rate": 1.713857750289911e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13603
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.90625,
      "epoch": 0.6573250869733281,
      "grad_norm": 0.0,
      "learning_rate": 1.7136161577116352e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13604
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 53.0,
      "epoch": 0.6573734054889834,
      "grad_norm": 0.0,
      "learning_rate": 1.7133745651333588e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 13605
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.75,
      "epoch": 0.6574217240046386,
      "grad_norm": 0.0,
      "learning_rate": 1.713132972555083e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13606
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.9375,
      "epoch": 0.6574700425202937,
      "grad_norm": 0.0,
      "learning_rate": 1.712891379976807e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13607
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.25,
      "epoch": 0.657518361035949,
      "grad_norm": 0.0,
      "learning_rate": 1.712649787398531e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13608
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.78125,
      "epoch": 0.6575666795516042,
      "grad_norm": 0.0,
      "learning_rate": 1.712408194820255e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13609
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.03125,
      "epoch": 0.6576149980672594,
      "grad_norm": 0.0,
      "learning_rate": 1.712166602241979e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13610
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.90625,
      "epoch": 0.6576633165829145,
      "grad_norm": 0.0,
      "learning_rate": 1.711925009663703e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 13611
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.40625,
      "epoch": 0.6577116350985698,
      "grad_norm": 0.0,
      "learning_rate": 1.7116834170854272e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13612
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.46875,
      "epoch": 0.657759953614225,
      "grad_norm": 0.0,
      "learning_rate": 1.711441824507151e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13613
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.375,
      "epoch": 0.6578082721298801,
      "grad_norm": 0.0,
      "learning_rate": 1.7112002319288752e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13614
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.96875,
      "epoch": 0.6578565906455354,
      "grad_norm": 0.0,
      "learning_rate": 1.7109586393505989e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13615
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.8125,
      "epoch": 0.6579049091611906,
      "grad_norm": 0.0,
      "learning_rate": 1.710717046772323e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13616
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.96875,
      "epoch": 0.6579532276768457,
      "grad_norm": 0.0,
      "learning_rate": 1.710475454194047e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13617
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 41.53125,
      "epoch": 0.658001546192501,
      "grad_norm": 0.0,
      "learning_rate": 1.710233861615771e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13618
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.90625,
      "epoch": 0.6580498647081562,
      "grad_norm": 0.0,
      "learning_rate": 1.709992269037495e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13619
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 46.25,
      "epoch": 0.6580981832238114,
      "grad_norm": 0.0,
      "learning_rate": 1.709750676459219e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13620
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.03125,
      "epoch": 0.6581465017394665,
      "grad_norm": 0.0,
      "learning_rate": 1.7095090838809431e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13621
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.21875,
      "epoch": 0.6581948202551218,
      "grad_norm": 0.0,
      "learning_rate": 1.7092674913026673e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13622
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.78125,
      "epoch": 0.658243138770777,
      "grad_norm": 0.0,
      "learning_rate": 1.7090258987243911e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 13623
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.3125,
      "epoch": 0.6582914572864321,
      "grad_norm": 0.0,
      "learning_rate": 1.7087843061461153e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13624
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.0,
      "epoch": 0.6583397758020874,
      "grad_norm": 0.0,
      "learning_rate": 1.7085427135678392e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13625
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.0625,
      "epoch": 0.6583880943177426,
      "grad_norm": 0.0,
      "learning_rate": 1.708301120989563e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13626
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.90625,
      "epoch": 0.6584364128333977,
      "grad_norm": 0.0,
      "learning_rate": 1.7080595284112872e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13627
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.625,
      "epoch": 0.6584847313490529,
      "grad_norm": 0.0,
      "learning_rate": 1.707817935833011e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13628
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 48.4375,
      "epoch": 0.6585330498647082,
      "grad_norm": 0.0,
      "learning_rate": 1.7075763432547352e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13629
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.03125,
      "epoch": 0.6585813683803634,
      "grad_norm": 0.0,
      "learning_rate": 1.707334750676459e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13630
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.1875,
      "epoch": 0.6586296868960185,
      "grad_norm": 0.0,
      "learning_rate": 1.7070931580981832e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13631
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 42.75,
      "epoch": 0.6586780054116738,
      "grad_norm": 0.0,
      "learning_rate": 1.7068515655199073e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13632
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.90625,
      "epoch": 0.658726323927329,
      "grad_norm": 0.0,
      "learning_rate": 1.7066099729416312e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13633
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 48.75,
      "epoch": 0.6587746424429841,
      "grad_norm": 0.0,
      "learning_rate": 1.7063683803633553e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 13634
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.5625,
      "epoch": 0.6588229609586393,
      "grad_norm": 0.0,
      "learning_rate": 1.7061267877850792e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13635
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.09375,
      "epoch": 0.6588712794742946,
      "grad_norm": 0.0,
      "learning_rate": 1.705885195206803e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13636
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.84375,
      "epoch": 0.6589195979899497,
      "grad_norm": 0.0,
      "learning_rate": 1.7056436026285272e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13637
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.0,
      "epoch": 0.6589679165056049,
      "grad_norm": 0.0,
      "learning_rate": 1.705402010050251e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13638
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.5625,
      "epoch": 0.6590162350212602,
      "grad_norm": 0.0,
      "learning_rate": 1.7051604174719752e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13639
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.125,
      "epoch": 0.6590645535369154,
      "grad_norm": 0.0,
      "learning_rate": 1.704918824893699e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13640
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.21875,
      "epoch": 0.6591128720525705,
      "grad_norm": 0.0,
      "learning_rate": 1.7046772323154232e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13641
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.0,
      "epoch": 0.6591611905682258,
      "grad_norm": 0.0,
      "learning_rate": 1.7044356397371474e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13642
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.53125,
      "epoch": 0.659209509083881,
      "grad_norm": 0.0,
      "learning_rate": 1.7041940471588712e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13643
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.71875,
      "epoch": 0.6592578275995361,
      "grad_norm": 0.0,
      "learning_rate": 1.7039524545805954e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13644
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.0625,
      "epoch": 0.6593061461151913,
      "grad_norm": 0.0,
      "learning_rate": 1.7037108620023192e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13645
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.71875,
      "epoch": 0.6593544646308466,
      "grad_norm": 0.0,
      "learning_rate": 1.703469269424043e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13646
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 43.6875,
      "epoch": 0.6594027831465017,
      "grad_norm": 0.0,
      "learning_rate": 1.7032276768457672e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13647
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.53125,
      "epoch": 0.6594511016621569,
      "grad_norm": 0.0,
      "learning_rate": 1.702986084267491e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13648
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.4375,
      "epoch": 0.6594994201778122,
      "grad_norm": 0.0,
      "learning_rate": 1.7027444916892152e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13649
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 54.125,
      "epoch": 0.6595477386934674,
      "grad_norm": 0.0,
      "learning_rate": 1.702502899110939e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 13650
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.5,
      "epoch": 0.6595960572091225,
      "grad_norm": 0.0,
      "learning_rate": 1.7022613065326633e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13651
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.4375,
      "epoch": 0.6596443757247777,
      "grad_norm": 0.0,
      "learning_rate": 1.7020197139543874e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13652
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.0625,
      "epoch": 0.659692694240433,
      "grad_norm": 0.0,
      "learning_rate": 1.7017781213761113e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13653
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 54.6875,
      "epoch": 0.6597410127560881,
      "grad_norm": 33.21116961267598,
      "learning_rate": 1.7015365287978354e-07,
      "loss": 0.0,
      "reward": 2.1875,
      "reward_std": 0.24359199404716492,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13654
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.96875,
      "epoch": 0.6597893312717433,
      "grad_norm": 0.0,
      "learning_rate": 1.7012949362195593e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13655
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.65625,
      "epoch": 0.6598376497873986,
      "grad_norm": 0.0,
      "learning_rate": 1.7010533436412834e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13656
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.3125,
      "epoch": 0.6598859683030537,
      "grad_norm": 0.0,
      "learning_rate": 1.7008117510630073e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13657
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.96875,
      "epoch": 0.6599342868187089,
      "grad_norm": 0.0,
      "learning_rate": 1.7005701584847312e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13658
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.28125,
      "epoch": 0.6599826053343641,
      "grad_norm": 0.0,
      "learning_rate": 1.7003285659064553e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13659
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.65625,
      "epoch": 0.6600309238500194,
      "grad_norm": 0.0,
      "learning_rate": 1.7000869733281792e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13660
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 59.53125,
      "epoch": 0.6600792423656745,
      "grad_norm": 70.82040888854553,
      "learning_rate": 1.6998453807499033e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.2314550280570984,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13661
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.0625,
      "epoch": 0.6601275608813297,
      "grad_norm": 0.0,
      "learning_rate": 1.6996037881716274e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13662
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 56.4375,
      "epoch": 0.660175879396985,
      "grad_norm": 0.0,
      "learning_rate": 1.6993621955933513e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13663
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.875,
      "epoch": 0.6602241979126401,
      "grad_norm": 0.0,
      "learning_rate": 1.6991206030150754e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13664
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 57.9375,
      "epoch": 0.6602725164282953,
      "grad_norm": 78.79554098055613,
      "learning_rate": 1.6988790104367993e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.35711580514907837,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13665
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 50.90625,
      "epoch": 0.6603208349439506,
      "grad_norm": 0.0,
      "learning_rate": 1.6986374178585234e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 13666
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.78125,
      "epoch": 0.6603691534596057,
      "grad_norm": 0.0,
      "learning_rate": 1.6983958252802473e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13667
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.21875,
      "epoch": 0.6604174719752609,
      "grad_norm": 0.0,
      "learning_rate": 1.6981542327019712e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13668
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.3125,
      "epoch": 0.6604657904909161,
      "grad_norm": 0.0,
      "learning_rate": 1.6979126401236953e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13669
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.09375,
      "epoch": 0.6605141090065714,
      "grad_norm": 0.0,
      "learning_rate": 1.6976710475454192e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13670
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.90625,
      "epoch": 0.6605624275222265,
      "grad_norm": 0.0,
      "learning_rate": 1.6974294549671433e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13671
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.03125,
      "epoch": 0.6606107460378817,
      "grad_norm": 0.0,
      "learning_rate": 1.6971878623888675e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13672
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.59375,
      "epoch": 0.660659064553537,
      "grad_norm": 0.0,
      "learning_rate": 1.6969462698105913e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13673
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.59375,
      "epoch": 0.6607073830691921,
      "grad_norm": 0.0,
      "learning_rate": 1.6967046772323155e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13674
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.375,
      "epoch": 0.6607557015848473,
      "grad_norm": 0.0,
      "learning_rate": 1.6964630846540394e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13675
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.4375,
      "epoch": 0.6608040201005025,
      "grad_norm": 0.0,
      "learning_rate": 1.6962214920757635e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13676
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.75,
      "epoch": 0.6608523386161577,
      "grad_norm": 0.0,
      "learning_rate": 1.6959798994974874e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13677
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.84375,
      "epoch": 0.6609006571318129,
      "grad_norm": 0.0,
      "learning_rate": 1.6957383069192112e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13678
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.84375,
      "epoch": 0.6609489756474681,
      "grad_norm": 0.0,
      "learning_rate": 1.6954967143409354e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13679
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 61.46875,
      "epoch": 0.6609972941631234,
      "grad_norm": 19.77934297191188,
      "learning_rate": 1.6952551217626592e-07,
      "loss": 0.0,
      "reward": 1.390625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 13680
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.21875,
      "epoch": 0.6610456126787785,
      "grad_norm": 0.0,
      "learning_rate": 1.6950135291843834e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13681
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 64.15625,
      "epoch": 0.6610939311944337,
      "grad_norm": 0.0,
      "learning_rate": 1.6947719366061075e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 13682
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.59375,
      "epoch": 0.661142249710089,
      "grad_norm": 0.0,
      "learning_rate": 1.6945303440278314e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13683
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.0625,
      "epoch": 0.6611905682257441,
      "grad_norm": 0.0,
      "learning_rate": 1.6942887514495555e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 13684
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 62.5,
      "epoch": 0.6612388867413993,
      "grad_norm": 0.0,
      "learning_rate": 1.6940471588712794e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 13685
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.59375,
      "epoch": 0.6612872052570545,
      "grad_norm": 0.0,
      "learning_rate": 1.6938055662930035e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13686
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 54.0625,
      "epoch": 0.6613355237727097,
      "grad_norm": 0.0,
      "learning_rate": 1.6935639737147277e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 13687
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.21875,
      "epoch": 0.6613838422883649,
      "grad_norm": 0.0,
      "learning_rate": 1.6933223811364513e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13688
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.34375,
      "epoch": 0.6614321608040201,
      "grad_norm": 0.0,
      "learning_rate": 1.6930807885581754e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13689
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.46875,
      "epoch": 0.6614804793196754,
      "grad_norm": 0.0,
      "learning_rate": 1.6928391959798993e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 13690
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.125,
      "epoch": 0.6615287978353305,
      "grad_norm": 0.0,
      "learning_rate": 1.6925976034016234e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13691
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.09375,
      "epoch": 0.6615771163509857,
      "grad_norm": 0.0,
      "learning_rate": 1.6923560108233476e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13692
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.5,
      "epoch": 0.6616254348666409,
      "grad_norm": 0.0,
      "learning_rate": 1.6921144182450714e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13693
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.53125,
      "epoch": 0.6616737533822961,
      "grad_norm": 0.0,
      "learning_rate": 1.6918728256667956e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13694
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.4375,
      "epoch": 0.6617220718979513,
      "grad_norm": 0.0,
      "learning_rate": 1.6916312330885194e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13695
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.21875,
      "epoch": 0.6617703904136065,
      "grad_norm": 0.0,
      "learning_rate": 1.6913896405102436e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13696
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.0,
      "epoch": 0.6618187089292616,
      "grad_norm": 0.0,
      "learning_rate": 1.6911480479319677e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13697
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.84375,
      "epoch": 0.6618670274449169,
      "grad_norm": 0.0,
      "learning_rate": 1.6909064553536913e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13698
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.125,
      "epoch": 0.6619153459605721,
      "grad_norm": 0.0,
      "learning_rate": 1.6906648627754154e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13699
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.78125,
      "epoch": 0.6619636644762273,
      "grad_norm": 0.0,
      "learning_rate": 1.6904232701971393e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13700
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.34375,
      "epoch": 0.6620119829918825,
      "grad_norm": 0.0,
      "learning_rate": 1.6901816776188635e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13701
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.15625,
      "epoch": 0.6620603015075377,
      "grad_norm": 0.0,
      "learning_rate": 1.6899400850405876e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13702
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.09375,
      "epoch": 0.6621086200231929,
      "grad_norm": 0.0,
      "learning_rate": 1.6896984924623115e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13703
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.09375,
      "epoch": 0.662156938538848,
      "grad_norm": 0.0,
      "learning_rate": 1.6894568998840356e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13704
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.21875,
      "epoch": 0.6622052570545033,
      "grad_norm": 0.0,
      "learning_rate": 1.6892153073057595e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13705
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.59375,
      "epoch": 0.6622535755701585,
      "grad_norm": 0.0,
      "learning_rate": 1.6889737147274836e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13706
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.8125,
      "epoch": 0.6623018940858136,
      "grad_norm": 0.0,
      "learning_rate": 1.6887321221492077e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13707
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.3125,
      "epoch": 0.6623502126014689,
      "grad_norm": 0.0,
      "learning_rate": 1.6884905295709313e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13708
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.78125,
      "epoch": 0.6623985311171241,
      "grad_norm": 0.0,
      "learning_rate": 1.6882489369926555e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13709
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.3125,
      "epoch": 0.6624468496327793,
      "grad_norm": 0.0,
      "learning_rate": 1.6880073444143794e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13710
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 58.875,
      "epoch": 0.6624951681484345,
      "grad_norm": 5.163359155413555,
      "learning_rate": 1.6877657518361035e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13711
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.125,
      "epoch": 0.6625434866640897,
      "grad_norm": 0.0,
      "learning_rate": 1.6875241592578276e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13712
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.8125,
      "epoch": 0.6625918051797449,
      "grad_norm": 0.0,
      "learning_rate": 1.6872825666795515e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13713
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 63.34375,
      "epoch": 0.6626401236954,
      "grad_norm": 0.0,
      "learning_rate": 1.6870409741012756e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13714
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.3125,
      "epoch": 0.6626884422110553,
      "grad_norm": 0.0,
      "learning_rate": 1.6867993815229995e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13715
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 60.34375,
      "epoch": 0.6627367607267105,
      "grad_norm": 0.0,
      "learning_rate": 1.6865577889447236e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 13716
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.6875,
      "epoch": 0.6627850792423656,
      "grad_norm": 0.0,
      "learning_rate": 1.6863161963664478e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13717
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.9375,
      "epoch": 0.6628333977580209,
      "grad_norm": 0.0,
      "learning_rate": 1.6860746037881714e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13718
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.0,
      "epoch": 0.6628817162736761,
      "grad_norm": 0.0,
      "learning_rate": 1.6858330112098955e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13719
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.3125,
      "epoch": 0.6629300347893313,
      "grad_norm": 0.0,
      "learning_rate": 1.6855914186316194e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13720
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.53125,
      "epoch": 0.6629783533049864,
      "grad_norm": 0.0,
      "learning_rate": 1.6853498260533435e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13721
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.5,
      "epoch": 0.6630266718206417,
      "grad_norm": 0.0,
      "learning_rate": 1.6851082334750677e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13722
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.625,
      "epoch": 0.6630749903362969,
      "grad_norm": 0.0,
      "learning_rate": 1.6848666408967915e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13723
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.4375,
      "epoch": 0.663123308851952,
      "grad_norm": 0.0,
      "learning_rate": 1.6846250483185157e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13724
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.3125,
      "epoch": 0.6631716273676073,
      "grad_norm": 0.0,
      "learning_rate": 1.6843834557402395e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13725
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.125,
      "epoch": 0.6632199458832625,
      "grad_norm": 0.0,
      "learning_rate": 1.6841418631619637e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13726
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.46875,
      "epoch": 0.6632682643989176,
      "grad_norm": 0.0,
      "learning_rate": 1.6839002705836878e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13727
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 59.90625,
      "epoch": 0.6633165829145728,
      "grad_norm": 0.0,
      "learning_rate": 1.6836586780054114e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13728
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.90625,
      "epoch": 0.6633649014302281,
      "grad_norm": 0.0,
      "learning_rate": 1.6834170854271356e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13729
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.96875,
      "epoch": 0.6634132199458833,
      "grad_norm": 0.0,
      "learning_rate": 1.6831754928488594e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13730
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.0,
      "epoch": 0.6634615384615384,
      "grad_norm": 0.0,
      "learning_rate": 1.6829339002705836e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13731
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.1875,
      "epoch": 0.6635098569771937,
      "grad_norm": 0.0,
      "learning_rate": 1.6826923076923077e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13732
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.15625,
      "epoch": 0.6635581754928489,
      "grad_norm": 0.0,
      "learning_rate": 1.6824507151140316e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13733
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.03125,
      "epoch": 0.663606494008504,
      "grad_norm": 0.0,
      "learning_rate": 1.6822091225357557e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13734
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.375,
      "epoch": 0.6636548125241593,
      "grad_norm": 0.0,
      "learning_rate": 1.6819675299574796e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13735
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.4375,
      "epoch": 0.6637031310398145,
      "grad_norm": 0.0,
      "learning_rate": 1.6817259373792037e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13736
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.3125,
      "epoch": 0.6637514495554696,
      "grad_norm": 0.0,
      "learning_rate": 1.6814843448009279e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13737
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.8125,
      "epoch": 0.6637997680711248,
      "grad_norm": 0.0,
      "learning_rate": 1.6812427522226517e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13738
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.0625,
      "epoch": 0.6638480865867801,
      "grad_norm": 0.0,
      "learning_rate": 1.6810011596443756e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13739
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.53125,
      "epoch": 0.6638964051024353,
      "grad_norm": 0.0,
      "learning_rate": 1.6807595670660995e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13740
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.6875,
      "epoch": 0.6639447236180904,
      "grad_norm": 0.0,
      "learning_rate": 1.6805179744878236e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13741
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.625,
      "epoch": 0.6639930421337457,
      "grad_norm": 0.0,
      "learning_rate": 1.6802763819095477e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13742
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.0625,
      "epoch": 0.6640413606494009,
      "grad_norm": 0.0,
      "learning_rate": 1.6800347893312716e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13743
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.03125,
      "epoch": 0.664089679165056,
      "grad_norm": 0.0,
      "learning_rate": 1.6797931967529958e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13744
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.65625,
      "epoch": 0.6641379976807112,
      "grad_norm": 0.0,
      "learning_rate": 1.6795516041747196e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13745
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.25,
      "epoch": 0.6641863161963665,
      "grad_norm": 0.0,
      "learning_rate": 1.6793100115964438e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13746
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.875,
      "epoch": 0.6642346347120216,
      "grad_norm": 0.0,
      "learning_rate": 1.679068419018168e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13747
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.59375,
      "epoch": 0.6642829532276768,
      "grad_norm": 0.0,
      "learning_rate": 1.6788268264398918e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13748
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 54.9375,
      "epoch": 0.6643312717433321,
      "grad_norm": 13.085057487607891,
      "learning_rate": 1.6785852338616156e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13749
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.375,
      "epoch": 0.6643795902589873,
      "grad_norm": 0.0,
      "learning_rate": 1.6783436412833395e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13750
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.5625,
      "epoch": 0.6644279087746424,
      "grad_norm": 0.0,
      "learning_rate": 1.6781020487050637e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13751
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.9375,
      "epoch": 0.6644762272902976,
      "grad_norm": 0.0,
      "learning_rate": 1.6778604561267878e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13752
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.1875,
      "epoch": 0.6645245458059529,
      "grad_norm": 0.0,
      "learning_rate": 1.6776188635485117e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13753
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.59375,
      "epoch": 0.664572864321608,
      "grad_norm": 0.0,
      "learning_rate": 1.6773772709702358e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13754
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.125,
      "epoch": 0.6646211828372632,
      "grad_norm": 0.0,
      "learning_rate": 1.6771356783919597e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13755
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.53125,
      "epoch": 0.6646695013529185,
      "grad_norm": 0.0,
      "learning_rate": 1.6768940858136838e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13756
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 63.59375,
      "epoch": 0.6647178198685736,
      "grad_norm": 11.248558115492669,
      "learning_rate": 1.676652493235408e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.26409149169921875,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13757
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.28125,
      "epoch": 0.6647661383842288,
      "grad_norm": 0.0,
      "learning_rate": 1.6764109006571318e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 13758
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.0,
      "epoch": 0.664814456899884,
      "grad_norm": 0.0,
      "learning_rate": 1.676169308078856e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13759
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.5625,
      "epoch": 0.6648627754155393,
      "grad_norm": 0.0,
      "learning_rate": 1.6759277155005796e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13760
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 50.78125,
      "epoch": 0.6649110939311944,
      "grad_norm": 0.0,
      "learning_rate": 1.6756861229223037e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13761
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.6875,
      "epoch": 0.6649594124468496,
      "grad_norm": 0.0,
      "learning_rate": 1.6754445303440278e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13762
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.125,
      "epoch": 0.6650077309625049,
      "grad_norm": 0.0,
      "learning_rate": 1.6752029377657517e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13763
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 45.84375,
      "epoch": 0.66505604947816,
      "grad_norm": 0.0,
      "learning_rate": 1.6749613451874758e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13764
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.0625,
      "epoch": 0.6651043679938152,
      "grad_norm": 0.0,
      "learning_rate": 1.6747197526091997e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13765
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 48.0625,
      "epoch": 0.6651526865094705,
      "grad_norm": 0.0,
      "learning_rate": 1.6744781600309238e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13766
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.4375,
      "epoch": 0.6652010050251256,
      "grad_norm": 0.0,
      "learning_rate": 1.674236567452648e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13767
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.6875,
      "epoch": 0.6652493235407808,
      "grad_norm": 0.0,
      "learning_rate": 1.6739949748743718e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13768
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.78125,
      "epoch": 0.665297642056436,
      "grad_norm": 0.0,
      "learning_rate": 1.673753382296096e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13769
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.53125,
      "epoch": 0.6653459605720913,
      "grad_norm": 0.0,
      "learning_rate": 1.6735117897178196e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13770
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.875,
      "epoch": 0.6653942790877464,
      "grad_norm": 0.0,
      "learning_rate": 1.6732701971395437e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13771
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.0,
      "epoch": 0.6654425976034016,
      "grad_norm": 0.0,
      "learning_rate": 1.6730286045612679e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13772
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 53.71875,
      "epoch": 0.6654909161190569,
      "grad_norm": 0.0,
      "learning_rate": 1.6727870119829917e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 13773
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.96875,
      "epoch": 0.665539234634712,
      "grad_norm": 0.0,
      "learning_rate": 1.672545419404716e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13774
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.28125,
      "epoch": 0.6655875531503672,
      "grad_norm": 0.0,
      "learning_rate": 1.6723038268264397e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13775
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.21875,
      "epoch": 0.6656358716660224,
      "grad_norm": 0.0,
      "learning_rate": 1.672062234248164e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13776
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.90625,
      "epoch": 0.6656841901816776,
      "grad_norm": 0.0,
      "learning_rate": 1.671820641669888e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13777
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.1875,
      "epoch": 0.6657325086973328,
      "grad_norm": 0.0,
      "learning_rate": 1.671579049091612e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13778
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.28125,
      "epoch": 0.665780827212988,
      "grad_norm": 0.0,
      "learning_rate": 1.671337456513336e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13779
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.46875,
      "epoch": 0.6658291457286433,
      "grad_norm": 0.0,
      "learning_rate": 1.6710958639350596e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13780
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 62.6875,
      "epoch": 0.6658774642442984,
      "grad_norm": 0.0,
      "learning_rate": 1.6708542713567838e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13781
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.8125,
      "epoch": 0.6659257827599536,
      "grad_norm": 0.0,
      "learning_rate": 1.670612678778508e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13782
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.0,
      "epoch": 0.6659741012756089,
      "grad_norm": 0.0,
      "learning_rate": 1.6703710862002318e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13783
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.46875,
      "epoch": 0.666022419791264,
      "grad_norm": 0.0,
      "learning_rate": 1.670129493621956e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13784
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.15625,
      "epoch": 0.6660707383069192,
      "grad_norm": 0.0,
      "learning_rate": 1.6698879010436798e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13785
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.65625,
      "epoch": 0.6661190568225744,
      "grad_norm": 0.0,
      "learning_rate": 1.669646308465404e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13786
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.4375,
      "epoch": 0.6661673753382296,
      "grad_norm": 0.0,
      "learning_rate": 1.669404715887128e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13787
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 63.6875,
      "epoch": 0.6662156938538848,
      "grad_norm": 0.0,
      "learning_rate": 1.669163123308852e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13788
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.6875,
      "epoch": 0.66626401236954,
      "grad_norm": 0.0,
      "learning_rate": 1.668921530730576e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13789
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.1875,
      "epoch": 0.6663123308851953,
      "grad_norm": 0.0,
      "learning_rate": 1.6686799381522997e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13790
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.59375,
      "epoch": 0.6663606494008504,
      "grad_norm": 0.0,
      "learning_rate": 1.6684383455740238e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13791
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.5625,
      "epoch": 0.6664089679165056,
      "grad_norm": 0.0,
      "learning_rate": 1.668196752995748e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13792
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 48.75,
      "epoch": 0.6664572864321608,
      "grad_norm": 0.0,
      "learning_rate": 1.6679551604174718e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13793
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.6875,
      "epoch": 0.666505604947816,
      "grad_norm": 0.0,
      "learning_rate": 1.667713567839196e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13794
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 40.0,
      "epoch": 0.6665539234634712,
      "grad_norm": 0.0,
      "learning_rate": 1.6674719752609198e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13795
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.90625,
      "epoch": 0.6666022419791264,
      "grad_norm": 0.0,
      "learning_rate": 1.667230382682644e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13796
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.03125,
      "epoch": 0.6666505604947816,
      "grad_norm": 0.0,
      "learning_rate": 1.666988790104368e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13797
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 60.96875,
      "epoch": 0.6666988790104368,
      "grad_norm": 0.0,
      "learning_rate": 1.666747197526092e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 13798
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.53125,
      "epoch": 0.666747197526092,
      "grad_norm": 0.0,
      "learning_rate": 1.666505604947816e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13799
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.34375,
      "epoch": 0.6667955160417472,
      "grad_norm": 0.0,
      "learning_rate": 1.6662640123695397e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13800
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 52.96875,
      "epoch": 0.6668438345574024,
      "grad_norm": 76.45923651728513,
      "learning_rate": 1.6660224197912638e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.18600594997406006,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13801
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 51.96875,
      "epoch": 0.6668921530730576,
      "grad_norm": 0.0,
      "learning_rate": 1.665780827212988e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13802
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.3125,
      "epoch": 0.6669404715887128,
      "grad_norm": 0.0,
      "learning_rate": 1.6655392346347119e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13803
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.5625,
      "epoch": 0.666988790104368,
      "grad_norm": 0.0,
      "learning_rate": 1.665297642056436e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13804
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.4375,
      "epoch": 0.6670371086200232,
      "grad_norm": 0.0,
      "learning_rate": 1.6650560494781599e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13805
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.15625,
      "epoch": 0.6670854271356784,
      "grad_norm": 0.0,
      "learning_rate": 1.664814456899884e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13806
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.71875,
      "epoch": 0.6671337456513335,
      "grad_norm": 0.0,
      "learning_rate": 1.6645728643216081e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13807
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.4375,
      "epoch": 0.6671820641669888,
      "grad_norm": 0.0,
      "learning_rate": 1.664331271743332e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13808
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.59375,
      "epoch": 0.667230382682644,
      "grad_norm": 0.0,
      "learning_rate": 1.6640896791650561e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13809
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 58.125,
      "epoch": 0.6672787011982992,
      "grad_norm": 16.821283493499426,
      "learning_rate": 1.66384808658678e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.2505248188972473,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13810
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.5625,
      "epoch": 0.6673270197139544,
      "grad_norm": 0.0,
      "learning_rate": 1.663606494008504e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13811
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.28125,
      "epoch": 0.6673753382296096,
      "grad_norm": 0.0,
      "learning_rate": 1.663364901430228e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13812
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 58.375,
      "epoch": 0.6674236567452648,
      "grad_norm": 0.0,
      "learning_rate": 1.663123308851952e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 13813
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 65.6875,
      "epoch": 0.66747197526092,
      "grad_norm": 32.63207131728328,
      "learning_rate": 1.662881716273676e-07,
      "loss": -0.0,
      "reward": 1.5,
      "reward_std": 0.23827512562274933,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13814
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 46.46875,
      "epoch": 0.6675202937765752,
      "grad_norm": 0.0,
      "learning_rate": 1.6626401236954e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 13815
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.03125,
      "epoch": 0.6675686122922304,
      "grad_norm": 0.0,
      "learning_rate": 1.662398531117124e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 13816
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.46875,
      "epoch": 0.6676169308078855,
      "grad_norm": 0.0,
      "learning_rate": 1.6621569385388482e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13817
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.90625,
      "epoch": 0.6676652493235408,
      "grad_norm": 0.0,
      "learning_rate": 1.661915345960572e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13818
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.3125,
      "epoch": 0.667713567839196,
      "grad_norm": 0.0,
      "learning_rate": 1.6616737533822962e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13819
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.03125,
      "epoch": 0.6677618863548512,
      "grad_norm": 0.0,
      "learning_rate": 1.66143216080402e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2583096921443939,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13820
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.5625,
      "epoch": 0.6678102048705064,
      "grad_norm": 0.0,
      "learning_rate": 1.661190568225744e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13821
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.65625,
      "epoch": 0.6678585233861616,
      "grad_norm": 0.0,
      "learning_rate": 1.660948975647468e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 13822
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.5625,
      "epoch": 0.6679068419018168,
      "grad_norm": 0.0,
      "learning_rate": 1.660707383069192e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13823
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.03125,
      "epoch": 0.6679551604174719,
      "grad_norm": 0.0,
      "learning_rate": 1.660465790490916e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13824
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.84375,
      "epoch": 0.6680034789331272,
      "grad_norm": 0.0,
      "learning_rate": 1.66022419791264e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13825
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.40625,
      "epoch": 0.6680517974487824,
      "grad_norm": 0.0,
      "learning_rate": 1.659982605334364e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13826
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 47.875,
      "epoch": 0.6681001159644375,
      "grad_norm": 0.0,
      "learning_rate": 1.6597410127560882e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13827
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.09375,
      "epoch": 0.6681484344800928,
      "grad_norm": 0.0,
      "learning_rate": 1.659499420177812e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13828
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.8125,
      "epoch": 0.668196752995748,
      "grad_norm": 0.0,
      "learning_rate": 1.6592578275995362e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13829
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 56.875,
      "epoch": 0.6682450715114032,
      "grad_norm": 0.0,
      "learning_rate": 1.65901623502126e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 13830
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 47.9375,
      "epoch": 0.6682933900270583,
      "grad_norm": 0.0,
      "learning_rate": 1.658774642442984e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13831
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 56.625,
      "epoch": 0.6683417085427136,
      "grad_norm": 0.0,
      "learning_rate": 1.658533049864708e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13832
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.90625,
      "epoch": 0.6683900270583688,
      "grad_norm": 0.0,
      "learning_rate": 1.658291457286432e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13833
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 56.65625,
      "epoch": 0.6684383455740239,
      "grad_norm": 0.0,
      "learning_rate": 1.658049864708156e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13834
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.15625,
      "epoch": 0.6684866640896792,
      "grad_norm": 0.0,
      "learning_rate": 1.65780827212988e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13835
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.28125,
      "epoch": 0.6685349826053344,
      "grad_norm": 0.0,
      "learning_rate": 1.657566679551604e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13836
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.96875,
      "epoch": 0.6685833011209895,
      "grad_norm": 0.0,
      "learning_rate": 1.6573250869733283e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13837
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.625,
      "epoch": 0.6686316196366447,
      "grad_norm": 0.0,
      "learning_rate": 1.657083494395052e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13838
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 62.28125,
      "epoch": 0.6686799381523,
      "grad_norm": 0.0,
      "learning_rate": 1.6568419018167763e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13839
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.46875,
      "epoch": 0.6687282566679552,
      "grad_norm": 0.0,
      "learning_rate": 1.6566003092385001e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13840
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.25,
      "epoch": 0.6687765751836103,
      "grad_norm": 0.0,
      "learning_rate": 1.6563587166602243e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 13841
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 49.3125,
      "epoch": 0.6688248936992656,
      "grad_norm": 0.0,
      "learning_rate": 1.6561171240819481e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13842
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.6875,
      "epoch": 0.6688732122149208,
      "grad_norm": 0.0,
      "learning_rate": 1.655875531503672e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13843
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.21875,
      "epoch": 0.6689215307305759,
      "grad_norm": 0.0,
      "learning_rate": 1.6556339389253961e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13844
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.65625,
      "epoch": 0.6689698492462312,
      "grad_norm": 0.0,
      "learning_rate": 1.65539234634712e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 13845
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 60.5625,
      "epoch": 0.6690181677618864,
      "grad_norm": 0.0,
      "learning_rate": 1.6551507537688442e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13846
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.9375,
      "epoch": 0.6690664862775415,
      "grad_norm": 0.0,
      "learning_rate": 1.6549091611905683e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13847
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.125,
      "epoch": 0.6691148047931967,
      "grad_norm": 0.0,
      "learning_rate": 1.6546675686122922e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13848
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.59375,
      "epoch": 0.669163123308852,
      "grad_norm": 0.0,
      "learning_rate": 1.6544259760340163e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13849
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.8125,
      "epoch": 0.6692114418245072,
      "grad_norm": 0.0,
      "learning_rate": 1.6541843834557402e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13850
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.6875,
      "epoch": 0.6692597603401623,
      "grad_norm": 0.0,
      "learning_rate": 1.6539427908774643e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13851
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.34375,
      "epoch": 0.6693080788558176,
      "grad_norm": 0.0,
      "learning_rate": 1.6537011982991882e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13852
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.34375,
      "epoch": 0.6693563973714728,
      "grad_norm": 0.0,
      "learning_rate": 1.653459605720912e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13853
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 58.21875,
      "epoch": 0.6694047158871279,
      "grad_norm": 0.0,
      "learning_rate": 1.6532180131426362e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13854
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.5,
      "epoch": 0.6694530344027831,
      "grad_norm": 0.0,
      "learning_rate": 1.65297642056436e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13855
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.3125,
      "epoch": 0.6695013529184384,
      "grad_norm": 0.0,
      "learning_rate": 1.6527348279860842e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13856
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.125,
      "epoch": 0.6695496714340935,
      "grad_norm": 0.0,
      "learning_rate": 1.6524932354078083e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13857
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.3125,
      "epoch": 0.6695979899497487,
      "grad_norm": 0.0,
      "learning_rate": 1.6522516428295322e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13858
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.84375,
      "epoch": 0.669646308465404,
      "grad_norm": 0.0,
      "learning_rate": 1.6520100502512563e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13859
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 60.09375,
      "epoch": 0.6696946269810592,
      "grad_norm": 0.0,
      "learning_rate": 1.6517684576729802e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13860
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.3125,
      "epoch": 0.6697429454967143,
      "grad_norm": 0.0,
      "learning_rate": 1.6515268650947043e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13861
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.75,
      "epoch": 0.6697912640123695,
      "grad_norm": 0.0,
      "learning_rate": 1.6512852725164285e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13862
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.5625,
      "epoch": 0.6698395825280248,
      "grad_norm": 0.0,
      "learning_rate": 1.651043679938152e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13863
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.6875,
      "epoch": 0.6698879010436799,
      "grad_norm": 0.0,
      "learning_rate": 1.6508020873598762e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13864
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.4375,
      "epoch": 0.6699362195593351,
      "grad_norm": 0.0,
      "learning_rate": 1.6505604947816e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13865
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.46875,
      "epoch": 0.6699845380749904,
      "grad_norm": 0.0,
      "learning_rate": 1.6503189022033242e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13866
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.0,
      "epoch": 0.6700328565906455,
      "grad_norm": 0.0,
      "learning_rate": 1.6500773096250484e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13867
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.4375,
      "epoch": 0.6700811751063007,
      "grad_norm": 0.0,
      "learning_rate": 1.6498357170467722e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13868
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.15625,
      "epoch": 0.670129493621956,
      "grad_norm": 0.0,
      "learning_rate": 1.6495941244684964e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13869
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.59375,
      "epoch": 0.6701778121376112,
      "grad_norm": 0.0,
      "learning_rate": 1.6493525318902203e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13870
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.125,
      "epoch": 0.6702261306532663,
      "grad_norm": 0.0,
      "learning_rate": 1.6491109393119444e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13871
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.1875,
      "epoch": 0.6702744491689215,
      "grad_norm": 0.0,
      "learning_rate": 1.6488693467336685e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13872
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.21875,
      "epoch": 0.6703227676845768,
      "grad_norm": 0.0,
      "learning_rate": 1.648627754155392e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13873
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 52.34375,
      "epoch": 0.6703710862002319,
      "grad_norm": 0.0,
      "learning_rate": 1.6483861615771163e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13874
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.03125,
      "epoch": 0.6704194047158871,
      "grad_norm": 0.0,
      "learning_rate": 1.6481445689988401e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13875
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 50.9375,
      "epoch": 0.6704677232315424,
      "grad_norm": 0.0,
      "learning_rate": 1.6479029764205643e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13876
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.59375,
      "epoch": 0.6705160417471975,
      "grad_norm": 0.0,
      "learning_rate": 1.6476613838422884e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13877
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.125,
      "epoch": 0.6705643602628527,
      "grad_norm": 0.0,
      "learning_rate": 1.6474197912640123e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13878
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.9375,
      "epoch": 0.6706126787785079,
      "grad_norm": 0.0,
      "learning_rate": 1.6471781986857364e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13879
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.25,
      "epoch": 0.6706609972941632,
      "grad_norm": 0.0,
      "learning_rate": 1.6469366061074603e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.2041158676147461,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13880
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.78125,
      "epoch": 0.6707093158098183,
      "grad_norm": 0.0,
      "learning_rate": 1.6466950135291844e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13881
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.75,
      "epoch": 0.6707576343254735,
      "grad_norm": 0.0,
      "learning_rate": 1.6464534209509086e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13882
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.71875,
      "epoch": 0.6708059528411288,
      "grad_norm": 0.0,
      "learning_rate": 1.6462118283726322e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13883
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.53125,
      "epoch": 0.6708542713567839,
      "grad_norm": 0.0,
      "learning_rate": 1.6459702357943563e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13884
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.8125,
      "epoch": 0.6709025898724391,
      "grad_norm": 0.0,
      "learning_rate": 1.6457286432160802e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13885
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.4375,
      "epoch": 0.6709509083880943,
      "grad_norm": 0.0,
      "learning_rate": 1.6454870506378043e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13886
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.71875,
      "epoch": 0.6709992269037495,
      "grad_norm": 0.0,
      "learning_rate": 1.6452454580595284e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13887
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.625,
      "epoch": 0.6710475454194047,
      "grad_norm": 0.0,
      "learning_rate": 1.6450038654812523e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13888
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.84375,
      "epoch": 0.6710958639350599,
      "grad_norm": 0.0,
      "learning_rate": 1.6447622729029765e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13889
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 56.15625,
      "epoch": 0.6711441824507152,
      "grad_norm": 0.0,
      "learning_rate": 1.6445206803247003e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13890
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 50.59375,
      "epoch": 0.6711925009663703,
      "grad_norm": 0.0,
      "learning_rate": 1.6442790877464245e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13891
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 58.53125,
      "epoch": 0.6712408194820255,
      "grad_norm": 0.0,
      "learning_rate": 1.6440374951681486e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13892
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 53.09375,
      "epoch": 0.6712891379976808,
      "grad_norm": 0.0,
      "learning_rate": 1.6437959025898722e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13893
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.375,
      "epoch": 0.6713374565133359,
      "grad_norm": 0.0,
      "learning_rate": 1.6435543100115963e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13894
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.21875,
      "epoch": 0.6713857750289911,
      "grad_norm": 0.0,
      "learning_rate": 1.6433127174333202e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13895
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 56.0625,
      "epoch": 0.6714340935446463,
      "grad_norm": 0.0,
      "learning_rate": 1.6430711248550444e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13896
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.40625,
      "epoch": 0.6714824120603015,
      "grad_norm": 0.0,
      "learning_rate": 1.6428295322767685e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13897
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.4375,
      "epoch": 0.6715307305759567,
      "grad_norm": 0.0,
      "learning_rate": 1.6425879396984924e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13898
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.9375,
      "epoch": 0.6715790490916119,
      "grad_norm": 0.0,
      "learning_rate": 1.6423463471202165e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13899
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.625,
      "epoch": 0.6716273676072672,
      "grad_norm": 0.0,
      "learning_rate": 1.6421047545419404e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 13900
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.75,
      "epoch": 0.6716756861229223,
      "grad_norm": 0.0,
      "learning_rate": 1.6418631619636645e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13901
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.4375,
      "epoch": 0.6717240046385775,
      "grad_norm": 0.0,
      "learning_rate": 1.6416215693853886e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13902
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.53125,
      "epoch": 0.6717723231542327,
      "grad_norm": 0.0,
      "learning_rate": 1.6413799768071122e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13903
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.5,
      "epoch": 0.6718206416698879,
      "grad_norm": 0.0,
      "learning_rate": 1.6411383842288364e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13904
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.46875,
      "epoch": 0.6718689601855431,
      "grad_norm": 0.0,
      "learning_rate": 1.6408967916505603e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13905
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.8125,
      "epoch": 0.6719172787011983,
      "grad_norm": 0.0,
      "learning_rate": 1.6406551990722844e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13906
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.84375,
      "epoch": 0.6719655972168534,
      "grad_norm": 0.0,
      "learning_rate": 1.6404136064940085e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13907
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 53.6875,
      "epoch": 0.6720139157325087,
      "grad_norm": 0.0,
      "learning_rate": 1.6401720139157324e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13908
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.125,
      "epoch": 0.6720622342481639,
      "grad_norm": 0.0,
      "learning_rate": 1.6399304213374565e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13909
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.3125,
      "epoch": 0.6721105527638191,
      "grad_norm": 0.0,
      "learning_rate": 1.6396888287591804e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2630178928375244,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13910
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.53125,
      "epoch": 0.6721588712794743,
      "grad_norm": 0.0,
      "learning_rate": 1.6394472361809045e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13911
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.71875,
      "epoch": 0.6722071897951295,
      "grad_norm": 0.0,
      "learning_rate": 1.6392056436026287e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13912
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 49.0625,
      "epoch": 0.6722555083107847,
      "grad_norm": 0.0,
      "learning_rate": 1.6389640510243526e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13913
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.9375,
      "epoch": 0.6723038268264399,
      "grad_norm": 0.0,
      "learning_rate": 1.6387224584460764e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13914
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.75,
      "epoch": 0.6723521453420951,
      "grad_norm": 0.0,
      "learning_rate": 1.6384808658678003e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13915
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 57.65625,
      "epoch": 0.6724004638577503,
      "grad_norm": 0.0,
      "learning_rate": 1.6382392732895244e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13916
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.0,
      "epoch": 0.6724487823734054,
      "grad_norm": 0.0,
      "learning_rate": 1.6379976807112486e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13917
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.6875,
      "epoch": 0.6724971008890607,
      "grad_norm": 0.0,
      "learning_rate": 1.6377560881329724e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13918
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 52.71875,
      "epoch": 0.6725454194047159,
      "grad_norm": 0.0,
      "learning_rate": 1.6375144955546966e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13919
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 51.15625,
      "epoch": 0.6725937379203711,
      "grad_norm": 0.0,
      "learning_rate": 1.6372729029764204e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13920
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 54.5625,
      "epoch": 0.6726420564360263,
      "grad_norm": 0.0,
      "learning_rate": 1.6370313103981446e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13921
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.84375,
      "epoch": 0.6726903749516815,
      "grad_norm": 0.0,
      "learning_rate": 1.6367897178198687e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13922
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 61.9375,
      "epoch": 0.6727386934673367,
      "grad_norm": 0.0,
      "learning_rate": 1.6365481252415926e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13923
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 61.25,
      "epoch": 0.6727870119829918,
      "grad_norm": 7.6869918989672,
      "learning_rate": 1.6363065326633165e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13924
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.34375,
      "epoch": 0.6728353304986471,
      "grad_norm": 0.0,
      "learning_rate": 1.6360649400850403e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13925
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.0,
      "epoch": 0.6728836490143023,
      "grad_norm": 0.0,
      "learning_rate": 1.6358233475067645e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13926
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.90625,
      "epoch": 0.6729319675299574,
      "grad_norm": 0.0,
      "learning_rate": 1.6355817549284886e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13927
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.59375,
      "epoch": 0.6729802860456127,
      "grad_norm": 0.0,
      "learning_rate": 1.6353401623502125e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13928
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.1875,
      "epoch": 0.6730286045612679,
      "grad_norm": 0.0,
      "learning_rate": 1.6350985697719366e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13929
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.9375,
      "epoch": 0.6730769230769231,
      "grad_norm": 0.0,
      "learning_rate": 1.6348569771936605e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13930
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.09375,
      "epoch": 0.6731252415925782,
      "grad_norm": 0.0,
      "learning_rate": 1.6346153846153846e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13931
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.9375,
      "epoch": 0.6731735601082335,
      "grad_norm": 0.0,
      "learning_rate": 1.6343737920371088e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 13932
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 55.21875,
      "epoch": 0.6732218786238887,
      "grad_norm": 0.0,
      "learning_rate": 1.6341321994588326e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13933
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.1875,
      "epoch": 0.6732701971395438,
      "grad_norm": 0.0,
      "learning_rate": 1.6338906068805565e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13934
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.03125,
      "epoch": 0.6733185156551991,
      "grad_norm": 0.0,
      "learning_rate": 1.6336490143022804e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13935
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.59375,
      "epoch": 0.6733668341708543,
      "grad_norm": 0.0,
      "learning_rate": 1.6334074217240045e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13936
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.75,
      "epoch": 0.6734151526865094,
      "grad_norm": 0.0,
      "learning_rate": 1.6331658291457286e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13937
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 65.0,
      "epoch": 0.6734634712021647,
      "grad_norm": 0.0,
      "learning_rate": 1.6329242365674525e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 13938
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.25,
      "epoch": 0.6735117897178199,
      "grad_norm": 0.0,
      "learning_rate": 1.6326826439891767e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13939
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.53125,
      "epoch": 0.6735601082334751,
      "grad_norm": 0.0,
      "learning_rate": 1.6324410514109005e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13940
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.65625,
      "epoch": 0.6736084267491302,
      "grad_norm": 0.0,
      "learning_rate": 1.6321994588326247e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13941
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.375,
      "epoch": 0.6736567452647855,
      "grad_norm": 0.0,
      "learning_rate": 1.6319578662543488e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 13942
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.09375,
      "epoch": 0.6737050637804407,
      "grad_norm": 0.0,
      "learning_rate": 1.6317162736760727e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 13943
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.5,
      "epoch": 0.6737533822960958,
      "grad_norm": 0.0,
      "learning_rate": 1.6314746810977968e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13944
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.5625,
      "epoch": 0.6738017008117511,
      "grad_norm": 0.0,
      "learning_rate": 1.6312330885195204e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13945
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.65625,
      "epoch": 0.6738500193274063,
      "grad_norm": 0.0,
      "learning_rate": 1.6309914959412445e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13946
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 60.65625,
      "epoch": 0.6738983378430614,
      "grad_norm": 43.277947714732846,
      "learning_rate": 1.6307499033629687e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.3029687702655792,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13947
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.5625,
      "epoch": 0.6739466563587166,
      "grad_norm": 0.0,
      "learning_rate": 1.6305083107846926e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 13948
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.28125,
      "epoch": 0.6739949748743719,
      "grad_norm": 0.0,
      "learning_rate": 1.6302667182064167e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13949
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.46875,
      "epoch": 0.6740432933900271,
      "grad_norm": 0.0,
      "learning_rate": 1.6300251256281406e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13950
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.3125,
      "epoch": 0.6740916119056822,
      "grad_norm": 0.0,
      "learning_rate": 1.6297835330498647e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13951
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.375,
      "epoch": 0.6741399304213375,
      "grad_norm": 0.0,
      "learning_rate": 1.6295419404715888e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13952
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.09375,
      "epoch": 0.6741882489369927,
      "grad_norm": 0.0,
      "learning_rate": 1.6293003478933127e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 13953
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.0625,
      "epoch": 0.6742365674526478,
      "grad_norm": 0.0,
      "learning_rate": 1.6290587553150368e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13954
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.21875,
      "epoch": 0.674284885968303,
      "grad_norm": 0.0,
      "learning_rate": 1.6288171627367605e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 13955
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.8125,
      "epoch": 0.6743332044839583,
      "grad_norm": 0.0,
      "learning_rate": 1.6285755701584846e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13956
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 51.5625,
      "epoch": 0.6743815229996134,
      "grad_norm": 0.0,
      "learning_rate": 1.6283339775802087e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13957
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.5625,
      "epoch": 0.6744298415152686,
      "grad_norm": 0.0,
      "learning_rate": 1.6280923850019326e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13958
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.40625,
      "epoch": 0.6744781600309239,
      "grad_norm": 0.0,
      "learning_rate": 1.6278507924236567e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13959
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.34375,
      "epoch": 0.6745264785465791,
      "grad_norm": 0.0,
      "learning_rate": 1.6276091998453806e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13960
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 56.71875,
      "epoch": 0.6745747970622342,
      "grad_norm": 0.0,
      "learning_rate": 1.6273676072671047e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 13961
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 57.40625,
      "epoch": 0.6746231155778895,
      "grad_norm": 0.0,
      "learning_rate": 1.627126014688829e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13962
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.03125,
      "epoch": 0.6746714340935447,
      "grad_norm": 0.0,
      "learning_rate": 1.6268844221105527e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13963
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.96875,
      "epoch": 0.6747197526091998,
      "grad_norm": 0.0,
      "learning_rate": 1.626642829532277e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 13964
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.375,
      "epoch": 0.674768071124855,
      "grad_norm": 0.0,
      "learning_rate": 1.6264012369540005e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13965
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.53125,
      "epoch": 0.6748163896405103,
      "grad_norm": 0.0,
      "learning_rate": 1.6261596443757246e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 13966
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.3125,
      "epoch": 0.6748647081561654,
      "grad_norm": 0.0,
      "learning_rate": 1.6259180517974488e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13967
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 69.125,
      "epoch": 0.6749130266718206,
      "grad_norm": 0.0,
      "learning_rate": 1.6256764592191726e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 13968
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.96875,
      "epoch": 0.6749613451874759,
      "grad_norm": 0.0,
      "learning_rate": 1.6254348666408968e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13969
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.21875,
      "epoch": 0.6750096637031311,
      "grad_norm": 0.0,
      "learning_rate": 1.6251932740626206e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13970
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.5,
      "epoch": 0.6750579822187862,
      "grad_norm": 0.0,
      "learning_rate": 1.6249516814843448e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13971
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.90625,
      "epoch": 0.6751063007344414,
      "grad_norm": 0.0,
      "learning_rate": 1.624710088906069e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 13972
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.0625,
      "epoch": 0.6751546192500967,
      "grad_norm": 0.0,
      "learning_rate": 1.6244684963277928e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13973
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.625,
      "epoch": 0.6752029377657518,
      "grad_norm": 0.0,
      "learning_rate": 1.624226903749517e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13974
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.84375,
      "epoch": 0.675251256281407,
      "grad_norm": 0.0,
      "learning_rate": 1.6239853111712405e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13975
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.84375,
      "epoch": 0.6752995747970623,
      "grad_norm": 0.0,
      "learning_rate": 1.6237437185929647e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 13976
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.1875,
      "epoch": 0.6753478933127174,
      "grad_norm": 0.0,
      "learning_rate": 1.6235021260146888e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13977
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.53125,
      "epoch": 0.6753962118283726,
      "grad_norm": 0.0,
      "learning_rate": 1.6232605334364127e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13978
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.96875,
      "epoch": 0.6754445303440278,
      "grad_norm": 0.0,
      "learning_rate": 1.6230189408581368e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 13979
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.78125,
      "epoch": 0.6754928488596831,
      "grad_norm": 0.0,
      "learning_rate": 1.6227773482798607e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13980
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.1875,
      "epoch": 0.6755411673753382,
      "grad_norm": 0.0,
      "learning_rate": 1.6225357557015848e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13981
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.125,
      "epoch": 0.6755894858909934,
      "grad_norm": 0.0,
      "learning_rate": 1.622294163123309e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13982
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.09375,
      "epoch": 0.6756378044066487,
      "grad_norm": 0.0,
      "learning_rate": 1.6220525705450328e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 13983
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.03125,
      "epoch": 0.6756861229223038,
      "grad_norm": 0.0,
      "learning_rate": 1.621810977966757e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13984
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.1875,
      "epoch": 0.675734441437959,
      "grad_norm": 0.0,
      "learning_rate": 1.6215693853884808e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13985
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.03125,
      "epoch": 0.6757827599536143,
      "grad_norm": 0.0,
      "learning_rate": 1.6213277928102047e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 13986
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.8125,
      "epoch": 0.6758310784692694,
      "grad_norm": 0.0,
      "learning_rate": 1.6210862002319288e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 13987
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.3125,
      "epoch": 0.6758793969849246,
      "grad_norm": 0.0,
      "learning_rate": 1.6208446076536527e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13988
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 59.03125,
      "epoch": 0.6759277155005798,
      "grad_norm": 0.0,
      "learning_rate": 1.6206030150753769e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13989
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 61.4375,
      "epoch": 0.6759760340162351,
      "grad_norm": 0.0,
      "learning_rate": 1.6203614224971007e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 13990
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.15625,
      "epoch": 0.6760243525318902,
      "grad_norm": 0.0,
      "learning_rate": 1.6201198299188249e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 13991
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.21875,
      "epoch": 0.6760726710475454,
      "grad_norm": 0.0,
      "learning_rate": 1.619878237340549e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 13992
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.125,
      "epoch": 0.6761209895632007,
      "grad_norm": 0.0,
      "learning_rate": 1.6196366447622729e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 13993
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.0,
      "epoch": 0.6761693080788558,
      "grad_norm": 0.0,
      "learning_rate": 1.619395052183997e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 13994
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.375,
      "epoch": 0.676217626594511,
      "grad_norm": 0.0,
      "learning_rate": 1.619153459605721e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 13995
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.8125,
      "epoch": 0.6762659451101662,
      "grad_norm": 0.0,
      "learning_rate": 1.6189118670274447e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 13996
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.15625,
      "epoch": 0.6763142636258214,
      "grad_norm": 0.0,
      "learning_rate": 1.618670274449169e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 13997
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.03125,
      "epoch": 0.6763625821414766,
      "grad_norm": 0.0,
      "learning_rate": 1.6184286818708928e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 13998
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 57.8125,
      "epoch": 0.6764109006571318,
      "grad_norm": 0.0,
      "learning_rate": 1.618187089292617e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 13999
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.28125,
      "epoch": 0.6764592191727871,
      "grad_norm": 0.0,
      "learning_rate": 1.6179454967143408e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14000
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 66.25,
      "epoch": 0.6765075376884422,
      "grad_norm": 0.0,
      "learning_rate": 1.617703904136065e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 14001
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.6875,
      "epoch": 0.6765558562040974,
      "grad_norm": 0.0,
      "learning_rate": 1.617462311557789e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14002
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 56.03125,
      "epoch": 0.6766041747197526,
      "grad_norm": 0.0,
      "learning_rate": 1.617220718979513e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14003
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.0625,
      "epoch": 0.6766524932354078,
      "grad_norm": 0.0,
      "learning_rate": 1.616979126401237e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14004
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.21875,
      "epoch": 0.676700811751063,
      "grad_norm": 0.0,
      "learning_rate": 1.616737533822961e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14005
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.5625,
      "epoch": 0.6767491302667182,
      "grad_norm": 0.0,
      "learning_rate": 1.6164959412446848e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14006
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 59.4375,
      "epoch": 0.6767974487823734,
      "grad_norm": 27.033241473913627,
      "learning_rate": 1.616254348666409e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.3298332095146179,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14007
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.78125,
      "epoch": 0.6768457672980286,
      "grad_norm": 0.0,
      "learning_rate": 1.6160127560881328e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14008
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.4375,
      "epoch": 0.6768940858136838,
      "grad_norm": 0.0,
      "learning_rate": 1.615771163509857e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14009
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.75,
      "epoch": 0.676942404329339,
      "grad_norm": 0.0,
      "learning_rate": 1.6155295709315808e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14010
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.46875,
      "epoch": 0.6769907228449942,
      "grad_norm": 0.0,
      "learning_rate": 1.615287978353305e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14011
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.3125,
      "epoch": 0.6770390413606494,
      "grad_norm": 0.0,
      "learning_rate": 1.615046385775029e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14012
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.53125,
      "epoch": 0.6770873598763046,
      "grad_norm": 0.0,
      "learning_rate": 1.614804793196753e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14013
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 52.875,
      "epoch": 0.6771356783919598,
      "grad_norm": 0.0,
      "learning_rate": 1.614563200618477e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14014
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.90625,
      "epoch": 0.677183996907615,
      "grad_norm": 0.0,
      "learning_rate": 1.614321608040201e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14015
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.84375,
      "epoch": 0.6772323154232702,
      "grad_norm": 0.0,
      "learning_rate": 1.614080015461925e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14016
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.625,
      "epoch": 0.6772806339389253,
      "grad_norm": 0.0,
      "learning_rate": 1.613838422883649e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14017
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.65625,
      "epoch": 0.6773289524545806,
      "grad_norm": 0.0,
      "learning_rate": 1.6135968303053728e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14018
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.90625,
      "epoch": 0.6773772709702358,
      "grad_norm": 0.0,
      "learning_rate": 1.613355237727097e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14019
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 61.46875,
      "epoch": 0.677425589485891,
      "grad_norm": 0.0,
      "learning_rate": 1.6131136451488208e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14020
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.0625,
      "epoch": 0.6774739080015462,
      "grad_norm": 0.0,
      "learning_rate": 1.612872052570545e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14021
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.71875,
      "epoch": 0.6775222265172014,
      "grad_norm": 0.0,
      "learning_rate": 1.612630459992269e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14022
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 62.09375,
      "epoch": 0.6775705450328566,
      "grad_norm": 0.0,
      "learning_rate": 1.612388867413993e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 14023
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.90625,
      "epoch": 0.6776188635485118,
      "grad_norm": 0.0,
      "learning_rate": 1.612147274835717e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 14024
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 54.40625,
      "epoch": 0.677667182064167,
      "grad_norm": 0.0,
      "learning_rate": 1.611905682257441e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14025
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 65.0,
      "epoch": 0.6777155005798222,
      "grad_norm": 0.0,
      "learning_rate": 1.611664089679165e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14026
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.125,
      "epoch": 0.6777638190954773,
      "grad_norm": 0.0,
      "learning_rate": 1.611422497100889e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14027
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 63.40625,
      "epoch": 0.6778121376111326,
      "grad_norm": 0.0,
      "learning_rate": 1.611180904522613e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14028
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.375,
      "epoch": 0.6778604561267878,
      "grad_norm": 0.0,
      "learning_rate": 1.610939311944337e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14029
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.03125,
      "epoch": 0.677908774642443,
      "grad_norm": 0.0,
      "learning_rate": 1.610697719366061e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14030
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.375,
      "epoch": 0.6779570931580982,
      "grad_norm": 0.0,
      "learning_rate": 1.610456126787785e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14031
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.90625,
      "epoch": 0.6780054116737534,
      "grad_norm": 0.0,
      "learning_rate": 1.6102145342095092e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14032
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.75,
      "epoch": 0.6780537301894086,
      "grad_norm": 40.49163283991163,
      "learning_rate": 1.609972941631233e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.3571978509426117,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14033
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 69.40625,
      "epoch": 0.6781020487050637,
      "grad_norm": 17.68142868961182,
      "learning_rate": 1.6097313490529572e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.323003351688385,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14034
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.4375,
      "epoch": 0.678150367220719,
      "grad_norm": 0.0,
      "learning_rate": 1.609489756474681e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14035
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.1875,
      "epoch": 0.6781986857363742,
      "grad_norm": 0.0,
      "learning_rate": 1.6092481638964052e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14036
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.15625,
      "epoch": 0.6782470042520293,
      "grad_norm": 0.0,
      "learning_rate": 1.609006571318129e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14037
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.5625,
      "epoch": 0.6782953227676846,
      "grad_norm": 0.0,
      "learning_rate": 1.608764978739853e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 14038
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.6875,
      "epoch": 0.6783436412833398,
      "grad_norm": 0.0,
      "learning_rate": 1.608523386161577e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14039
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.875,
      "epoch": 0.678391959798995,
      "grad_norm": 0.0,
      "learning_rate": 1.608281793583301e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14040
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.9375,
      "epoch": 0.6784402783146501,
      "grad_norm": 0.0,
      "learning_rate": 1.608040201005025e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14041
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.03125,
      "epoch": 0.6784885968303054,
      "grad_norm": 0.0,
      "learning_rate": 1.6077986084267492e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14042
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.09375,
      "epoch": 0.6785369153459606,
      "grad_norm": 0.0,
      "learning_rate": 1.607557015848473e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14043
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.15625,
      "epoch": 0.6785852338616157,
      "grad_norm": 0.0,
      "learning_rate": 1.6073154232701972e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14044
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.9375,
      "epoch": 0.678633552377271,
      "grad_norm": 0.0,
      "learning_rate": 1.607073830691921e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14045
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.78125,
      "epoch": 0.6786818708929262,
      "grad_norm": 0.0,
      "learning_rate": 1.6068322381136452e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14046
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.625,
      "epoch": 0.6787301894085813,
      "grad_norm": 0.0,
      "learning_rate": 1.6065906455353693e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14047
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.03125,
      "epoch": 0.6787785079242366,
      "grad_norm": 0.0,
      "learning_rate": 1.606349052957093e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14048
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.84375,
      "epoch": 0.6788268264398918,
      "grad_norm": 0.0,
      "learning_rate": 1.606107460378817e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14049
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.5625,
      "epoch": 0.678875144955547,
      "grad_norm": 0.0,
      "learning_rate": 1.605865867800541e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14050
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.96875,
      "epoch": 0.6789234634712021,
      "grad_norm": 0.0,
      "learning_rate": 1.605624275222265e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14051
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.6875,
      "epoch": 0.6789717819868574,
      "grad_norm": 0.0,
      "learning_rate": 1.6053826826439892e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14052
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.65625,
      "epoch": 0.6790201005025126,
      "grad_norm": 0.0,
      "learning_rate": 1.605141090065713e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14053
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.90625,
      "epoch": 0.6790684190181677,
      "grad_norm": 0.0,
      "learning_rate": 1.6048994974874372e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14054
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.75,
      "epoch": 0.679116737533823,
      "grad_norm": 11.188876359113994,
      "learning_rate": 1.604657904909161e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14055
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.65625,
      "epoch": 0.6791650560494782,
      "grad_norm": 0.0,
      "learning_rate": 1.6044163123308852e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23568853735923767,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14056
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.3125,
      "epoch": 0.6792133745651333,
      "grad_norm": 0.0,
      "learning_rate": 1.6041747197526094e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14057
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.15625,
      "epoch": 0.6792616930807885,
      "grad_norm": 0.0,
      "learning_rate": 1.603933127174333e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.19044628739356995,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14058
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.78125,
      "epoch": 0.6793100115964438,
      "grad_norm": 0.0,
      "learning_rate": 1.603691534596057e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14059
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.46875,
      "epoch": 0.679358330112099,
      "grad_norm": 0.0,
      "learning_rate": 1.603449942017781e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14060
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.625,
      "epoch": 0.6794066486277541,
      "grad_norm": 0.0,
      "learning_rate": 1.6032083494395051e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14061
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.65625,
      "epoch": 0.6794549671434094,
      "grad_norm": 0.0,
      "learning_rate": 1.6029667568612293e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14062
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.125,
      "epoch": 0.6795032856590646,
      "grad_norm": 0.0,
      "learning_rate": 1.6027251642829531e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14063
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.9375,
      "epoch": 0.6795516041747197,
      "grad_norm": 0.0,
      "learning_rate": 1.6024835717046773e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14064
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.8125,
      "epoch": 0.679599922690375,
      "grad_norm": 0.0,
      "learning_rate": 1.6022419791264011e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14065
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.5625,
      "epoch": 0.6796482412060302,
      "grad_norm": 0.0,
      "learning_rate": 1.6020003865481253e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14066
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.625,
      "epoch": 0.6796965597216853,
      "grad_norm": 0.0,
      "learning_rate": 1.6017587939698494e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14067
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.125,
      "epoch": 0.6797448782373405,
      "grad_norm": 0.0,
      "learning_rate": 1.601517201391573e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14068
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.3125,
      "epoch": 0.6797931967529958,
      "grad_norm": 0.0,
      "learning_rate": 1.6012756088132972e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14069
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.15625,
      "epoch": 0.679841515268651,
      "grad_norm": 0.0,
      "learning_rate": 1.601034016235021e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14070
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.90625,
      "epoch": 0.6798898337843061,
      "grad_norm": 0.0,
      "learning_rate": 1.6007924236567452e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14071
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.21875,
      "epoch": 0.6799381522999614,
      "grad_norm": 0.0,
      "learning_rate": 1.6005508310784693e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14072
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.3125,
      "epoch": 0.6799864708156166,
      "grad_norm": 0.0,
      "learning_rate": 1.6003092385001932e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14073
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.15625,
      "epoch": 0.6800347893312717,
      "grad_norm": 0.0,
      "learning_rate": 1.6000676459219173e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14074
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.71875,
      "epoch": 0.6800831078469269,
      "grad_norm": 0.0,
      "learning_rate": 1.5998260533436412e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14075
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.84375,
      "epoch": 0.6801314263625822,
      "grad_norm": 0.0,
      "learning_rate": 1.5995844607653653e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14076
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.09375,
      "epoch": 0.6801797448782373,
      "grad_norm": 0.0,
      "learning_rate": 1.5993428681870895e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14077
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.21875,
      "epoch": 0.6802280633938925,
      "grad_norm": 0.0,
      "learning_rate": 1.599101275608813e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14078
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.71875,
      "epoch": 0.6802763819095478,
      "grad_norm": 0.0,
      "learning_rate": 1.5988596830305372e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14079
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.5625,
      "epoch": 0.680324700425203,
      "grad_norm": 0.0,
      "learning_rate": 1.598618090452261e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 14080
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.125,
      "epoch": 0.6803730189408581,
      "grad_norm": 0.0,
      "learning_rate": 1.5983764978739852e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14081
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.8125,
      "epoch": 0.6804213374565133,
      "grad_norm": 0.0,
      "learning_rate": 1.5981349052957093e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14082
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.71875,
      "epoch": 0.6804696559721686,
      "grad_norm": 0.0,
      "learning_rate": 1.5978933127174332e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14083
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.125,
      "epoch": 0.6805179744878237,
      "grad_norm": 0.0,
      "learning_rate": 1.5976517201391574e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14084
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 58.09375,
      "epoch": 0.6805662930034789,
      "grad_norm": 0.0,
      "learning_rate": 1.5974101275608812e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14085
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 64.78125,
      "epoch": 0.6806146115191342,
      "grad_norm": 0.0,
      "learning_rate": 1.5971685349826054e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14086
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.375,
      "epoch": 0.6806629300347893,
      "grad_norm": 0.0,
      "learning_rate": 1.5969269424043295e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14087
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.75,
      "epoch": 0.6807112485504445,
      "grad_norm": 0.0,
      "learning_rate": 1.5966853498260534e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14088
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 63.75,
      "epoch": 0.6807595670660997,
      "grad_norm": 0.0,
      "learning_rate": 1.5964437572477772e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14089
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.6875,
      "epoch": 0.680807885581755,
      "grad_norm": 0.0,
      "learning_rate": 1.596202164669501e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14090
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.46875,
      "epoch": 0.6808562040974101,
      "grad_norm": 0.0,
      "learning_rate": 1.5959605720912253e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 14091
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.25,
      "epoch": 0.6809045226130653,
      "grad_norm": 0.0,
      "learning_rate": 1.5957189795129494e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14092
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 67.25,
      "epoch": 0.6809528411287206,
      "grad_norm": 0.0,
      "learning_rate": 1.5954773869346733e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14093
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.3125,
      "epoch": 0.6810011596443757,
      "grad_norm": 0.0,
      "learning_rate": 1.5952357943563974e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14094
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 59.4375,
      "epoch": 0.6810494781600309,
      "grad_norm": 0.0,
      "learning_rate": 1.5949942017781213e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14095
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.46875,
      "epoch": 0.6810977966756862,
      "grad_norm": 0.0,
      "learning_rate": 1.5947526091998454e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14096
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.21875,
      "epoch": 0.6811461151913413,
      "grad_norm": 0.0,
      "learning_rate": 1.5945110166215695e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14097
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.71875,
      "epoch": 0.6811944337069965,
      "grad_norm": 0.0,
      "learning_rate": 1.5942694240432934e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14098
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.84375,
      "epoch": 0.6812427522226517,
      "grad_norm": 0.0,
      "learning_rate": 1.5940278314650173e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14099
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.34375,
      "epoch": 0.681291070738307,
      "grad_norm": 0.0,
      "learning_rate": 1.5937862388867412e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14100
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.3125,
      "epoch": 0.6813393892539621,
      "grad_norm": 0.0,
      "learning_rate": 1.5935446463084653e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14101
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 74.28125,
      "epoch": 0.6813877077696173,
      "grad_norm": 50.64523095786729,
      "learning_rate": 1.5933030537301894e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.29854315519332886,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14102
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.375,
      "epoch": 0.6814360262852726,
      "grad_norm": 0.0,
      "learning_rate": 1.5930614611519133e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14103
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.125,
      "epoch": 0.6814843448009277,
      "grad_norm": 0.0,
      "learning_rate": 1.5928198685736374e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14104
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.28125,
      "epoch": 0.6815326633165829,
      "grad_norm": 0.0,
      "learning_rate": 1.5925782759953613e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14105
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.90625,
      "epoch": 0.6815809818322381,
      "grad_norm": 0.0,
      "learning_rate": 1.5923366834170854e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14106
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 55.59375,
      "epoch": 0.6816293003478933,
      "grad_norm": 0.0,
      "learning_rate": 1.5920950908388096e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14107
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.5,
      "epoch": 0.6816776188635485,
      "grad_norm": 0.0,
      "learning_rate": 1.5918534982605335e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14108
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.15625,
      "epoch": 0.6817259373792037,
      "grad_norm": 0.0,
      "learning_rate": 1.5916119056822573e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14109
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.46875,
      "epoch": 0.681774255894859,
      "grad_norm": 0.0,
      "learning_rate": 1.5913703131039812e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14110
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.5,
      "epoch": 0.6818225744105141,
      "grad_norm": 0.0,
      "learning_rate": 1.5911287205257053e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14111
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.75,
      "epoch": 0.6818708929261693,
      "grad_norm": 0.0,
      "learning_rate": 1.5908871279474295e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14112
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.25,
      "epoch": 0.6819192114418245,
      "grad_norm": 0.0,
      "learning_rate": 1.5906455353691533e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14113
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.15625,
      "epoch": 0.6819675299574797,
      "grad_norm": 0.0,
      "learning_rate": 1.5904039427908775e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14114
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.03125,
      "epoch": 0.6820158484731349,
      "grad_norm": 0.0,
      "learning_rate": 1.5901623502126013e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14115
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.90625,
      "epoch": 0.6820641669887901,
      "grad_norm": 0.0,
      "learning_rate": 1.5899207576343255e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14116
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.65625,
      "epoch": 0.6821124855044453,
      "grad_norm": 0.0,
      "learning_rate": 1.5896791650560496e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14117
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 65.84375,
      "epoch": 0.6821608040201005,
      "grad_norm": 0.0,
      "learning_rate": 1.5894375724777735e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14118
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.59375,
      "epoch": 0.6822091225357557,
      "grad_norm": 0.0,
      "learning_rate": 1.5891959798994976e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14119
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.5625,
      "epoch": 0.682257441051411,
      "grad_norm": 0.0,
      "learning_rate": 1.5889543873212212e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14120
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.71875,
      "epoch": 0.6823057595670661,
      "grad_norm": 0.0,
      "learning_rate": 1.5887127947429454e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14121
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.75,
      "epoch": 0.6823540780827213,
      "grad_norm": 0.0,
      "learning_rate": 1.5884712021646695e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 14122
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.53125,
      "epoch": 0.6824023965983765,
      "grad_norm": 0.0,
      "learning_rate": 1.5882296095863934e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14123
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.40625,
      "epoch": 0.6824507151140317,
      "grad_norm": 0.0,
      "learning_rate": 1.5879880170081175e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14124
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.65625,
      "epoch": 0.6824990336296869,
      "grad_norm": 0.0,
      "learning_rate": 1.5877464244298414e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14125
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 65.21875,
      "epoch": 0.6825473521453421,
      "grad_norm": 0.0,
      "learning_rate": 1.5875048318515655e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14126
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 59.65625,
      "epoch": 0.6825956706609972,
      "grad_norm": 0.0,
      "learning_rate": 1.5872632392732897e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14127
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.09375,
      "epoch": 0.6826439891766525,
      "grad_norm": 0.0,
      "learning_rate": 1.5870216466950135e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14128
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 65.5625,
      "epoch": 0.6826923076923077,
      "grad_norm": 0.0,
      "learning_rate": 1.5867800541167377e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14129
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 68.40625,
      "epoch": 0.6827406262079629,
      "grad_norm": 0.0,
      "learning_rate": 1.5865384615384613e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14130
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.1875,
      "epoch": 0.6827889447236181,
      "grad_norm": 0.0,
      "learning_rate": 1.5862968689601854e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14131
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.65625,
      "epoch": 0.6828372632392733,
      "grad_norm": 0.0,
      "learning_rate": 1.5860552763819095e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14132
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.90625,
      "epoch": 0.6828855817549285,
      "grad_norm": 0.0,
      "learning_rate": 1.5858136838036334e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14133
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.8125,
      "epoch": 0.6829339002705836,
      "grad_norm": 0.0,
      "learning_rate": 1.5855720912253576e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14134
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.59375,
      "epoch": 0.6829822187862389,
      "grad_norm": 0.0,
      "learning_rate": 1.5853304986470814e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14135
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.375,
      "epoch": 0.6830305373018941,
      "grad_norm": 0.0,
      "learning_rate": 1.5850889060688056e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14136
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.8125,
      "epoch": 0.6830788558175492,
      "grad_norm": 0.0,
      "learning_rate": 1.5848473134905297e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14137
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.5,
      "epoch": 0.6831271743332045,
      "grad_norm": 0.0,
      "learning_rate": 1.5846057209122536e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14138
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.53125,
      "epoch": 0.6831754928488597,
      "grad_norm": 0.0,
      "learning_rate": 1.5843641283339777e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 14139
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 60.0625,
      "epoch": 0.6832238113645149,
      "grad_norm": 0.0,
      "learning_rate": 1.5841225357557013e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 14140
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.6875,
      "epoch": 0.68327212988017,
      "grad_norm": 0.0,
      "learning_rate": 1.5838809431774254e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14141
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.5625,
      "epoch": 0.6833204483958253,
      "grad_norm": 0.0,
      "learning_rate": 1.5836393505991496e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14142
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.875,
      "epoch": 0.6833687669114805,
      "grad_norm": 0.0,
      "learning_rate": 1.5833977580208735e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14143
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 64.8125,
      "epoch": 0.6834170854271356,
      "grad_norm": 0.0,
      "learning_rate": 1.5831561654425976e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14144
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.15625,
      "epoch": 0.6834654039427909,
      "grad_norm": 0.0,
      "learning_rate": 1.5829145728643215e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14145
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.90625,
      "epoch": 0.6835137224584461,
      "grad_norm": 0.0,
      "learning_rate": 1.5826729802860456e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14146
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.28125,
      "epoch": 0.6835620409741012,
      "grad_norm": 0.0,
      "learning_rate": 1.5824313877077697e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14147
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 62.34375,
      "epoch": 0.6836103594897565,
      "grad_norm": 8.163831386089539,
      "learning_rate": 1.5821897951294936e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.32781746983528137,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 0.96875,
      "step": 14148
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 55.03125,
      "epoch": 0.6836586780054117,
      "grad_norm": 0.0,
      "learning_rate": 1.5819482025512177e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14149
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.9375,
      "epoch": 0.6837069965210669,
      "grad_norm": 0.0,
      "learning_rate": 1.5817066099729414e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14150
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.15625,
      "epoch": 0.683755315036722,
      "grad_norm": 0.0,
      "learning_rate": 1.5814650173946655e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14151
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.90625,
      "epoch": 0.6838036335523773,
      "grad_norm": 0.0,
      "learning_rate": 1.5812234248163896e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14152
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.21875,
      "epoch": 0.6838519520680325,
      "grad_norm": 0.0,
      "learning_rate": 1.5809818322381135e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14153
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.46875,
      "epoch": 0.6839002705836876,
      "grad_norm": 0.0,
      "learning_rate": 1.5807402396598376e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14154
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.0625,
      "epoch": 0.6839485890993429,
      "grad_norm": 0.0,
      "learning_rate": 1.5804986470815615e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14155
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.9375,
      "epoch": 0.6839969076149981,
      "grad_norm": 0.0,
      "learning_rate": 1.5802570545032856e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14156
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.34375,
      "epoch": 0.6840452261306532,
      "grad_norm": 0.0,
      "learning_rate": 1.5800154619250098e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14157
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.875,
      "epoch": 0.6840935446463084,
      "grad_norm": 0.0,
      "learning_rate": 1.5797738693467336e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14158
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.71875,
      "epoch": 0.6841418631619637,
      "grad_norm": 0.0,
      "learning_rate": 1.5795322767684578e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14159
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.625,
      "epoch": 0.6841901816776189,
      "grad_norm": 0.0,
      "learning_rate": 1.5792906841901814e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14160
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.84375,
      "epoch": 0.684238500193274,
      "grad_norm": 0.0,
      "learning_rate": 1.5790490916119055e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14161
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.0625,
      "epoch": 0.6842868187089293,
      "grad_norm": 0.0,
      "learning_rate": 1.5788074990336297e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14162
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.78125,
      "epoch": 0.6843351372245845,
      "grad_norm": 0.0,
      "learning_rate": 1.5785659064553535e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14163
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.71875,
      "epoch": 0.6843834557402396,
      "grad_norm": 0.0,
      "learning_rate": 1.5783243138770777e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14164
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 61.96875,
      "epoch": 0.6844317742558949,
      "grad_norm": 0.0,
      "learning_rate": 1.5780827212988015e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14165
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.84375,
      "epoch": 0.6844800927715501,
      "grad_norm": 0.0,
      "learning_rate": 1.5778411287205257e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14166
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.4375,
      "epoch": 0.6845284112872052,
      "grad_norm": 0.0,
      "learning_rate": 1.5775995361422498e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14167
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.09375,
      "epoch": 0.6845767298028604,
      "grad_norm": 0.0,
      "learning_rate": 1.5773579435639737e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14168
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.59375,
      "epoch": 0.6846250483185157,
      "grad_norm": 0.0,
      "learning_rate": 1.5771163509856978e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14169
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.8125,
      "epoch": 0.6846733668341709,
      "grad_norm": 0.0,
      "learning_rate": 1.5768747584074217e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14170
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.6875,
      "epoch": 0.684721685349826,
      "grad_norm": 0.0,
      "learning_rate": 1.5766331658291456e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14171
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.4375,
      "epoch": 0.6847700038654813,
      "grad_norm": 0.0,
      "learning_rate": 1.5763915732508697e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14172
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.75,
      "epoch": 0.6848183223811365,
      "grad_norm": 0.0,
      "learning_rate": 1.5761499806725936e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14173
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.90625,
      "epoch": 0.6848666408967916,
      "grad_norm": 0.0,
      "learning_rate": 1.5759083880943177e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14174
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 64.0,
      "epoch": 0.6849149594124468,
      "grad_norm": 0.0,
      "learning_rate": 1.5756667955160416e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14175
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.625,
      "epoch": 0.6849632779281021,
      "grad_norm": 0.0,
      "learning_rate": 1.5754252029377657e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14176
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.65625,
      "epoch": 0.6850115964437572,
      "grad_norm": 0.0,
      "learning_rate": 1.5751836103594899e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14177
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.0,
      "epoch": 0.6850599149594124,
      "grad_norm": 0.0,
      "learning_rate": 1.5749420177812137e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14178
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 75.1875,
      "epoch": 0.6851082334750677,
      "grad_norm": 15.606901965796187,
      "learning_rate": 1.5747004252029379e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14179
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.03125,
      "epoch": 0.6851565519907229,
      "grad_norm": 0.0,
      "learning_rate": 1.5744588326246617e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14180
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 60.21875,
      "epoch": 0.685204870506378,
      "grad_norm": 0.0,
      "learning_rate": 1.5742172400463856e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14181
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.59375,
      "epoch": 0.6852531890220332,
      "grad_norm": 0.0,
      "learning_rate": 1.5739756474681097e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14182
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.9375,
      "epoch": 0.6853015075376885,
      "grad_norm": 0.0,
      "learning_rate": 1.5737340548898336e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14183
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.3125,
      "epoch": 0.6853498260533436,
      "grad_norm": 0.0,
      "learning_rate": 1.5734924623115578e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14184
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.0,
      "epoch": 0.6853981445689988,
      "grad_norm": 0.0,
      "learning_rate": 1.5732508697332816e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14185
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.46875,
      "epoch": 0.6854464630846541,
      "grad_norm": 0.0,
      "learning_rate": 1.5730092771550058e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14186
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.53125,
      "epoch": 0.6854947816003092,
      "grad_norm": 0.0,
      "learning_rate": 1.57276768457673e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14187
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.90625,
      "epoch": 0.6855431001159644,
      "grad_norm": 0.0,
      "learning_rate": 1.5725260919984538e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14188
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 65.03125,
      "epoch": 0.6855914186316197,
      "grad_norm": 32.09716869792565,
      "learning_rate": 1.572284499420178e-07,
      "loss": -0.0,
      "reward": 2.34375,
      "reward_std": 0.3014557659626007,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14189
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.15625,
      "epoch": 0.6856397371472749,
      "grad_norm": 0.0,
      "learning_rate": 1.5720429068419018e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14190
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 70.15625,
      "epoch": 0.68568805566293,
      "grad_norm": 0.0,
      "learning_rate": 1.5718013142636256e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 14191
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.46875,
      "epoch": 0.6857363741785852,
      "grad_norm": 0.0,
      "learning_rate": 1.5715597216853498e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14192
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.15625,
      "epoch": 0.6857846926942405,
      "grad_norm": 0.0,
      "learning_rate": 1.5713181291070737e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14193
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.5625,
      "epoch": 0.6858330112098956,
      "grad_norm": 0.0,
      "learning_rate": 1.5710765365287978e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14194
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.90625,
      "epoch": 0.6858813297255508,
      "grad_norm": 0.0,
      "learning_rate": 1.5708349439505217e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14195
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.34375,
      "epoch": 0.6859296482412061,
      "grad_norm": 0.0,
      "learning_rate": 1.5705933513722458e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14196
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.375,
      "epoch": 0.6859779667568612,
      "grad_norm": 0.0,
      "learning_rate": 1.57035175879397e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14197
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.125,
      "epoch": 0.6860262852725164,
      "grad_norm": 0.0,
      "learning_rate": 1.5701101662156938e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14198
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.1875,
      "epoch": 0.6860746037881716,
      "grad_norm": 0.0,
      "learning_rate": 1.569868573637418e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14199
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.03125,
      "epoch": 0.6861229223038269,
      "grad_norm": 0.0,
      "learning_rate": 1.5696269810591418e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14200
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 62.6875,
      "epoch": 0.686171240819482,
      "grad_norm": 0.0,
      "learning_rate": 1.569385388480866e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14201
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.625,
      "epoch": 0.6862195593351372,
      "grad_norm": 0.0,
      "learning_rate": 1.5691437959025898e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14202
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 61.0625,
      "epoch": 0.6862678778507925,
      "grad_norm": 0.0,
      "learning_rate": 1.5689022033243137e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14203
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.40625,
      "epoch": 0.6863161963664476,
      "grad_norm": 0.0,
      "learning_rate": 1.5686606107460378e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14204
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.40625,
      "epoch": 0.6863645148821028,
      "grad_norm": 0.0,
      "learning_rate": 1.5684190181677617e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14205
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.6875,
      "epoch": 0.686412833397758,
      "grad_norm": 0.0,
      "learning_rate": 1.5681774255894858e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14206
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 58.09375,
      "epoch": 0.6864611519134132,
      "grad_norm": 0.0,
      "learning_rate": 1.56793583301121e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14207
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.09375,
      "epoch": 0.6865094704290684,
      "grad_norm": 0.0,
      "learning_rate": 1.5676942404329338e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14208
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.28125,
      "epoch": 0.6865577889447236,
      "grad_norm": 0.0,
      "learning_rate": 1.567452647854658e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14209
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.90625,
      "epoch": 0.6866061074603789,
      "grad_norm": 0.0,
      "learning_rate": 1.5672110552763819e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14210
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 61.53125,
      "epoch": 0.686654425976034,
      "grad_norm": 0.0,
      "learning_rate": 1.566969462698106e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14211
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.40625,
      "epoch": 0.6867027444916892,
      "grad_norm": 0.0,
      "learning_rate": 1.5667278701198299e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14212
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 62.03125,
      "epoch": 0.6867510630073445,
      "grad_norm": 0.0,
      "learning_rate": 1.5664862775415537e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14213
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.3125,
      "epoch": 0.6867993815229996,
      "grad_norm": 0.0,
      "learning_rate": 1.566244684963278e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14214
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.75,
      "epoch": 0.6868477000386548,
      "grad_norm": 0.0,
      "learning_rate": 1.5660030923850017e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14215
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.875,
      "epoch": 0.68689601855431,
      "grad_norm": 0.0,
      "learning_rate": 1.565761499806726e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14216
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.4375,
      "epoch": 0.6869443370699652,
      "grad_norm": 0.0,
      "learning_rate": 1.56551990722845e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14217
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.0,
      "epoch": 0.6869926555856204,
      "grad_norm": 0.0,
      "learning_rate": 1.565278314650174e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14218
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.09375,
      "epoch": 0.6870409741012756,
      "grad_norm": 0.0,
      "learning_rate": 1.565036722071898e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14219
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.75,
      "epoch": 0.6870892926169309,
      "grad_norm": 0.0,
      "learning_rate": 1.564795129493622e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14220
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.5,
      "epoch": 0.687137611132586,
      "grad_norm": 0.0,
      "learning_rate": 1.564553536915346e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 14221
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 63.625,
      "epoch": 0.6871859296482412,
      "grad_norm": 0.0,
      "learning_rate": 1.5643119443370702e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14222
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 69.90625,
      "epoch": 0.6872342481638964,
      "grad_norm": 15.178183043384548,
      "learning_rate": 1.5640703517587938e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14223
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.96875,
      "epoch": 0.6872825666795516,
      "grad_norm": 0.0,
      "learning_rate": 1.563828759180518e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14224
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.8125,
      "epoch": 0.6873308851952068,
      "grad_norm": 0.0,
      "learning_rate": 1.5635871666022418e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14225
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.3125,
      "epoch": 0.687379203710862,
      "grad_norm": 0.0,
      "learning_rate": 1.563345574023966e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14226
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.59375,
      "epoch": 0.6874275222265172,
      "grad_norm": 0.0,
      "learning_rate": 1.56310398144569e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14227
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.65625,
      "epoch": 0.6874758407421724,
      "grad_norm": 0.0,
      "learning_rate": 1.562862388867414e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14228
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.53125,
      "epoch": 0.6875241592578276,
      "grad_norm": 0.0,
      "learning_rate": 1.562620796289138e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14229
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 63.71875,
      "epoch": 0.6875724777734828,
      "grad_norm": 0.0,
      "learning_rate": 1.562379203710862e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14230
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.53125,
      "epoch": 0.687620796289138,
      "grad_norm": 0.0,
      "learning_rate": 1.562137611132586e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14231
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.625,
      "epoch": 0.6876691148047932,
      "grad_norm": 0.0,
      "learning_rate": 1.5618960185543097e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14232
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.53125,
      "epoch": 0.6877174333204484,
      "grad_norm": 0.0,
      "learning_rate": 1.5616544259760338e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14233
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.78125,
      "epoch": 0.6877657518361036,
      "grad_norm": 0.0,
      "learning_rate": 1.561412833397758e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14234
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.8125,
      "epoch": 0.6878140703517588,
      "grad_norm": 0.0,
      "learning_rate": 1.5611712408194818e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14235
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.375,
      "epoch": 0.687862388867414,
      "grad_norm": 0.0,
      "learning_rate": 1.560929648241206e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14236
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 60.65625,
      "epoch": 0.6879107073830691,
      "grad_norm": 0.0,
      "learning_rate": 1.5606880556629298e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14237
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.59375,
      "epoch": 0.6879590258987244,
      "grad_norm": 0.0,
      "learning_rate": 1.560446463084654e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14238
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.78125,
      "epoch": 0.6880073444143796,
      "grad_norm": 0.0,
      "learning_rate": 1.560204870506378e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14239
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.34375,
      "epoch": 0.6880556629300348,
      "grad_norm": 0.0,
      "learning_rate": 1.559963277928102e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14240
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.625,
      "epoch": 0.68810398144569,
      "grad_norm": 0.0,
      "learning_rate": 1.559721685349826e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14241
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.1875,
      "epoch": 0.6881522999613452,
      "grad_norm": 0.0,
      "learning_rate": 1.55948009277155e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14242
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.5625,
      "epoch": 0.6882006184770004,
      "grad_norm": 0.0,
      "learning_rate": 1.5592385001932739e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14243
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.8125,
      "epoch": 0.6882489369926555,
      "grad_norm": 0.0,
      "learning_rate": 1.558996907614998e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14244
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.78125,
      "epoch": 0.6882972555083108,
      "grad_norm": 0.0,
      "learning_rate": 1.5587553150367219e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14245
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.21875,
      "epoch": 0.688345574023966,
      "grad_norm": 0.0,
      "learning_rate": 1.558513722458446e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14246
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.5,
      "epoch": 0.6883938925396211,
      "grad_norm": 0.0,
      "learning_rate": 1.5582721298801699e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14247
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.0625,
      "epoch": 0.6884422110552764,
      "grad_norm": 0.0,
      "learning_rate": 1.558030537301894e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14248
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.8125,
      "epoch": 0.6884905295709316,
      "grad_norm": 0.0,
      "learning_rate": 1.5577889447236181e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14249
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.53125,
      "epoch": 0.6885388480865868,
      "grad_norm": 0.0,
      "learning_rate": 1.557547352145342e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14250
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 81.34375,
      "epoch": 0.688587166602242,
      "grad_norm": 16.109163326517287,
      "learning_rate": 1.5573057595670661e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.19408094882965088,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14251
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.625,
      "epoch": 0.6886354851178972,
      "grad_norm": 0.0,
      "learning_rate": 1.55706416698879e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14252
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.5,
      "epoch": 0.6886838036335524,
      "grad_norm": 0.0,
      "learning_rate": 1.556822574410514e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14253
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.96875,
      "epoch": 0.6887321221492075,
      "grad_norm": 0.0,
      "learning_rate": 1.556580981832238e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14254
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.28125,
      "epoch": 0.6887804406648628,
      "grad_norm": 0.0,
      "learning_rate": 1.556339389253962e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14255
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.0,
      "epoch": 0.688828759180518,
      "grad_norm": 0.0,
      "learning_rate": 1.556097796675686e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14256
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.0625,
      "epoch": 0.6888770776961731,
      "grad_norm": 0.0,
      "learning_rate": 1.55585620409741e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14257
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.0625,
      "epoch": 0.6889253962118284,
      "grad_norm": 0.0,
      "learning_rate": 1.555614611519134e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14258
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.5,
      "epoch": 0.6889737147274836,
      "grad_norm": 0.0,
      "learning_rate": 1.5553730189408582e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14259
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.875,
      "epoch": 0.6890220332431388,
      "grad_norm": 0.0,
      "learning_rate": 1.555131426362582e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14260
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.34375,
      "epoch": 0.6890703517587939,
      "grad_norm": 0.0,
      "learning_rate": 1.5548898337843062e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14261
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.5625,
      "epoch": 0.6891186702744492,
      "grad_norm": 0.0,
      "learning_rate": 1.55464824120603e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14262
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.375,
      "epoch": 0.6891669887901044,
      "grad_norm": 0.0,
      "learning_rate": 1.554406648627754e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14263
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 67.4375,
      "epoch": 0.6892153073057595,
      "grad_norm": 0.0,
      "learning_rate": 1.554165056049478e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14264
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.46875,
      "epoch": 0.6892636258214148,
      "grad_norm": 0.0,
      "learning_rate": 1.553923463471202e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 14265
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.09375,
      "epoch": 0.68931194433707,
      "grad_norm": 0.0,
      "learning_rate": 1.553681870892926e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14266
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.71875,
      "epoch": 0.6893602628527251,
      "grad_norm": 0.0,
      "learning_rate": 1.55344027831465e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14267
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.78125,
      "epoch": 0.6894085813683803,
      "grad_norm": 0.0,
      "learning_rate": 1.553198685736374e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14268
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.65625,
      "epoch": 0.6894568998840356,
      "grad_norm": 0.0,
      "learning_rate": 1.5529570931580982e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.2561880350112915,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14269
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.5,
      "epoch": 0.6895052183996908,
      "grad_norm": 0.0,
      "learning_rate": 1.552715500579822e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14270
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.09375,
      "epoch": 0.6895535369153459,
      "grad_norm": 0.0,
      "learning_rate": 1.5524739080015462e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14271
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.40625,
      "epoch": 0.6896018554310012,
      "grad_norm": 0.0,
      "learning_rate": 1.55223231542327e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14272
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.4375,
      "epoch": 0.6896501739466564,
      "grad_norm": 0.0,
      "learning_rate": 1.5519907228449942e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14273
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.28125,
      "epoch": 0.6896984924623115,
      "grad_norm": 0.0,
      "learning_rate": 1.551749130266718e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14274
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.9375,
      "epoch": 0.6897468109779668,
      "grad_norm": 0.0,
      "learning_rate": 1.551507537688442e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14275
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.90625,
      "epoch": 0.689795129493622,
      "grad_norm": 0.0,
      "learning_rate": 1.551265945110166e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14276
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.59375,
      "epoch": 0.6898434480092771,
      "grad_norm": 0.0,
      "learning_rate": 1.55102435253189e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14277
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.875,
      "epoch": 0.6898917665249323,
      "grad_norm": 0.0,
      "learning_rate": 1.550782759953614e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14278
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.59375,
      "epoch": 0.6899400850405876,
      "grad_norm": 0.0,
      "learning_rate": 1.5505411673753383e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14279
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.65625,
      "epoch": 0.6899884035562428,
      "grad_norm": 0.0,
      "learning_rate": 1.550299574797062e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14280
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.78125,
      "epoch": 0.6900367220718979,
      "grad_norm": 0.0,
      "learning_rate": 1.5500579822187863e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14281
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.21875,
      "epoch": 0.6900850405875532,
      "grad_norm": 0.0,
      "learning_rate": 1.5498163896405101e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14282
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 69.71875,
      "epoch": 0.6901333591032084,
      "grad_norm": 14.11266250037901,
      "learning_rate": 1.5495747970622343e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.22097086906433105,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14283
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.15625,
      "epoch": 0.6901816776188635,
      "grad_norm": 0.0,
      "learning_rate": 1.5493332044839581e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14284
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.09375,
      "epoch": 0.6902299961345187,
      "grad_norm": 0.0,
      "learning_rate": 1.549091611905682e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14285
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.5,
      "epoch": 0.690278314650174,
      "grad_norm": 0.0,
      "learning_rate": 1.5488500193274062e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14286
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.71875,
      "epoch": 0.6903266331658291,
      "grad_norm": 0.0,
      "learning_rate": 1.54860842674913e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14287
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.03125,
      "epoch": 0.6903749516814843,
      "grad_norm": 0.0,
      "learning_rate": 1.5483668341708542e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14288
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.65625,
      "epoch": 0.6904232701971396,
      "grad_norm": 0.0,
      "learning_rate": 1.5481252415925783e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14289
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.59375,
      "epoch": 0.6904715887127948,
      "grad_norm": 0.0,
      "learning_rate": 1.5478836490143022e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14290
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.625,
      "epoch": 0.6905199072284499,
      "grad_norm": 0.0,
      "learning_rate": 1.5476420564360263e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14291
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.90625,
      "epoch": 0.6905682257441051,
      "grad_norm": 0.0,
      "learning_rate": 1.5474004638577502e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14292
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 83.5,
      "epoch": 0.6906165442597604,
      "grad_norm": 0.0,
      "learning_rate": 1.5471588712794743e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 14293
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.53125,
      "epoch": 0.6906648627754155,
      "grad_norm": 0.0,
      "learning_rate": 1.5469172787011982e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14294
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.75,
      "epoch": 0.6907131812910707,
      "grad_norm": 0.0,
      "learning_rate": 1.546675686122922e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14295
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.0,
      "epoch": 0.690761499806726,
      "grad_norm": 0.0,
      "learning_rate": 1.5464340935446462e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14296
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.1875,
      "epoch": 0.6908098183223811,
      "grad_norm": 0.0,
      "learning_rate": 1.54619250096637e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14297
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.5,
      "epoch": 0.6908581368380363,
      "grad_norm": 0.0,
      "learning_rate": 1.5459509083880942e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14298
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.59375,
      "epoch": 0.6909064553536916,
      "grad_norm": 0.0,
      "learning_rate": 1.5457093158098183e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14299
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.09375,
      "epoch": 0.6909547738693468,
      "grad_norm": 0.0,
      "learning_rate": 1.5454677232315422e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 14300
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 69.65625,
      "epoch": 0.6910030923850019,
      "grad_norm": 0.0,
      "learning_rate": 1.5452261306532663e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14301
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.46875,
      "epoch": 0.6910514109006571,
      "grad_norm": 0.0,
      "learning_rate": 1.5449845380749902e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14302
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.46875,
      "epoch": 0.6910997294163124,
      "grad_norm": 0.0,
      "learning_rate": 1.5447429454967144e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14303
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.40625,
      "epoch": 0.6911480479319675,
      "grad_norm": 0.0,
      "learning_rate": 1.5445013529184385e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14304
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.4375,
      "epoch": 0.6911963664476227,
      "grad_norm": 0.0,
      "learning_rate": 1.544259760340162e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14305
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.78125,
      "epoch": 0.691244684963278,
      "grad_norm": 0.0,
      "learning_rate": 1.5440181677618862e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14306
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.0625,
      "epoch": 0.6912930034789331,
      "grad_norm": 0.0,
      "learning_rate": 1.54377657518361e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14307
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.21875,
      "epoch": 0.6913413219945883,
      "grad_norm": 0.0,
      "learning_rate": 1.5435349826053342e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14308
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.125,
      "epoch": 0.6913896405102435,
      "grad_norm": 0.0,
      "learning_rate": 1.5432933900270584e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14309
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.34375,
      "epoch": 0.6914379590258988,
      "grad_norm": 0.0,
      "learning_rate": 1.5430517974487822e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14310
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.21875,
      "epoch": 0.6914862775415539,
      "grad_norm": 0.0,
      "learning_rate": 1.5428102048705064e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14311
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.625,
      "epoch": 0.6915345960572091,
      "grad_norm": 0.0,
      "learning_rate": 1.5425686122922303e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14312
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.0625,
      "epoch": 0.6915829145728644,
      "grad_norm": 0.0,
      "learning_rate": 1.5423270197139544e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14313
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.5,
      "epoch": 0.6916312330885195,
      "grad_norm": 0.0,
      "learning_rate": 1.5420854271356785e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14314
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.1875,
      "epoch": 0.6916795516041747,
      "grad_norm": 0.0,
      "learning_rate": 1.5418438345574021e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14315
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.3125,
      "epoch": 0.69172787011983,
      "grad_norm": 0.0,
      "learning_rate": 1.5416022419791263e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14316
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 87.1875,
      "epoch": 0.6917761886354851,
      "grad_norm": 0.0,
      "learning_rate": 1.5413606494008501e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14317
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.25,
      "epoch": 0.6918245071511403,
      "grad_norm": 0.0,
      "learning_rate": 1.5411190568225743e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14318
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.1875,
      "epoch": 0.6918728256667955,
      "grad_norm": 0.0,
      "learning_rate": 1.5408774642442984e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14319
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.0625,
      "epoch": 0.6919211441824508,
      "grad_norm": 0.0,
      "learning_rate": 1.5406358716660223e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14320
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.3125,
      "epoch": 0.6919694626981059,
      "grad_norm": 0.0,
      "learning_rate": 1.5403942790877464e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14321
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.25,
      "epoch": 0.6920177812137611,
      "grad_norm": 0.0,
      "learning_rate": 1.5401526865094703e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14322
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.5,
      "epoch": 0.6920660997294164,
      "grad_norm": 0.0,
      "learning_rate": 1.5399110939311944e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14323
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.34375,
      "epoch": 0.6921144182450715,
      "grad_norm": 0.0,
      "learning_rate": 1.5396695013529186e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14324
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.875,
      "epoch": 0.6921627367607267,
      "grad_norm": 0.0,
      "learning_rate": 1.5394279087746422e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 14325
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.3125,
      "epoch": 0.6922110552763819,
      "grad_norm": 0.0,
      "learning_rate": 1.5391863161963663e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14326
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.5,
      "epoch": 0.6922593737920371,
      "grad_norm": 0.0,
      "learning_rate": 1.5389447236180902e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14327
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.59375,
      "epoch": 0.6923076923076923,
      "grad_norm": 0.0,
      "learning_rate": 1.5387031310398143e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14328
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.125,
      "epoch": 0.6923560108233475,
      "grad_norm": 0.0,
      "learning_rate": 1.5384615384615385e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14329
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.875,
      "epoch": 0.6924043293390028,
      "grad_norm": 0.0,
      "learning_rate": 1.5382199458832623e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14330
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.125,
      "epoch": 0.6924526478546579,
      "grad_norm": 0.0,
      "learning_rate": 1.5379783533049865e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14331
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.71875,
      "epoch": 0.6925009663703131,
      "grad_norm": 0.0,
      "learning_rate": 1.5377367607267103e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14332
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.46875,
      "epoch": 0.6925492848859683,
      "grad_norm": 0.0,
      "learning_rate": 1.5374951681484345e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14333
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 65.34375,
      "epoch": 0.6925976034016235,
      "grad_norm": 0.0,
      "learning_rate": 1.5372535755701586e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14334
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.5625,
      "epoch": 0.6926459219172787,
      "grad_norm": 0.0,
      "learning_rate": 1.5370119829918822e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14335
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.4375,
      "epoch": 0.6926942404329339,
      "grad_norm": 0.0,
      "learning_rate": 1.5367703904136063e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14336
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.6875,
      "epoch": 0.692742558948589,
      "grad_norm": 0.0,
      "learning_rate": 1.5365287978353302e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14337
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.15625,
      "epoch": 0.6927908774642443,
      "grad_norm": 0.0,
      "learning_rate": 1.5362872052570544e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14338
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 75.625,
      "epoch": 0.6928391959798995,
      "grad_norm": 0.0,
      "learning_rate": 1.5360456126787785e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14339
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.53125,
      "epoch": 0.6928875144955547,
      "grad_norm": 0.0,
      "learning_rate": 1.5358040201005024e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14340
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.5625,
      "epoch": 0.6929358330112099,
      "grad_norm": 0.0,
      "learning_rate": 1.5355624275222265e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14341
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.46875,
      "epoch": 0.6929841515268651,
      "grad_norm": 0.0,
      "learning_rate": 1.5353208349439504e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14342
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.875,
      "epoch": 0.6930324700425203,
      "grad_norm": 0.0,
      "learning_rate": 1.5350792423656745e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14343
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.15625,
      "epoch": 0.6930807885581755,
      "grad_norm": 0.0,
      "learning_rate": 1.5348376497873986e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14344
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.75,
      "epoch": 0.6931291070738307,
      "grad_norm": 0.0,
      "learning_rate": 1.5345960572091225e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14345
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.21875,
      "epoch": 0.6931774255894859,
      "grad_norm": 0.0,
      "learning_rate": 1.5343544646308464e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14346
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.21875,
      "epoch": 0.693225744105141,
      "grad_norm": 0.0,
      "learning_rate": 1.5341128720525703e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14347
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.21875,
      "epoch": 0.6932740626207963,
      "grad_norm": 0.0,
      "learning_rate": 1.5338712794742944e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14348
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.1875,
      "epoch": 0.6933223811364515,
      "grad_norm": 0.0,
      "learning_rate": 1.5336296868960185e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14349
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.0625,
      "epoch": 0.6933706996521067,
      "grad_norm": 0.0,
      "learning_rate": 1.5333880943177424e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14350
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.125,
      "epoch": 0.6934190181677619,
      "grad_norm": 0.0,
      "learning_rate": 1.5331465017394665e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14351
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.3125,
      "epoch": 0.6934673366834171,
      "grad_norm": 0.0,
      "learning_rate": 1.5329049091611904e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14352
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.46875,
      "epoch": 0.6935156551990723,
      "grad_norm": 0.0,
      "learning_rate": 1.5326633165829145e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14353
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.125,
      "epoch": 0.6935639737147274,
      "grad_norm": 0.0,
      "learning_rate": 1.5324217240046387e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14354
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.59375,
      "epoch": 0.6936122922303827,
      "grad_norm": 0.0,
      "learning_rate": 1.5321801314263626e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14355
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.125,
      "epoch": 0.6936606107460379,
      "grad_norm": 0.0,
      "learning_rate": 1.5319385388480864e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14356
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.09375,
      "epoch": 0.693708929261693,
      "grad_norm": 0.0,
      "learning_rate": 1.5316969462698103e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14357
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.78125,
      "epoch": 0.6937572477773483,
      "grad_norm": 0.0,
      "learning_rate": 1.5314553536915344e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14358
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 66.78125,
      "epoch": 0.6938055662930035,
      "grad_norm": 0.0,
      "learning_rate": 1.5312137611132586e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14359
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.875,
      "epoch": 0.6938538848086587,
      "grad_norm": 0.0,
      "learning_rate": 1.5309721685349824e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14360
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.84375,
      "epoch": 0.6939022033243138,
      "grad_norm": 0.0,
      "learning_rate": 1.5307305759567066e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14361
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.125,
      "epoch": 0.6939505218399691,
      "grad_norm": 0.0,
      "learning_rate": 1.5304889833784305e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14362
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.59375,
      "epoch": 0.6939988403556243,
      "grad_norm": 0.0,
      "learning_rate": 1.5302473908001546e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14363
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.0625,
      "epoch": 0.6940471588712794,
      "grad_norm": 0.0,
      "learning_rate": 1.5300057982218787e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14364
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.84375,
      "epoch": 0.6940954773869347,
      "grad_norm": 0.0,
      "learning_rate": 1.5297642056436026e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14365
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.3125,
      "epoch": 0.6941437959025899,
      "grad_norm": 0.0,
      "learning_rate": 1.5295226130653265e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14366
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.34375,
      "epoch": 0.694192114418245,
      "grad_norm": 0.0,
      "learning_rate": 1.5292810204870503e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14367
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.5,
      "epoch": 0.6942404329339003,
      "grad_norm": 0.0,
      "learning_rate": 1.5290394279087745e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14368
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.40625,
      "epoch": 0.6942887514495555,
      "grad_norm": 0.0,
      "learning_rate": 1.5287978353304986e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14369
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.1875,
      "epoch": 0.6943370699652107,
      "grad_norm": 0.0,
      "learning_rate": 1.5285562427522225e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 14370
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.65625,
      "epoch": 0.6943853884808658,
      "grad_norm": 0.0,
      "learning_rate": 1.5283146501739466e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14371
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.8125,
      "epoch": 0.6944337069965211,
      "grad_norm": 0.0,
      "learning_rate": 1.5280730575956705e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14372
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.34375,
      "epoch": 0.6944820255121763,
      "grad_norm": 0.0,
      "learning_rate": 1.5278314650173946e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14373
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.15625,
      "epoch": 0.6945303440278314,
      "grad_norm": 0.0,
      "learning_rate": 1.5275898724391188e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14374
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.125,
      "epoch": 0.6945786625434867,
      "grad_norm": 0.0,
      "learning_rate": 1.5273482798608426e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14375
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.28125,
      "epoch": 0.6946269810591419,
      "grad_norm": 0.0,
      "learning_rate": 1.5271066872825668e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 14376
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.6875,
      "epoch": 0.694675299574797,
      "grad_norm": 0.0,
      "learning_rate": 1.5268650947042904e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14377
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.3125,
      "epoch": 0.6947236180904522,
      "grad_norm": 0.0,
      "learning_rate": 1.5266235021260145e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14378
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.875,
      "epoch": 0.6947719366061075,
      "grad_norm": 0.0,
      "learning_rate": 1.5263819095477386e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14379
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.5625,
      "epoch": 0.6948202551217627,
      "grad_norm": 0.0,
      "learning_rate": 1.5261403169694625e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14380
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.8125,
      "epoch": 0.6948685736374178,
      "grad_norm": 0.0,
      "learning_rate": 1.5258987243911867e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14381
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 84.375,
      "epoch": 0.6949168921530731,
      "grad_norm": 5.849872507272588,
      "learning_rate": 1.5256571318129105e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14382
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.0,
      "epoch": 0.6949652106687283,
      "grad_norm": 0.0,
      "learning_rate": 1.5254155392346347e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14383
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.0625,
      "epoch": 0.6950135291843834,
      "grad_norm": 0.0,
      "learning_rate": 1.5251739466563588e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14384
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.6875,
      "epoch": 0.6950618477000386,
      "grad_norm": 0.0,
      "learning_rate": 1.5249323540780827e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14385
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.4375,
      "epoch": 0.6951101662156939,
      "grad_norm": 0.0,
      "learning_rate": 1.5246907614998068e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14386
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.59375,
      "epoch": 0.695158484731349,
      "grad_norm": 0.0,
      "learning_rate": 1.5244491689215304e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14387
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.5625,
      "epoch": 0.6952068032470042,
      "grad_norm": 0.0,
      "learning_rate": 1.5242075763432546e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14388
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.9375,
      "epoch": 0.6952551217626595,
      "grad_norm": 0.0,
      "learning_rate": 1.5239659837649787e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 14389
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.4375,
      "epoch": 0.6953034402783147,
      "grad_norm": 0.0,
      "learning_rate": 1.5237243911867026e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14390
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.8125,
      "epoch": 0.6953517587939698,
      "grad_norm": 0.0,
      "learning_rate": 1.5234827986084267e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14391
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.21875,
      "epoch": 0.695400077309625,
      "grad_norm": 0.0,
      "learning_rate": 1.5232412060301506e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14392
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.875,
      "epoch": 0.6954483958252803,
      "grad_norm": 0.0,
      "learning_rate": 1.5229996134518747e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14393
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.34375,
      "epoch": 0.6954967143409354,
      "grad_norm": 0.0,
      "learning_rate": 1.5227580208735988e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14394
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.3125,
      "epoch": 0.6955450328565906,
      "grad_norm": 0.0,
      "learning_rate": 1.5225164282953227e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14395
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.65625,
      "epoch": 0.6955933513722459,
      "grad_norm": 0.0,
      "learning_rate": 1.5222748357170468e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14396
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.5,
      "epoch": 0.695641669887901,
      "grad_norm": 0.0,
      "learning_rate": 1.5220332431387705e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14397
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.90625,
      "epoch": 0.6956899884035562,
      "grad_norm": 0.0,
      "learning_rate": 1.5217916505604946e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14398
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 66.5,
      "epoch": 0.6957383069192115,
      "grad_norm": 0.0,
      "learning_rate": 1.5215500579822187e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14399
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.71875,
      "epoch": 0.6957866254348667,
      "grad_norm": 0.0,
      "learning_rate": 1.5213084654039426e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14400
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.6875,
      "epoch": 0.6958349439505218,
      "grad_norm": 0.0,
      "learning_rate": 1.5210668728256667e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14401
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.375,
      "epoch": 0.695883262466177,
      "grad_norm": 0.0,
      "learning_rate": 1.5208252802473906e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14402
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.625,
      "epoch": 0.6959315809818323,
      "grad_norm": 0.0,
      "learning_rate": 1.5205836876691147e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14403
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.1875,
      "epoch": 0.6959798994974874,
      "grad_norm": 0.0,
      "learning_rate": 1.520342095090839e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14404
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.15625,
      "epoch": 0.6960282180131426,
      "grad_norm": 0.0,
      "learning_rate": 1.5201005025125628e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14405
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.9375,
      "epoch": 0.6960765365287979,
      "grad_norm": 0.0,
      "learning_rate": 1.519858909934287e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14406
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.3125,
      "epoch": 0.696124855044453,
      "grad_norm": 0.0,
      "learning_rate": 1.5196173173560105e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14407
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.65625,
      "epoch": 0.6961731735601082,
      "grad_norm": 0.0,
      "learning_rate": 1.5193757247777346e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14408
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.0,
      "epoch": 0.6962214920757634,
      "grad_norm": 0.0,
      "learning_rate": 1.5191341321994588e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14409
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.84375,
      "epoch": 0.6962698105914187,
      "grad_norm": 0.0,
      "learning_rate": 1.5188925396211826e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14410
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.65625,
      "epoch": 0.6963181291070738,
      "grad_norm": 0.0,
      "learning_rate": 1.5186509470429068e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14411
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.40625,
      "epoch": 0.696366447622729,
      "grad_norm": 0.0,
      "learning_rate": 1.5184093544646306e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14412
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.125,
      "epoch": 0.6964147661383843,
      "grad_norm": 0.0,
      "learning_rate": 1.5181677618863548e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14413
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.90625,
      "epoch": 0.6964630846540394,
      "grad_norm": 0.0,
      "learning_rate": 1.517926169308079e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14414
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.375,
      "epoch": 0.6965114031696946,
      "grad_norm": 0.0,
      "learning_rate": 1.5176845767298028e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14415
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.5,
      "epoch": 0.6965597216853499,
      "grad_norm": 0.0,
      "learning_rate": 1.517442984151527e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14416
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.9375,
      "epoch": 0.696608040201005,
      "grad_norm": 0.0,
      "learning_rate": 1.5172013915732505e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14417
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 69.90625,
      "epoch": 0.6966563587166602,
      "grad_norm": 0.0,
      "learning_rate": 1.5169597989949747e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14418
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.875,
      "epoch": 0.6967046772323154,
      "grad_norm": 0.0,
      "learning_rate": 1.5167182064166988e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14419
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.71875,
      "epoch": 0.6967529957479707,
      "grad_norm": 0.0,
      "learning_rate": 1.5164766138384227e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14420
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.96875,
      "epoch": 0.6968013142636258,
      "grad_norm": 0.0,
      "learning_rate": 1.5162350212601468e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14421
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.25,
      "epoch": 0.696849632779281,
      "grad_norm": 0.0,
      "learning_rate": 1.5159934286818707e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14422
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.1875,
      "epoch": 0.6968979512949363,
      "grad_norm": 0.0,
      "learning_rate": 1.5157518361035948e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14423
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.3125,
      "epoch": 0.6969462698105914,
      "grad_norm": 0.0,
      "learning_rate": 1.515510243525319e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 14424
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.03125,
      "epoch": 0.6969945883262466,
      "grad_norm": 0.0,
      "learning_rate": 1.5152686509470428e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14425
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.90625,
      "epoch": 0.6970429068419018,
      "grad_norm": 0.0,
      "learning_rate": 1.515027058368767e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14426
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.75,
      "epoch": 0.697091225357557,
      "grad_norm": 0.0,
      "learning_rate": 1.5147854657904908e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14427
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.15625,
      "epoch": 0.6971395438732122,
      "grad_norm": 0.0,
      "learning_rate": 1.5145438732122147e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14428
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.8125,
      "epoch": 0.6971878623888674,
      "grad_norm": 0.0,
      "learning_rate": 1.5143022806339388e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14429
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.71875,
      "epoch": 0.6972361809045227,
      "grad_norm": 0.0,
      "learning_rate": 1.5140606880556627e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819833278656,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14430
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.09375,
      "epoch": 0.6972844994201778,
      "grad_norm": 0.0,
      "learning_rate": 1.5138190954773869e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 14431
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.9375,
      "epoch": 0.697332817935833,
      "grad_norm": 0.0,
      "learning_rate": 1.5135775028991107e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14432
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.15625,
      "epoch": 0.6973811364514882,
      "grad_norm": 0.0,
      "learning_rate": 1.5133359103208349e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14433
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.09375,
      "epoch": 0.6974294549671434,
      "grad_norm": 0.0,
      "learning_rate": 1.513094317742559e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14434
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.25,
      "epoch": 0.6974777734827986,
      "grad_norm": 0.0,
      "learning_rate": 1.512852725164283e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14435
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.40625,
      "epoch": 0.6975260919984538,
      "grad_norm": 0.0,
      "learning_rate": 1.512611132586007e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14436
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.03125,
      "epoch": 0.697574410514109,
      "grad_norm": 0.0,
      "learning_rate": 1.512369540007731e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14437
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.6875,
      "epoch": 0.6976227290297642,
      "grad_norm": 0.0,
      "learning_rate": 1.5121279474294547e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14438
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.46875,
      "epoch": 0.6976710475454194,
      "grad_norm": 0.0,
      "learning_rate": 1.511886354851179e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14439
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.8125,
      "epoch": 0.6977193660610747,
      "grad_norm": 0.0,
      "learning_rate": 1.5116447622729028e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14440
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.28125,
      "epoch": 0.6977676845767298,
      "grad_norm": 0.0,
      "learning_rate": 1.511403169694627e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14441
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.15625,
      "epoch": 0.697816003092385,
      "grad_norm": 0.0,
      "learning_rate": 1.5111615771163508e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14442
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 83.9375,
      "epoch": 0.6978643216080402,
      "grad_norm": 0.0,
      "learning_rate": 1.510919984538075e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 14443
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.1875,
      "epoch": 0.6979126401236954,
      "grad_norm": 0.0,
      "learning_rate": 1.510678391959799e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 14444
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.5625,
      "epoch": 0.6979609586393506,
      "grad_norm": 0.0,
      "learning_rate": 1.510436799381523e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14445
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 69.1875,
      "epoch": 0.6980092771550058,
      "grad_norm": 0.0,
      "learning_rate": 1.510195206803247e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 14446
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.46875,
      "epoch": 0.698057595670661,
      "grad_norm": 0.0,
      "learning_rate": 1.509953614224971e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14447
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.9375,
      "epoch": 0.6981059141863162,
      "grad_norm": 0.0,
      "learning_rate": 1.509712021646695e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14448
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.0,
      "epoch": 0.6981542327019714,
      "grad_norm": 0.0,
      "learning_rate": 1.509470429068419e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14449
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.5,
      "epoch": 0.6982025512176266,
      "grad_norm": 0.0,
      "learning_rate": 1.5092288364901428e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14450
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.65625,
      "epoch": 0.6982508697332818,
      "grad_norm": 0.0,
      "learning_rate": 1.508987243911867e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14451
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.09375,
      "epoch": 0.698299188248937,
      "grad_norm": 0.0,
      "learning_rate": 1.5087456513335908e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14452
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.8125,
      "epoch": 0.6983475067645922,
      "grad_norm": 0.0,
      "learning_rate": 1.508504058755315e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14453
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.40625,
      "epoch": 0.6983958252802474,
      "grad_norm": 0.0,
      "learning_rate": 1.508262466177039e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14454
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.09375,
      "epoch": 0.6984441437959026,
      "grad_norm": 0.0,
      "learning_rate": 1.508020873598763e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14455
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.5625,
      "epoch": 0.6984924623115578,
      "grad_norm": 0.0,
      "learning_rate": 1.507779281020487e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14456
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.09375,
      "epoch": 0.6985407808272129,
      "grad_norm": 0.0,
      "learning_rate": 1.507537688442211e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14457
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.40625,
      "epoch": 0.6985890993428682,
      "grad_norm": 0.0,
      "learning_rate": 1.507296095863935e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14458
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.53125,
      "epoch": 0.6986374178585234,
      "grad_norm": 0.0,
      "learning_rate": 1.507054503285659e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14459
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.59375,
      "epoch": 0.6986857363741786,
      "grad_norm": 0.0,
      "learning_rate": 1.5068129107073828e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14460
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.4375,
      "epoch": 0.6987340548898338,
      "grad_norm": 0.0,
      "learning_rate": 1.506571318129107e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14461
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.21875,
      "epoch": 0.698782373405489,
      "grad_norm": 0.0,
      "learning_rate": 1.5063297255508308e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14462
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.59375,
      "epoch": 0.6988306919211442,
      "grad_norm": 0.0,
      "learning_rate": 1.506088132972555e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14463
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.125,
      "epoch": 0.6988790104367993,
      "grad_norm": 0.0,
      "learning_rate": 1.505846540394279e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14464
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 79.96875,
      "epoch": 0.6989273289524546,
      "grad_norm": 0.0,
      "learning_rate": 1.505604947816003e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 14465
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.65625,
      "epoch": 0.6989756474681098,
      "grad_norm": 0.0,
      "learning_rate": 1.505363355237727e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14466
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.5,
      "epoch": 0.6990239659837649,
      "grad_norm": 0.0,
      "learning_rate": 1.505121762659451e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14467
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.96875,
      "epoch": 0.6990722844994202,
      "grad_norm": 0.0,
      "learning_rate": 1.504880170081175e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14468
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.40625,
      "epoch": 0.6991206030150754,
      "grad_norm": 0.0,
      "learning_rate": 1.504638577502899e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14469
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.375,
      "epoch": 0.6991689215307306,
      "grad_norm": 0.0,
      "learning_rate": 1.504396984924623e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14470
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.65625,
      "epoch": 0.6992172400463857,
      "grad_norm": 0.0,
      "learning_rate": 1.504155392346347e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14471
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.625,
      "epoch": 0.699265558562041,
      "grad_norm": 0.0,
      "learning_rate": 1.503913799768071e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14472
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.96875,
      "epoch": 0.6993138770776962,
      "grad_norm": 0.0,
      "learning_rate": 1.503672207189795e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14473
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.8125,
      "epoch": 0.6993621955933513,
      "grad_norm": 0.0,
      "learning_rate": 1.5034306146115192e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14474
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.96875,
      "epoch": 0.6994105141090066,
      "grad_norm": 0.0,
      "learning_rate": 1.503189022033243e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14475
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.375,
      "epoch": 0.6994588326246618,
      "grad_norm": 0.0,
      "learning_rate": 1.5029474294549672e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14476
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.25,
      "epoch": 0.6995071511403169,
      "grad_norm": 0.0,
      "learning_rate": 1.502705836876691e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14477
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.90625,
      "epoch": 0.6995554696559722,
      "grad_norm": 0.0,
      "learning_rate": 1.5024642442984152e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14478
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.59375,
      "epoch": 0.6996037881716274,
      "grad_norm": 0.0,
      "learning_rate": 1.5022226517201393e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14479
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 71.59375,
      "epoch": 0.6996521066872826,
      "grad_norm": 0.0,
      "learning_rate": 1.501981059141863e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14480
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.125,
      "epoch": 0.6997004252029377,
      "grad_norm": 0.0,
      "learning_rate": 1.501739466563587e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14481
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.1875,
      "epoch": 0.699748743718593,
      "grad_norm": 0.0,
      "learning_rate": 1.501497873985311e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14482
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.71875,
      "epoch": 0.6997970622342482,
      "grad_norm": 0.0,
      "learning_rate": 1.501256281407035e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14483
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.75,
      "epoch": 0.6998453807499033,
      "grad_norm": 0.0,
      "learning_rate": 1.5010146888287592e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14484
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.75,
      "epoch": 0.6998936992655586,
      "grad_norm": 0.0,
      "learning_rate": 1.500773096250483e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14485
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.625,
      "epoch": 0.6999420177812138,
      "grad_norm": 0.0,
      "learning_rate": 1.5005315036722072e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14486
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 74.96875,
      "epoch": 0.6999903362968689,
      "grad_norm": 18.397694732670335,
      "learning_rate": 1.500289911093931e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.2472364753484726,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14487
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.125,
      "epoch": 0.7000386548125241,
      "grad_norm": 0.0,
      "learning_rate": 1.5000483185156552e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14488
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.5625,
      "epoch": 0.7000869733281794,
      "grad_norm": 0.0,
      "learning_rate": 1.4998067259373793e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14489
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.0,
      "epoch": 0.7001352918438346,
      "grad_norm": 0.0,
      "learning_rate": 1.499565133359103e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14490
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.375,
      "epoch": 0.7001836103594897,
      "grad_norm": 0.0,
      "learning_rate": 1.499323540780827e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14491
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.09375,
      "epoch": 0.700231928875145,
      "grad_norm": 0.0,
      "learning_rate": 1.499081948202551e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14492
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.78125,
      "epoch": 0.7002802473908002,
      "grad_norm": 0.0,
      "learning_rate": 1.498840355624275e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14493
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.34375,
      "epoch": 0.7003285659064553,
      "grad_norm": 0.0,
      "learning_rate": 1.4985987630459992e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14494
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.15625,
      "epoch": 0.7003768844221105,
      "grad_norm": 0.0,
      "learning_rate": 1.498357170467723e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14495
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.9375,
      "epoch": 0.7004252029377658,
      "grad_norm": 0.0,
      "learning_rate": 1.4981155778894472e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14496
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.90625,
      "epoch": 0.7004735214534209,
      "grad_norm": 0.0,
      "learning_rate": 1.497873985311171e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14497
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 68.4375,
      "epoch": 0.7005218399690761,
      "grad_norm": 0.0,
      "learning_rate": 1.4976323927328953e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14498
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.1875,
      "epoch": 0.7005701584847314,
      "grad_norm": 0.0,
      "learning_rate": 1.4973908001546194e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14499
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.40625,
      "epoch": 0.7006184770003866,
      "grad_norm": 0.0,
      "learning_rate": 1.497149207576343e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14500
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.4375,
      "epoch": 0.7006667955160417,
      "grad_norm": 0.0,
      "learning_rate": 1.496907614998067e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14501
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.4375,
      "epoch": 0.700715114031697,
      "grad_norm": 0.0,
      "learning_rate": 1.496666022419791e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 14502
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.0,
      "epoch": 0.7007634325473522,
      "grad_norm": 0.0,
      "learning_rate": 1.4964244298415151e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14503
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.3125,
      "epoch": 0.7008117510630073,
      "grad_norm": 0.0,
      "learning_rate": 1.4961828372632393e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14504
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.28125,
      "epoch": 0.7008600695786625,
      "grad_norm": 0.0,
      "learning_rate": 1.4959412446849631e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14505
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.125,
      "epoch": 0.7009083880943178,
      "grad_norm": 0.0,
      "learning_rate": 1.4956996521066873e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14506
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.78125,
      "epoch": 0.7009567066099729,
      "grad_norm": 0.0,
      "learning_rate": 1.4954580595284112e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14507
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.8125,
      "epoch": 0.7010050251256281,
      "grad_norm": 0.0,
      "learning_rate": 1.4952164669501353e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14508
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.53125,
      "epoch": 0.7010533436412834,
      "grad_norm": 0.0,
      "learning_rate": 1.4949748743718594e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14509
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.0625,
      "epoch": 0.7011016621569386,
      "grad_norm": 0.0,
      "learning_rate": 1.494733281793583e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14510
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.78125,
      "epoch": 0.7011499806725937,
      "grad_norm": 0.0,
      "learning_rate": 1.4944916892153072e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14511
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.5,
      "epoch": 0.7011982991882489,
      "grad_norm": 0.0,
      "learning_rate": 1.494250096637031e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14512
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.21875,
      "epoch": 0.7012466177039042,
      "grad_norm": 0.0,
      "learning_rate": 1.4940085040587552e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14513
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.40625,
      "epoch": 0.7012949362195593,
      "grad_norm": 0.0,
      "learning_rate": 1.4937669114804793e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14514
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.5,
      "epoch": 0.7013432547352145,
      "grad_norm": 0.0,
      "learning_rate": 1.4935253189022032e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14515
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.21875,
      "epoch": 0.7013915732508698,
      "grad_norm": 0.0,
      "learning_rate": 1.4932837263239273e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14516
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.65625,
      "epoch": 0.7014398917665249,
      "grad_norm": 0.0,
      "learning_rate": 1.4930421337456512e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14517
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.09375,
      "epoch": 0.7014882102821801,
      "grad_norm": 0.0,
      "learning_rate": 1.4928005411673753e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14518
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.78125,
      "epoch": 0.7015365287978353,
      "grad_norm": 0.0,
      "learning_rate": 1.4925589485890995e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14519
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.625,
      "epoch": 0.7015848473134906,
      "grad_norm": 0.0,
      "learning_rate": 1.492317356010823e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14520
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.40625,
      "epoch": 0.7016331658291457,
      "grad_norm": 0.0,
      "learning_rate": 1.4920757634325472e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14521
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.59375,
      "epoch": 0.7016814843448009,
      "grad_norm": 0.0,
      "learning_rate": 1.491834170854271e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14522
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.25,
      "epoch": 0.7017298028604562,
      "grad_norm": 0.0,
      "learning_rate": 1.4915925782759952e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14523
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.53125,
      "epoch": 0.7017781213761113,
      "grad_norm": 0.0,
      "learning_rate": 1.4913509856977194e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14524
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.125,
      "epoch": 0.7018264398917665,
      "grad_norm": 0.0,
      "learning_rate": 1.4911093931194432e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14525
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.03125,
      "epoch": 0.7018747584074218,
      "grad_norm": 0.0,
      "learning_rate": 1.4908678005411674e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14526
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.125,
      "epoch": 0.7019230769230769,
      "grad_norm": 0.0,
      "learning_rate": 1.4906262079628912e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14527
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.59375,
      "epoch": 0.7019713954387321,
      "grad_norm": 0.0,
      "learning_rate": 1.4903846153846154e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14528
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.59375,
      "epoch": 0.7020197139543873,
      "grad_norm": 0.0,
      "learning_rate": 1.4901430228063395e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14529
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.59375,
      "epoch": 0.7020680324700426,
      "grad_norm": 0.0,
      "learning_rate": 1.4899014302280634e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14530
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.40625,
      "epoch": 0.7021163509856977,
      "grad_norm": 0.0,
      "learning_rate": 1.4896598376497872e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14531
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.96875,
      "epoch": 0.7021646695013529,
      "grad_norm": 0.0,
      "learning_rate": 1.489418245071511e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14532
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.125,
      "epoch": 0.7022129880170082,
      "grad_norm": 0.0,
      "learning_rate": 1.4891766524932353e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14533
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.09375,
      "epoch": 0.7022613065326633,
      "grad_norm": 0.0,
      "learning_rate": 1.4889350599149594e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14534
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.84375,
      "epoch": 0.7023096250483185,
      "grad_norm": 0.0,
      "learning_rate": 1.4886934673366833e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14535
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.5625,
      "epoch": 0.7023579435639737,
      "grad_norm": 0.0,
      "learning_rate": 1.4884518747584074e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14536
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.875,
      "epoch": 0.7024062620796289,
      "grad_norm": 0.0,
      "learning_rate": 1.4882102821801313e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 14537
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.09375,
      "epoch": 0.7024545805952841,
      "grad_norm": 0.0,
      "learning_rate": 1.4879686896018554e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14538
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.71875,
      "epoch": 0.7025028991109393,
      "grad_norm": 0.0,
      "learning_rate": 1.4877270970235795e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14539
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.46875,
      "epoch": 0.7025512176265946,
      "grad_norm": 0.0,
      "learning_rate": 1.4874855044453034e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14540
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.1875,
      "epoch": 0.7025995361422497,
      "grad_norm": 0.0,
      "learning_rate": 1.4872439118670273e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14541
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.75,
      "epoch": 0.7026478546579049,
      "grad_norm": 0.0,
      "learning_rate": 1.4870023192887512e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14542
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.71875,
      "epoch": 0.7026961731735601,
      "grad_norm": 0.0,
      "learning_rate": 1.4867607267104753e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14543
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.28125,
      "epoch": 0.7027444916892153,
      "grad_norm": 0.0,
      "learning_rate": 1.4865191341321994e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14544
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.96875,
      "epoch": 0.7027928102048705,
      "grad_norm": 0.0,
      "learning_rate": 1.4862775415539233e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14545
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.0625,
      "epoch": 0.7028411287205257,
      "grad_norm": 0.0,
      "learning_rate": 1.4860359489756474e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 14546
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.0625,
      "epoch": 0.7028894472361809,
      "grad_norm": 0.0,
      "learning_rate": 1.4857943563973713e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2519446909427643,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14547
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.96875,
      "epoch": 0.7029377657518361,
      "grad_norm": 0.0,
      "learning_rate": 1.4855527638190954e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14548
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.46875,
      "epoch": 0.7029860842674913,
      "grad_norm": 0.0,
      "learning_rate": 1.4853111712408196e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14549
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.78125,
      "epoch": 0.7030344027831466,
      "grad_norm": 0.0,
      "learning_rate": 1.4850695786625435e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14550
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.3125,
      "epoch": 0.7030827212988017,
      "grad_norm": 0.0,
      "learning_rate": 1.4848279860842676e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23568855226039886,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14551
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.75,
      "epoch": 0.7031310398144569,
      "grad_norm": 0.0,
      "learning_rate": 1.4845863935059912e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14552
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.125,
      "epoch": 0.7031793583301121,
      "grad_norm": 13.805843440754506,
      "learning_rate": 1.4843448009277153e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14553
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.28125,
      "epoch": 0.7032276768457673,
      "grad_norm": 0.0,
      "learning_rate": 1.4841032083494395e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14554
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.5625,
      "epoch": 0.7032759953614225,
      "grad_norm": 0.0,
      "learning_rate": 1.4838616157711633e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14555
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.875,
      "epoch": 0.7033243138770777,
      "grad_norm": 0.0,
      "learning_rate": 1.4836200231928875e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14556
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.96875,
      "epoch": 0.7033726323927328,
      "grad_norm": 0.0,
      "learning_rate": 1.4833784306146113e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14557
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.71875,
      "epoch": 0.7034209509083881,
      "grad_norm": 0.0,
      "learning_rate": 1.4831368380363355e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14558
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.71875,
      "epoch": 0.7034692694240433,
      "grad_norm": 0.0,
      "learning_rate": 1.4828952454580596e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14559
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.9375,
      "epoch": 0.7035175879396985,
      "grad_norm": 0.0,
      "learning_rate": 1.4826536528797835e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14560
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.84375,
      "epoch": 0.7035659064553537,
      "grad_norm": 0.0,
      "learning_rate": 1.4824120603015076e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14561
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.375,
      "epoch": 0.7036142249710089,
      "grad_norm": 0.0,
      "learning_rate": 1.4821704677232312e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14562
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.40625,
      "epoch": 0.7036625434866641,
      "grad_norm": 0.0,
      "learning_rate": 1.4819288751449554e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14563
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.5,
      "epoch": 0.7037108620023192,
      "grad_norm": 0.0,
      "learning_rate": 1.4816872825666795e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14564
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.8125,
      "epoch": 0.7037591805179745,
      "grad_norm": 0.0,
      "learning_rate": 1.4814456899884034e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14565
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.65625,
      "epoch": 0.7038074990336297,
      "grad_norm": 0.0,
      "learning_rate": 1.4812040974101275e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14566
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.5,
      "epoch": 0.7038558175492848,
      "grad_norm": 0.0,
      "learning_rate": 1.4809625048318514e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14567
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.5625,
      "epoch": 0.7039041360649401,
      "grad_norm": 0.0,
      "learning_rate": 1.4807209122535755e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14568
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.3125,
      "epoch": 0.7039524545805953,
      "grad_norm": 0.0,
      "learning_rate": 1.4804793196752997e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14569
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.6875,
      "epoch": 0.7040007730962505,
      "grad_norm": 0.0,
      "learning_rate": 1.4802377270970235e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 14570
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.5625,
      "epoch": 0.7040490916119057,
      "grad_norm": 0.0,
      "learning_rate": 1.4799961345187477e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23356688022613525,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14571
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.25,
      "epoch": 0.7040974101275609,
      "grad_norm": 0.0,
      "learning_rate": 1.4797545419404713e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14572
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.0625,
      "epoch": 0.7041457286432161,
      "grad_norm": 0.0,
      "learning_rate": 1.4795129493621954e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14573
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.84375,
      "epoch": 0.7041940471588712,
      "grad_norm": 0.0,
      "learning_rate": 1.4792713567839195e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14574
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.15625,
      "epoch": 0.7042423656745265,
      "grad_norm": 0.0,
      "learning_rate": 1.4790297642056434e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14575
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.09375,
      "epoch": 0.7042906841901817,
      "grad_norm": 0.0,
      "learning_rate": 1.4787881716273676e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14576
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.1875,
      "epoch": 0.7043390027058368,
      "grad_norm": 0.0,
      "learning_rate": 1.4785465790490914e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14577
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.25,
      "epoch": 0.7043873212214921,
      "grad_norm": 0.0,
      "learning_rate": 1.4783049864708156e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 14578
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.0,
      "epoch": 0.7044356397371473,
      "grad_norm": 0.0,
      "learning_rate": 1.4780633938925397e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14579
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.84375,
      "epoch": 0.7044839582528025,
      "grad_norm": 0.0,
      "learning_rate": 1.4778218013142636e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14580
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.96875,
      "epoch": 0.7045322767684576,
      "grad_norm": 0.0,
      "learning_rate": 1.4775802087359877e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14581
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.65625,
      "epoch": 0.7045805952841129,
      "grad_norm": 0.0,
      "learning_rate": 1.4773386161577113e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14582
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.6875,
      "epoch": 0.7046289137997681,
      "grad_norm": 0.0,
      "learning_rate": 1.4770970235794355e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14583
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.4375,
      "epoch": 0.7046772323154232,
      "grad_norm": 0.0,
      "learning_rate": 1.4768554310011596e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2630178928375244,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14584
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.6875,
      "epoch": 0.7047255508310785,
      "grad_norm": 0.0,
      "learning_rate": 1.4766138384228835e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14585
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.6875,
      "epoch": 0.7047738693467337,
      "grad_norm": 0.0,
      "learning_rate": 1.4763722458446076e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2472364604473114,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14586
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 68.5,
      "epoch": 0.7048221878623888,
      "grad_norm": 0.0,
      "learning_rate": 1.4761306532663315e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 14587
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.03125,
      "epoch": 0.704870506378044,
      "grad_norm": 0.0,
      "learning_rate": 1.4758890606880556e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14588
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.59375,
      "epoch": 0.7049188248936993,
      "grad_norm": 0.0,
      "learning_rate": 1.4756474681097797e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14589
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.75,
      "epoch": 0.7049671434093545,
      "grad_norm": 0.0,
      "learning_rate": 1.4754058755315036e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14590
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.4375,
      "epoch": 0.7050154619250096,
      "grad_norm": 0.0,
      "learning_rate": 1.4751642829532277e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14591
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 66.15625,
      "epoch": 0.7050637804406649,
      "grad_norm": 0.0,
      "learning_rate": 1.4749226903749514e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14592
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.28125,
      "epoch": 0.7051120989563201,
      "grad_norm": 0.0,
      "learning_rate": 1.4746810977966755e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14593
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 83.15625,
      "epoch": 0.7051604174719752,
      "grad_norm": 52.977765093129825,
      "learning_rate": 1.4744395052183996e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.2529362142086029,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14594
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.0,
      "epoch": 0.7052087359876305,
      "grad_norm": 0.0,
      "learning_rate": 1.4741979126401235e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14595
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.625,
      "epoch": 0.7052570545032857,
      "grad_norm": 0.0,
      "learning_rate": 1.4739563200618476e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14596
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.9375,
      "epoch": 0.7053053730189408,
      "grad_norm": 0.0,
      "learning_rate": 1.4737147274835715e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14597
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.53125,
      "epoch": 0.705353691534596,
      "grad_norm": 0.0,
      "learning_rate": 1.4734731349052956e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14598
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 68.75,
      "epoch": 0.7054020100502513,
      "grad_norm": 0.0,
      "learning_rate": 1.4732315423270198e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14599
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.4375,
      "epoch": 0.7054503285659065,
      "grad_norm": 0.0,
      "learning_rate": 1.4729899497487437e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.23827511072158813,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14600
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.8125,
      "epoch": 0.7054986470815616,
      "grad_norm": 0.0,
      "learning_rate": 1.4727483571704678e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14601
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.6875,
      "epoch": 0.7055469655972169,
      "grad_norm": 0.0,
      "learning_rate": 1.4725067645921917e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14602
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.125,
      "epoch": 0.7055952841128721,
      "grad_norm": 0.0,
      "learning_rate": 1.4722651720139155e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14603
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 64.6875,
      "epoch": 0.7056436026285272,
      "grad_norm": 0.0,
      "learning_rate": 1.4720235794356397e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14604
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.75,
      "epoch": 0.7056919211441824,
      "grad_norm": 0.0,
      "learning_rate": 1.4717819868573635e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14605
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.40625,
      "epoch": 0.7057402396598377,
      "grad_norm": 0.0,
      "learning_rate": 1.4715403942790877e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14606
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.90625,
      "epoch": 0.7057885581754928,
      "grad_norm": 0.0,
      "learning_rate": 1.4712988017008115e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14607
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.03125,
      "epoch": 0.705836876691148,
      "grad_norm": 0.0,
      "learning_rate": 1.4710572091225357e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14608
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.0,
      "epoch": 0.7058851952068033,
      "grad_norm": 0.0,
      "learning_rate": 1.4708156165442598e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14609
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.78125,
      "epoch": 0.7059335137224585,
      "grad_norm": 0.0,
      "learning_rate": 1.4705740239659837e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14610
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.28125,
      "epoch": 0.7059818322381136,
      "grad_norm": 0.0,
      "learning_rate": 1.4703324313877078e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14611
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.875,
      "epoch": 0.7060301507537688,
      "grad_norm": 0.0,
      "learning_rate": 1.4700908388094317e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14612
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.21875,
      "epoch": 0.7060784692694241,
      "grad_norm": 0.0,
      "learning_rate": 1.4698492462311556e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14613
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.625,
      "epoch": 0.7061267877850792,
      "grad_norm": 0.0,
      "learning_rate": 1.4696076536528797e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14614
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.3125,
      "epoch": 0.7061751063007344,
      "grad_norm": 0.0,
      "learning_rate": 1.4693660610746036e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14615
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.53125,
      "epoch": 0.7062234248163897,
      "grad_norm": 0.0,
      "learning_rate": 1.4691244684963277e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14616
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.9375,
      "epoch": 0.7062717433320448,
      "grad_norm": 0.0,
      "learning_rate": 1.4688828759180516e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14617
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 74.96875,
      "epoch": 0.7063200618477,
      "grad_norm": 0.0,
      "learning_rate": 1.4686412833397757e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14618
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.1875,
      "epoch": 0.7063683803633553,
      "grad_norm": 0.0,
      "learning_rate": 1.4683996907614999e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14619
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.96875,
      "epoch": 0.7064166988790105,
      "grad_norm": 0.0,
      "learning_rate": 1.4681580981832237e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14620
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 65.6875,
      "epoch": 0.7064650173946656,
      "grad_norm": 0.0,
      "learning_rate": 1.4679165056049479e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14621
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.34375,
      "epoch": 0.7065133359103208,
      "grad_norm": 0.0,
      "learning_rate": 1.4676749130266717e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14622
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.875,
      "epoch": 0.7065616544259761,
      "grad_norm": 0.0,
      "learning_rate": 1.4674333204483956e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14623
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.5625,
      "epoch": 0.7066099729416312,
      "grad_norm": 0.0,
      "learning_rate": 1.4671917278701197e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14624
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.125,
      "epoch": 0.7066582914572864,
      "grad_norm": 0.0,
      "learning_rate": 1.4669501352918436e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14625
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 63.4375,
      "epoch": 0.7067066099729417,
      "grad_norm": 0.0,
      "learning_rate": 1.4667085427135678e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14626
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.15625,
      "epoch": 0.7067549284885968,
      "grad_norm": 0.0,
      "learning_rate": 1.4664669501352916e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14627
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.46875,
      "epoch": 0.706803247004252,
      "grad_norm": 16.794919178065907,
      "learning_rate": 1.4662253575570158e-07,
      "loss": -0.0,
      "reward": 1.859375,
      "reward_std": 0.29713037610054016,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14628
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.625,
      "epoch": 0.7068515655199072,
      "grad_norm": 0.0,
      "learning_rate": 1.46598376497874e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14629
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.0625,
      "epoch": 0.7068998840355625,
      "grad_norm": 0.0,
      "learning_rate": 1.4657421724004638e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14630
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.90625,
      "epoch": 0.7069482025512176,
      "grad_norm": 0.0,
      "learning_rate": 1.465500579822188e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14631
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.53125,
      "epoch": 0.7069965210668728,
      "grad_norm": 0.0,
      "learning_rate": 1.4652589872439118e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14632
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.5625,
      "epoch": 0.7070448395825281,
      "grad_norm": 0.0,
      "learning_rate": 1.465017394665636e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14633
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 81.71875,
      "epoch": 0.7070931580981832,
      "grad_norm": 23.53966235009113,
      "learning_rate": 1.4647758020873598e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.468485027551651,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14634
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.9375,
      "epoch": 0.7071414766138384,
      "grad_norm": 0.0,
      "learning_rate": 1.4645342095090837e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14635
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 71.75,
      "epoch": 0.7071897951294936,
      "grad_norm": 0.0,
      "learning_rate": 1.4642926169308078e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14636
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 69.375,
      "epoch": 0.7072381136451488,
      "grad_norm": 0.0,
      "learning_rate": 1.4640510243525317e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 14637
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.125,
      "epoch": 0.707286432160804,
      "grad_norm": 0.0,
      "learning_rate": 1.4638094317742558e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14638
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.09375,
      "epoch": 0.7073347506764592,
      "grad_norm": 0.0,
      "learning_rate": 1.46356783919598e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14639
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.8125,
      "epoch": 0.7073830691921145,
      "grad_norm": 0.0,
      "learning_rate": 1.4633262466177038e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14640
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.90625,
      "epoch": 0.7074313877077696,
      "grad_norm": 0.0,
      "learning_rate": 1.463084654039428e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14641
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 71.0,
      "epoch": 0.7074797062234248,
      "grad_norm": 39.62010440509934,
      "learning_rate": 1.4628430614611518e-07,
      "loss": -0.0,
      "reward": 2.328125,
      "reward_std": 0.28460076451301575,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14642
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.84375,
      "epoch": 0.70752802473908,
      "grad_norm": 0.0,
      "learning_rate": 1.462601468882876e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14643
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.375,
      "epoch": 0.7075763432547352,
      "grad_norm": 0.0,
      "learning_rate": 1.4623598763045998e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14644
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.28125,
      "epoch": 0.7076246617703904,
      "grad_norm": 0.0,
      "learning_rate": 1.4621182837263237e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14645
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.9375,
      "epoch": 0.7076729802860456,
      "grad_norm": 0.0,
      "learning_rate": 1.4618766911480478e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14646
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.78125,
      "epoch": 0.7077212988017008,
      "grad_norm": 0.0,
      "learning_rate": 1.4616350985697717e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14647
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.25,
      "epoch": 0.707769617317356,
      "grad_norm": 0.0,
      "learning_rate": 1.4613935059914958e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14648
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.1875,
      "epoch": 0.7078179358330112,
      "grad_norm": 0.0,
      "learning_rate": 1.46115191341322e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14649
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.1875,
      "epoch": 0.7078662543486665,
      "grad_norm": 0.0,
      "learning_rate": 1.4609103208349438e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14650
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.0625,
      "epoch": 0.7079145728643216,
      "grad_norm": 0.0,
      "learning_rate": 1.460668728256668e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14651
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.875,
      "epoch": 0.7079628913799768,
      "grad_norm": 0.0,
      "learning_rate": 1.4604271356783919e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14652
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.90625,
      "epoch": 0.708011209895632,
      "grad_norm": 0.0,
      "learning_rate": 1.460185543100116e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14653
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.4375,
      "epoch": 0.7080595284112872,
      "grad_norm": 0.0,
      "learning_rate": 1.4599439505218399e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14654
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 74.90625,
      "epoch": 0.7081078469269424,
      "grad_norm": 23.406875929291342,
      "learning_rate": 1.4597023579435637e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.26409146189689636,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14655
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.59375,
      "epoch": 0.7081561654425976,
      "grad_norm": 0.0,
      "learning_rate": 1.459460765365288e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14656
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.5625,
      "epoch": 0.7082044839582528,
      "grad_norm": 0.0,
      "learning_rate": 1.4592191727870117e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14657
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.0625,
      "epoch": 0.708252802473908,
      "grad_norm": 0.0,
      "learning_rate": 1.458977580208736e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14658
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.5625,
      "epoch": 0.7083011209895632,
      "grad_norm": 0.0,
      "learning_rate": 1.45873598763046e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14659
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.59375,
      "epoch": 0.7083494395052184,
      "grad_norm": 0.0,
      "learning_rate": 1.458494395052184e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2540663480758667,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14660
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.0,
      "epoch": 0.7083977580208736,
      "grad_norm": 0.0,
      "learning_rate": 1.458252802473908e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14661
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 79.8125,
      "epoch": 0.7084460765365288,
      "grad_norm": 9.661498905903425,
      "learning_rate": 1.458011209895632e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.31300365924835205,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14662
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.21875,
      "epoch": 0.708494395052184,
      "grad_norm": 0.0,
      "learning_rate": 1.457769617317356e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14663
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.6875,
      "epoch": 0.7085427135678392,
      "grad_norm": 0.0,
      "learning_rate": 1.4575280247390802e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14664
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.375,
      "epoch": 0.7085910320834944,
      "grad_norm": 0.0,
      "learning_rate": 1.4572864321608038e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14665
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.6875,
      "epoch": 0.7086393505991496,
      "grad_norm": 0.0,
      "learning_rate": 1.457044839582528e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14666
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.78125,
      "epoch": 0.7086876691148047,
      "grad_norm": 0.0,
      "learning_rate": 1.4568032470042518e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14667
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.09375,
      "epoch": 0.70873598763046,
      "grad_norm": 0.0,
      "learning_rate": 1.456561654425976e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14668
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.34375,
      "epoch": 0.7087843061461152,
      "grad_norm": 0.0,
      "learning_rate": 1.4563200618477e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14669
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.0,
      "epoch": 0.7088326246617704,
      "grad_norm": 0.0,
      "learning_rate": 1.456078469269424e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14670
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.53125,
      "epoch": 0.7088809431774256,
      "grad_norm": 0.0,
      "learning_rate": 1.455836876691148e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 14671
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.75,
      "epoch": 0.7089292616930808,
      "grad_norm": 0.0,
      "learning_rate": 1.455595284112872e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 14672
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.46875,
      "epoch": 0.708977580208736,
      "grad_norm": 0.0,
      "learning_rate": 1.455353691534596e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14673
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.125,
      "epoch": 0.7090258987243911,
      "grad_norm": 0.0,
      "learning_rate": 1.4551120989563202e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14674
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.125,
      "epoch": 0.7090742172400464,
      "grad_norm": 0.0,
      "learning_rate": 1.4548705063780438e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14675
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.0,
      "epoch": 0.7091225357557016,
      "grad_norm": 0.0,
      "learning_rate": 1.454628913799768e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14676
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.28125,
      "epoch": 0.7091708542713567,
      "grad_norm": 0.0,
      "learning_rate": 1.4543873212214918e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14677
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.3125,
      "epoch": 0.709219172787012,
      "grad_norm": 0.0,
      "learning_rate": 1.454145728643216e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14678
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.15625,
      "epoch": 0.7092674913026672,
      "grad_norm": 0.0,
      "learning_rate": 1.45390413606494e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14679
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.09375,
      "epoch": 0.7093158098183224,
      "grad_norm": 0.0,
      "learning_rate": 1.453662543486664e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 14680
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.75,
      "epoch": 0.7093641283339776,
      "grad_norm": 0.0,
      "learning_rate": 1.453420950908388e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14681
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.78125,
      "epoch": 0.7094124468496328,
      "grad_norm": 0.0,
      "learning_rate": 1.453179358330112e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14682
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.65625,
      "epoch": 0.709460765365288,
      "grad_norm": 0.0,
      "learning_rate": 1.452937765751836e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14683
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.25,
      "epoch": 0.7095090838809431,
      "grad_norm": 0.0,
      "learning_rate": 1.4526961731735602e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 14684
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.09375,
      "epoch": 0.7095574023965984,
      "grad_norm": 0.0,
      "learning_rate": 1.4524545805952839e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14685
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.65625,
      "epoch": 0.7096057209122536,
      "grad_norm": 0.0,
      "learning_rate": 1.452212988017008e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14686
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.15625,
      "epoch": 0.7096540394279087,
      "grad_norm": 0.0,
      "learning_rate": 1.4519713954387319e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14687
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.25,
      "epoch": 0.709702357943564,
      "grad_norm": 0.0,
      "learning_rate": 1.451729802860456e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14688
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.0,
      "epoch": 0.7097506764592192,
      "grad_norm": 0.0,
      "learning_rate": 1.4514882102821801e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14689
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.1875,
      "epoch": 0.7097989949748744,
      "grad_norm": 0.0,
      "learning_rate": 1.451246617703904e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14690
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.28125,
      "epoch": 0.7098473134905295,
      "grad_norm": 0.0,
      "learning_rate": 1.4510050251256281e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14691
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 67.1875,
      "epoch": 0.7098956320061848,
      "grad_norm": 0.0,
      "learning_rate": 1.450763432547352e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14692
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.375,
      "epoch": 0.70994395052184,
      "grad_norm": 0.0,
      "learning_rate": 1.4505218399690761e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14693
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.6875,
      "epoch": 0.7099922690374951,
      "grad_norm": 0.0,
      "learning_rate": 1.4502802473908003e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14694
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.125,
      "epoch": 0.7100405875531504,
      "grad_norm": 0.0,
      "learning_rate": 1.450038654812524e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14695
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.625,
      "epoch": 0.7100889060688056,
      "grad_norm": 0.0,
      "learning_rate": 1.449797062234248e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14696
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.78125,
      "epoch": 0.7101372245844607,
      "grad_norm": 0.0,
      "learning_rate": 1.449555469655972e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14697
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.34375,
      "epoch": 0.710185543100116,
      "grad_norm": 0.0,
      "learning_rate": 1.449313877077696e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14698
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.0,
      "epoch": 0.7102338616157712,
      "grad_norm": 0.0,
      "learning_rate": 1.4490722844994202e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14699
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.28125,
      "epoch": 0.7102821801314264,
      "grad_norm": 0.0,
      "learning_rate": 1.448830691921144e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 14700
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.3125,
      "epoch": 0.7103304986470815,
      "grad_norm": 0.0,
      "learning_rate": 1.4485890993428682e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 14701
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.40625,
      "epoch": 0.7103788171627368,
      "grad_norm": 0.0,
      "learning_rate": 1.448347506764592e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14702
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.71875,
      "epoch": 0.710427135678392,
      "grad_norm": 0.0,
      "learning_rate": 1.4481059141863162e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14703
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.59375,
      "epoch": 0.7104754541940471,
      "grad_norm": 0.0,
      "learning_rate": 1.4478643216080403e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14704
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.875,
      "epoch": 0.7105237727097024,
      "grad_norm": 0.0,
      "learning_rate": 1.4476227290297642e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14705
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.59375,
      "epoch": 0.7105720912253576,
      "grad_norm": 0.0,
      "learning_rate": 1.447381136451488e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14706
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 70.4375,
      "epoch": 0.7106204097410127,
      "grad_norm": 0.0,
      "learning_rate": 1.447139543873212e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14707
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.84375,
      "epoch": 0.7106687282566679,
      "grad_norm": 0.0,
      "learning_rate": 1.446897951294936e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14708
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 74.375,
      "epoch": 0.7107170467723232,
      "grad_norm": 0.0,
      "learning_rate": 1.4466563587166602e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14709
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.9375,
      "epoch": 0.7107653652879784,
      "grad_norm": 0.0,
      "learning_rate": 1.446414766138384e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14710
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 91.875,
      "epoch": 0.7108136838036335,
      "grad_norm": 0.0,
      "learning_rate": 1.4461731735601082e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 14711
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.6875,
      "epoch": 0.7108620023192888,
      "grad_norm": 0.0,
      "learning_rate": 1.445931580981832e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 14712
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 76.0625,
      "epoch": 0.710910320834944,
      "grad_norm": 7.0564871903283235,
      "learning_rate": 1.4456899884035562e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14713
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.375,
      "epoch": 0.7109586393505991,
      "grad_norm": 0.0,
      "learning_rate": 1.4454483958252804e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14714
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.375,
      "epoch": 0.7110069578662543,
      "grad_norm": 0.0,
      "learning_rate": 1.4452068032470042e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.2109457552433014,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14715
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.375,
      "epoch": 0.7110552763819096,
      "grad_norm": 0.0,
      "learning_rate": 1.444965210668728e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14716
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.1875,
      "epoch": 0.7111035948975647,
      "grad_norm": 0.0,
      "learning_rate": 1.444723618090452e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14717
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.75,
      "epoch": 0.7111519134132199,
      "grad_norm": 0.0,
      "learning_rate": 1.444482025512176e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14718
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.34375,
      "epoch": 0.7112002319288752,
      "grad_norm": 0.0,
      "learning_rate": 1.4442404329339003e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14719
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.28125,
      "epoch": 0.7112485504445304,
      "grad_norm": 0.0,
      "learning_rate": 1.443998840355624e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14720
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.28125,
      "epoch": 0.7112968689601855,
      "grad_norm": 0.0,
      "learning_rate": 1.4437572477773483e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14721
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.34375,
      "epoch": 0.7113451874758407,
      "grad_norm": 0.0,
      "learning_rate": 1.443515655199072e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14722
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.75,
      "epoch": 0.711393505991496,
      "grad_norm": 0.0,
      "learning_rate": 1.4432740626207963e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14723
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.53125,
      "epoch": 0.7114418245071511,
      "grad_norm": 0.0,
      "learning_rate": 1.4430324700425204e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14724
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.46875,
      "epoch": 0.7114901430228063,
      "grad_norm": 0.0,
      "learning_rate": 1.4427908774642443e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880052089691,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14725
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.15625,
      "epoch": 0.7115384615384616,
      "grad_norm": 0.0,
      "learning_rate": 1.4425492848859681e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14726
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.9375,
      "epoch": 0.7115867800541167,
      "grad_norm": 0.0,
      "learning_rate": 1.442307692307692e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14727
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.84375,
      "epoch": 0.7116350985697719,
      "grad_norm": 0.0,
      "learning_rate": 1.4420660997294162e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14728
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.0625,
      "epoch": 0.7116834170854272,
      "grad_norm": 0.0,
      "learning_rate": 1.4418245071511403e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14729
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.875,
      "epoch": 0.7117317356010824,
      "grad_norm": 0.0,
      "learning_rate": 1.4415829145728642e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14730
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.53125,
      "epoch": 0.7117800541167375,
      "grad_norm": 0.0,
      "learning_rate": 1.4413413219945883e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 14731
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.59375,
      "epoch": 0.7118283726323927,
      "grad_norm": 0.0,
      "learning_rate": 1.4410997294163122e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14732
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 94.15625,
      "epoch": 0.711876691148048,
      "grad_norm": 11.206154766231677,
      "learning_rate": 1.4408581368380363e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14733
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.34375,
      "epoch": 0.7119250096637031,
      "grad_norm": 0.0,
      "learning_rate": 1.4406165442597604e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14734
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.28125,
      "epoch": 0.7119733281793583,
      "grad_norm": 0.0,
      "learning_rate": 1.4403749516814843e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 14735
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 88.25,
      "epoch": 0.7120216466950136,
      "grad_norm": 73.31594855315142,
      "learning_rate": 1.4401333591032085e-07,
      "loss": 0.0,
      "reward": 1.546875,
      "reward_std": 0.188242569565773,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14736
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.625,
      "epoch": 0.7120699652106687,
      "grad_norm": 0.0,
      "learning_rate": 1.439891766524932e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14737
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.3125,
      "epoch": 0.7121182837263239,
      "grad_norm": 0.0,
      "learning_rate": 1.4396501739466562e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14738
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.59375,
      "epoch": 0.7121666022419791,
      "grad_norm": 0.0,
      "learning_rate": 1.4394085813683803e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14739
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 80.625,
      "epoch": 0.7122149207576344,
      "grad_norm": 12.485023557111283,
      "learning_rate": 1.4391669887901042e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.2641943693161011,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14740
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 72.5625,
      "epoch": 0.7122632392732895,
      "grad_norm": 0.0,
      "learning_rate": 1.4389253962118283e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14741
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 65.96875,
      "epoch": 0.7123115577889447,
      "grad_norm": 0.0,
      "learning_rate": 1.4386838036335522e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14742
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.4375,
      "epoch": 0.7123598763046,
      "grad_norm": 0.0,
      "learning_rate": 1.4384422110552763e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14743
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.625,
      "epoch": 0.7124081948202551,
      "grad_norm": 0.0,
      "learning_rate": 1.4382006184770005e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14744
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.84375,
      "epoch": 0.7124565133359103,
      "grad_norm": 0.0,
      "learning_rate": 1.4379590258987244e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14745
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.1875,
      "epoch": 0.7125048318515655,
      "grad_norm": 0.0,
      "learning_rate": 1.4377174333204485e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14746
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.375,
      "epoch": 0.7125531503672207,
      "grad_norm": 0.0,
      "learning_rate": 1.437475840742172e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14747
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.96875,
      "epoch": 0.7126014688828759,
      "grad_norm": 0.0,
      "learning_rate": 1.4372342481638962e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14748
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.28125,
      "epoch": 0.7126497873985311,
      "grad_norm": 0.0,
      "learning_rate": 1.4369926555856204e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14749
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.25,
      "epoch": 0.7126981059141864,
      "grad_norm": 0.0,
      "learning_rate": 1.4367510630073442e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14750
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 64.6875,
      "epoch": 0.7127464244298415,
      "grad_norm": 0.0,
      "learning_rate": 1.4365094704290684e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14751
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.8125,
      "epoch": 0.7127947429454967,
      "grad_norm": 0.0,
      "learning_rate": 1.4362678778507922e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14752
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.15625,
      "epoch": 0.712843061461152,
      "grad_norm": 0.0,
      "learning_rate": 1.4360262852725164e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14753
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.375,
      "epoch": 0.7128913799768071,
      "grad_norm": 0.0,
      "learning_rate": 1.4357846926942405e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14754
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 73.59375,
      "epoch": 0.7129396984924623,
      "grad_norm": 0.0,
      "learning_rate": 1.4355431001159644e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14755
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.96875,
      "epoch": 0.7129880170081175,
      "grad_norm": 0.0,
      "learning_rate": 1.4353015075376885e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14756
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.96875,
      "epoch": 0.7130363355237727,
      "grad_norm": 0.0,
      "learning_rate": 1.4350599149594121e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14757
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.3125,
      "epoch": 0.7130846540394279,
      "grad_norm": 10.211190039715088,
      "learning_rate": 1.4348183223811363e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14758
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.8125,
      "epoch": 0.7131329725550831,
      "grad_norm": 0.0,
      "learning_rate": 1.4345767298028604e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14759
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.1875,
      "epoch": 0.7131812910707384,
      "grad_norm": 0.0,
      "learning_rate": 1.4343351372245843e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14760
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.96875,
      "epoch": 0.7132296095863935,
      "grad_norm": 0.0,
      "learning_rate": 1.4340935446463084e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14761
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.53125,
      "epoch": 0.7132779281020487,
      "grad_norm": 0.0,
      "learning_rate": 1.4338519520680323e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14762
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 62.96875,
      "epoch": 0.7133262466177039,
      "grad_norm": 0.0,
      "learning_rate": 1.4336103594897564e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14763
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 95.375,
      "epoch": 0.7133745651333591,
      "grad_norm": 0.0,
      "learning_rate": 1.4333687669114806e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 14764
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 71.8125,
      "epoch": 0.7134228836490143,
      "grad_norm": 5.439640893637235,
      "learning_rate": 1.4331271743332044e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14765
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.03125,
      "epoch": 0.7134712021646695,
      "grad_norm": 0.0,
      "learning_rate": 1.4328855817549286e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14766
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.9375,
      "epoch": 0.7135195206803246,
      "grad_norm": 0.0,
      "learning_rate": 1.4326439891766522e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14767
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.84375,
      "epoch": 0.7135678391959799,
      "grad_norm": 0.0,
      "learning_rate": 1.4324023965983763e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14768
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.3125,
      "epoch": 0.7136161577116351,
      "grad_norm": 0.0,
      "learning_rate": 1.4321608040201004e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14769
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 68.53125,
      "epoch": 0.7136644762272903,
      "grad_norm": 0.0,
      "learning_rate": 1.4319192114418243e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14770
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.5625,
      "epoch": 0.7137127947429455,
      "grad_norm": 0.0,
      "learning_rate": 1.4316776188635485e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14771
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 74.3125,
      "epoch": 0.7137611132586007,
      "grad_norm": 0.0,
      "learning_rate": 1.4314360262852723e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 14772
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.65625,
      "epoch": 0.7138094317742559,
      "grad_norm": 0.0,
      "learning_rate": 1.4311944337069965e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14773
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.8125,
      "epoch": 0.713857750289911,
      "grad_norm": 0.0,
      "learning_rate": 1.4309528411287206e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14774
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.6875,
      "epoch": 0.7139060688055663,
      "grad_norm": 0.0,
      "learning_rate": 1.4307112485504445e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14775
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 70.75,
      "epoch": 0.7139543873212215,
      "grad_norm": 39.956959102839946,
      "learning_rate": 1.4304696559721686e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.21316051483154297,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14776
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.6875,
      "epoch": 0.7140027058368766,
      "grad_norm": 0.0,
      "learning_rate": 1.4302280633938925e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2130674123764038,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14777
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.15625,
      "epoch": 0.7140510243525319,
      "grad_norm": 0.0,
      "learning_rate": 1.4299864708156164e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14778
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.40625,
      "epoch": 0.7140993428681871,
      "grad_norm": 0.0,
      "learning_rate": 1.4297448782373405e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14779
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 67.1875,
      "epoch": 0.7141476613838423,
      "grad_norm": 0.0,
      "learning_rate": 1.4295032856590644e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14780
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.0,
      "epoch": 0.7141959798994975,
      "grad_norm": 0.0,
      "learning_rate": 1.4292616930807885e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14781
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.5,
      "epoch": 0.7142442984151527,
      "grad_norm": 0.0,
      "learning_rate": 1.4290201005025124e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14782
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.5,
      "epoch": 0.7142926169308079,
      "grad_norm": 0.0,
      "learning_rate": 1.4287785079242365e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14783
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.8125,
      "epoch": 0.714340935446463,
      "grad_norm": 0.0,
      "learning_rate": 1.4285369153459606e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14784
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.71875,
      "epoch": 0.7143892539621183,
      "grad_norm": 0.0,
      "learning_rate": 1.4282953227676845e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14785
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 77.5,
      "epoch": 0.7144375724777735,
      "grad_norm": 0.0,
      "learning_rate": 1.4280537301894086e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14786
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.125,
      "epoch": 0.7144858909934286,
      "grad_norm": 0.0,
      "learning_rate": 1.4278121376111325e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14787
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.09375,
      "epoch": 0.7145342095090839,
      "grad_norm": 0.0,
      "learning_rate": 1.4275705450328564e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14788
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.28125,
      "epoch": 0.7145825280247391,
      "grad_norm": 0.0,
      "learning_rate": 1.4273289524545805e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24251843988895416,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14789
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.96875,
      "epoch": 0.7146308465403943,
      "grad_norm": 0.0,
      "learning_rate": 1.4270873598763044e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14790
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 72.5,
      "epoch": 0.7146791650560494,
      "grad_norm": 0.0,
      "learning_rate": 1.4268457672980285e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14791
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 102.375,
      "epoch": 0.7147274835717047,
      "grad_norm": 4.51842947865349,
      "learning_rate": 1.4266041747197524e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14792
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.28125,
      "epoch": 0.7147758020873599,
      "grad_norm": 0.0,
      "learning_rate": 1.4263625821414765e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14793
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.1875,
      "epoch": 0.714824120603015,
      "grad_norm": 0.0,
      "learning_rate": 1.4261209895632007e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14794
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.71875,
      "epoch": 0.7148724391186703,
      "grad_norm": 0.0,
      "learning_rate": 1.4258793969849246e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14795
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.8125,
      "epoch": 0.7149207576343255,
      "grad_norm": 0.0,
      "learning_rate": 1.4256378044066487e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14796
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.75,
      "epoch": 0.7149690761499806,
      "grad_norm": 0.0,
      "learning_rate": 1.4253962118283726e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14797
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.0625,
      "epoch": 0.7150173946656359,
      "grad_norm": 0.0,
      "learning_rate": 1.4251546192500964e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14798
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.96875,
      "epoch": 0.7150657131812911,
      "grad_norm": 0.0,
      "learning_rate": 1.4249130266718206e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14799
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 92.28125,
      "epoch": 0.7151140316969463,
      "grad_norm": 7.02102424939944,
      "learning_rate": 1.4246714340935444e-07,
      "loss": -0.0,
      "reward": 2.34375,
      "reward_std": 0.2925041913986206,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14800
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.5,
      "epoch": 0.7151623502126014,
      "grad_norm": 0.0,
      "learning_rate": 1.4244298415152686e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14801
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.75,
      "epoch": 0.7152106687282567,
      "grad_norm": 0.0,
      "learning_rate": 1.4241882489369924e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14802
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 76.34375,
      "epoch": 0.7152589872439119,
      "grad_norm": 5.358661215943018,
      "learning_rate": 1.4239466563587166e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14803
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.03125,
      "epoch": 0.715307305759567,
      "grad_norm": 0.0,
      "learning_rate": 1.4237050637804407e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14804
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.625,
      "epoch": 0.7153556242752223,
      "grad_norm": 0.0,
      "learning_rate": 1.4234634712021646e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14805
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.96875,
      "epoch": 0.7154039427908775,
      "grad_norm": 0.0,
      "learning_rate": 1.4232218786238887e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14806
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.0625,
      "epoch": 0.7154522613065326,
      "grad_norm": 0.0,
      "learning_rate": 1.4229802860456126e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14807
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.21875,
      "epoch": 0.7155005798221878,
      "grad_norm": 0.0,
      "learning_rate": 1.4227386934673367e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14808
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.96875,
      "epoch": 0.7155488983378431,
      "grad_norm": 0.0,
      "learning_rate": 1.4224971008890606e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 14809
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.21875,
      "epoch": 0.7155972168534983,
      "grad_norm": 0.0,
      "learning_rate": 1.4222555083107845e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14810
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.65625,
      "epoch": 0.7156455353691534,
      "grad_norm": 0.0,
      "learning_rate": 1.4220139157325086e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14811
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.625,
      "epoch": 0.7156938538848087,
      "grad_norm": 0.0,
      "learning_rate": 1.4217723231542325e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14812
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.5,
      "epoch": 0.7157421724004639,
      "grad_norm": 0.0,
      "learning_rate": 1.4215307305759566e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14813
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.78125,
      "epoch": 0.715790490916119,
      "grad_norm": 0.0,
      "learning_rate": 1.4212891379976808e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14814
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.65625,
      "epoch": 0.7158388094317742,
      "grad_norm": 0.0,
      "learning_rate": 1.4210475454194046e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14815
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.21875,
      "epoch": 0.7158871279474295,
      "grad_norm": 0.0,
      "learning_rate": 1.4208059528411288e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14816
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.28125,
      "epoch": 0.7159354464630846,
      "grad_norm": 0.0,
      "learning_rate": 1.4205643602628526e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14817
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.90625,
      "epoch": 0.7159837649787398,
      "grad_norm": 0.0,
      "learning_rate": 1.4203227676845768e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14818
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.0,
      "epoch": 0.7160320834943951,
      "grad_norm": 0.0,
      "learning_rate": 1.4200811751063006e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14819
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.21875,
      "epoch": 0.7160804020100503,
      "grad_norm": 0.0,
      "learning_rate": 1.4198395825280245e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14820
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.09375,
      "epoch": 0.7161287205257054,
      "grad_norm": 0.0,
      "learning_rate": 1.4195979899497487e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14821
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.65625,
      "epoch": 0.7161770390413607,
      "grad_norm": 0.0,
      "learning_rate": 1.4193563973714725e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14822
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.5625,
      "epoch": 0.7162253575570159,
      "grad_norm": 0.0,
      "learning_rate": 1.4191148047931967e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14823
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.78125,
      "epoch": 0.716273676072671,
      "grad_norm": 0.0,
      "learning_rate": 1.4188732122149208e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14824
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.28125,
      "epoch": 0.7163219945883262,
      "grad_norm": 0.0,
      "learning_rate": 1.4186316196366447e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14825
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.375,
      "epoch": 0.7163703131039815,
      "grad_norm": 0.0,
      "learning_rate": 1.4183900270583688e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14826
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 104.65625,
      "epoch": 0.7164186316196366,
      "grad_norm": 7.095585316673841,
      "learning_rate": 1.4181484344800927e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.25726157426834106,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14827
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.1875,
      "epoch": 0.7164669501352918,
      "grad_norm": 0.0,
      "learning_rate": 1.4179068419018168e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14828
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 90.5,
      "epoch": 0.7165152686509471,
      "grad_norm": 18.579392727721434,
      "learning_rate": 1.4176652493235407e-07,
      "loss": 0.0,
      "reward": 1.875,
      "reward_std": 0.21316051483154297,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14829
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.1875,
      "epoch": 0.7165635871666023,
      "grad_norm": 0.0,
      "learning_rate": 1.4174236567452646e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 14830
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.40625,
      "epoch": 0.7166119056822574,
      "grad_norm": 0.0,
      "learning_rate": 1.4171820641669887e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14831
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 90.5,
      "epoch": 0.7166602241979126,
      "grad_norm": 30.061372010568196,
      "learning_rate": 1.4169404715887126e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.15580293536186218,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14832
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.3125,
      "epoch": 0.7167085427135679,
      "grad_norm": 0.0,
      "learning_rate": 1.4166988790104367e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14833
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.875,
      "epoch": 0.716756861229223,
      "grad_norm": 0.0,
      "learning_rate": 1.4164572864321608e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14834
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.5,
      "epoch": 0.7168051797448782,
      "grad_norm": 0.0,
      "learning_rate": 1.4162156938538847e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14835
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.75,
      "epoch": 0.7168534982605335,
      "grad_norm": 0.0,
      "learning_rate": 1.4159741012756088e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14836
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.59375,
      "epoch": 0.7169018167761886,
      "grad_norm": 0.0,
      "learning_rate": 1.4157325086973327e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14837
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.5,
      "epoch": 0.7169501352918438,
      "grad_norm": 0.0,
      "learning_rate": 1.4154909161190569e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14838
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.0,
      "epoch": 0.716998453807499,
      "grad_norm": 0.0,
      "learning_rate": 1.415249323540781e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14839
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.8125,
      "epoch": 0.7170467723231543,
      "grad_norm": 0.0,
      "learning_rate": 1.4150077309625046e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14840
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.0,
      "epoch": 0.7170950908388094,
      "grad_norm": 0.0,
      "learning_rate": 1.4147661383842287e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14841
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.0625,
      "epoch": 0.7171434093544646,
      "grad_norm": 0.0,
      "learning_rate": 1.4145245458059526e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14842
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.6875,
      "epoch": 0.7171917278701199,
      "grad_norm": 0.0,
      "learning_rate": 1.4142829532276767e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14843
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.09375,
      "epoch": 0.717240046385775,
      "grad_norm": 0.0,
      "learning_rate": 1.414041360649401e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14844
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.71875,
      "epoch": 0.7172883649014302,
      "grad_norm": 0.0,
      "learning_rate": 1.4137997680711247e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14845
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.34375,
      "epoch": 0.7173366834170855,
      "grad_norm": 0.0,
      "learning_rate": 1.413558175492849e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14846
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.1875,
      "epoch": 0.7173850019327406,
      "grad_norm": 0.0,
      "learning_rate": 1.4133165829145728e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14847
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.21875,
      "epoch": 0.7174333204483958,
      "grad_norm": 0.0,
      "learning_rate": 1.413074990336297e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14848
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.28125,
      "epoch": 0.717481638964051,
      "grad_norm": 0.0,
      "learning_rate": 1.412833397758021e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14849
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.375,
      "epoch": 0.7175299574797063,
      "grad_norm": 0.0,
      "learning_rate": 1.4125918051797446e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14850
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.53125,
      "epoch": 0.7175782759953614,
      "grad_norm": 0.0,
      "learning_rate": 1.4123502126014688e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14851
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.25,
      "epoch": 0.7176265945110166,
      "grad_norm": 0.0,
      "learning_rate": 1.4121086200231926e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14852
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.09375,
      "epoch": 0.7176749130266719,
      "grad_norm": 0.0,
      "learning_rate": 1.4118670274449168e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14853
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.46875,
      "epoch": 0.717723231542327,
      "grad_norm": 0.0,
      "learning_rate": 1.411625434866641e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24251843988895416,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14854
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.59375,
      "epoch": 0.7177715500579822,
      "grad_norm": 0.0,
      "learning_rate": 1.4113838422883648e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14855
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 91.0625,
      "epoch": 0.7178198685736374,
      "grad_norm": 46.794639530974855,
      "learning_rate": 1.411142249710089e-07,
      "loss": 0.0,
      "reward": 1.46875,
      "reward_std": 0.2346404492855072,
      "rewards/accuracy_reward": 0.34375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14856
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.8125,
      "epoch": 0.7178681870892926,
      "grad_norm": 0.0,
      "learning_rate": 1.4109006571318128e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 14857
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.6875,
      "epoch": 0.7179165056049478,
      "grad_norm": 0.0,
      "learning_rate": 1.410659064553537e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14858
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.71875,
      "epoch": 0.717964824120603,
      "grad_norm": 0.0,
      "learning_rate": 1.410417471975261e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14859
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.34375,
      "epoch": 0.7180131426362583,
      "grad_norm": 0.0,
      "learning_rate": 1.4101758793969847e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14860
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.1875,
      "epoch": 0.7180614611519134,
      "grad_norm": 0.0,
      "learning_rate": 1.4099342868187088e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14861
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.1875,
      "epoch": 0.7181097796675686,
      "grad_norm": 0.0,
      "learning_rate": 1.4096926942404327e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14862
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.46875,
      "epoch": 0.7181580981832238,
      "grad_norm": 0.0,
      "learning_rate": 1.4094511016621568e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 14863
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.25,
      "epoch": 0.718206416698879,
      "grad_norm": 0.0,
      "learning_rate": 1.409209509083881e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14864
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.4375,
      "epoch": 0.7182547352145342,
      "grad_norm": 0.0,
      "learning_rate": 1.4089679165056048e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14865
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 76.1875,
      "epoch": 0.7183030537301894,
      "grad_norm": 0.0,
      "learning_rate": 1.408726323927329e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14866
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.3125,
      "epoch": 0.7183513722458446,
      "grad_norm": 0.0,
      "learning_rate": 1.4084847313490528e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14867
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 87.0625,
      "epoch": 0.7183996907614998,
      "grad_norm": 6.583054724583663,
      "learning_rate": 1.408243138770777e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14868
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.65625,
      "epoch": 0.718448009277155,
      "grad_norm": 0.0,
      "learning_rate": 1.408001546192501e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14869
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.25,
      "epoch": 0.7184963277928103,
      "grad_norm": 0.0,
      "learning_rate": 1.4077599536142247e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14870
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.09375,
      "epoch": 0.7185446463084654,
      "grad_norm": 0.0,
      "learning_rate": 1.4075183610359488e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14871
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.03125,
      "epoch": 0.7185929648241206,
      "grad_norm": 0.0,
      "learning_rate": 1.4072767684576727e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14872
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.21875,
      "epoch": 0.7186412833397758,
      "grad_norm": 0.0,
      "learning_rate": 1.4070351758793969e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14873
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.375,
      "epoch": 0.718689601855431,
      "grad_norm": 0.0,
      "learning_rate": 1.406793583301121e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14874
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.65625,
      "epoch": 0.7187379203710862,
      "grad_norm": 0.0,
      "learning_rate": 1.4065519907228449e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14875
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 91.09375,
      "epoch": 0.7187862388867414,
      "grad_norm": 0.0,
      "learning_rate": 1.406310398144569e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 14876
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.59375,
      "epoch": 0.7188345574023965,
      "grad_norm": 0.0,
      "learning_rate": 1.406068805566293e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14877
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.0625,
      "epoch": 0.7188828759180518,
      "grad_norm": 0.0,
      "learning_rate": 1.405827212988017e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14878
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.625,
      "epoch": 0.718931194433707,
      "grad_norm": 0.0,
      "learning_rate": 1.4055856204097411e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 14879
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.5,
      "epoch": 0.7189795129493622,
      "grad_norm": 0.0,
      "learning_rate": 1.4053440278314648e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14880
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.46875,
      "epoch": 0.7190278314650174,
      "grad_norm": 0.0,
      "learning_rate": 1.405102435253189e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14881
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.84375,
      "epoch": 0.7190761499806726,
      "grad_norm": 0.0,
      "learning_rate": 1.4048608426749128e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14882
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.3125,
      "epoch": 0.7191244684963278,
      "grad_norm": 0.0,
      "learning_rate": 1.404619250096637e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14883
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.75,
      "epoch": 0.719172787011983,
      "grad_norm": 0.0,
      "learning_rate": 1.404377657518361e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14884
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.25,
      "epoch": 0.7192211055276382,
      "grad_norm": 0.0,
      "learning_rate": 1.404136064940085e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14885
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.75,
      "epoch": 0.7192694240432934,
      "grad_norm": 0.0,
      "learning_rate": 1.403894472361809e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14886
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.625,
      "epoch": 0.7193177425589485,
      "grad_norm": 0.0,
      "learning_rate": 1.403652879783533e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14887
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 83.03125,
      "epoch": 0.7193660610746038,
      "grad_norm": 0.0,
      "learning_rate": 1.403411287205257e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14888
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.0625,
      "epoch": 0.719414379590259,
      "grad_norm": 0.0,
      "learning_rate": 1.4031696946269812e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14889
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.78125,
      "epoch": 0.7194626981059142,
      "grad_norm": 0.0,
      "learning_rate": 1.402928102048705e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14890
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.96875,
      "epoch": 0.7195110166215694,
      "grad_norm": 0.0,
      "learning_rate": 1.402686509470429e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 14891
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 85.21875,
      "epoch": 0.7195593351372246,
      "grad_norm": 12.607313248973858,
      "learning_rate": 1.4024449168921528e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.24359199404716492,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14892
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 88.90625,
      "epoch": 0.7196076536528798,
      "grad_norm": 71.3318628223751,
      "learning_rate": 1.402203324313877e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19999711215496063,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14893
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 93.875,
      "epoch": 0.7196559721685349,
      "grad_norm": 0.0,
      "learning_rate": 1.401961731735601e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 14894
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.84375,
      "epoch": 0.7197042906841902,
      "grad_norm": 0.0,
      "learning_rate": 1.401720139157325e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14895
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.25,
      "epoch": 0.7197526091998454,
      "grad_norm": 0.0,
      "learning_rate": 1.401478546579049e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14896
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.9375,
      "epoch": 0.7198009277155005,
      "grad_norm": 0.0,
      "learning_rate": 1.401236954000773e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14897
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.25,
      "epoch": 0.7198492462311558,
      "grad_norm": 0.0,
      "learning_rate": 1.400995361422497e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14898
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.46875,
      "epoch": 0.719897564746811,
      "grad_norm": 0.0,
      "learning_rate": 1.4007537688442212e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14899
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 93.5,
      "epoch": 0.7199458832624662,
      "grad_norm": 11.106610249780628,
      "learning_rate": 1.400512176265945e-07,
      "loss": 0.0,
      "reward": 1.578125,
      "reward_std": 0.17598573863506317,
      "rewards/accuracy_reward": 0.40625,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14900
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.65625,
      "epoch": 0.7199942017781213,
      "grad_norm": 0.0,
      "learning_rate": 1.400270583687669e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14901
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.375,
      "epoch": 0.7200425202937766,
      "grad_norm": 0.0,
      "learning_rate": 1.4000289911093928e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 14902
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.875,
      "epoch": 0.7200908388094318,
      "grad_norm": 0.0,
      "learning_rate": 1.399787398531117e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14903
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.46875,
      "epoch": 0.7201391573250869,
      "grad_norm": 0.0,
      "learning_rate": 1.399545805952841e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14904
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.65625,
      "epoch": 0.7201874758407422,
      "grad_norm": 0.0,
      "learning_rate": 1.399304213374565e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14905
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.8125,
      "epoch": 0.7202357943563974,
      "grad_norm": 0.0,
      "learning_rate": 1.399062620796289e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14906
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.8125,
      "epoch": 0.7202841128720525,
      "grad_norm": 0.0,
      "learning_rate": 1.398821028218013e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 14907
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.3125,
      "epoch": 0.7203324313877078,
      "grad_norm": 0.0,
      "learning_rate": 1.398579435639737e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14908
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.5,
      "epoch": 0.720380749903363,
      "grad_norm": 0.0,
      "learning_rate": 1.3983378430614613e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14909
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.1875,
      "epoch": 0.7204290684190182,
      "grad_norm": 0.0,
      "learning_rate": 1.3980962504831851e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14910
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.8125,
      "epoch": 0.7204773869346733,
      "grad_norm": 0.0,
      "learning_rate": 1.3978546579049093e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 14911
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.40625,
      "epoch": 0.7205257054503286,
      "grad_norm": 0.0,
      "learning_rate": 1.397613065326633e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14912
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.53125,
      "epoch": 0.7205740239659838,
      "grad_norm": 0.0,
      "learning_rate": 1.397371472748357e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14913
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.03125,
      "epoch": 0.7206223424816389,
      "grad_norm": 0.0,
      "learning_rate": 1.3971298801700812e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14914
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.40625,
      "epoch": 0.7206706609972942,
      "grad_norm": 0.0,
      "learning_rate": 1.396888287591805e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14915
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.15625,
      "epoch": 0.7207189795129494,
      "grad_norm": 0.0,
      "learning_rate": 1.3966466950135292e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14916
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.6875,
      "epoch": 0.7207672980286045,
      "grad_norm": 0.0,
      "learning_rate": 1.396405102435253e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14917
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.3125,
      "epoch": 0.7208156165442597,
      "grad_norm": 0.0,
      "learning_rate": 1.3961635098569772e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 14918
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 93.40625,
      "epoch": 0.720863935059915,
      "grad_norm": 0.0,
      "learning_rate": 1.3959219172787013e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 14919
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.21875,
      "epoch": 0.7209122535755702,
      "grad_norm": 0.0,
      "learning_rate": 1.3956803247004252e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14920
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.15625,
      "epoch": 0.7209605720912253,
      "grad_norm": 0.0,
      "learning_rate": 1.3954387321221493e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14921
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.625,
      "epoch": 0.7210088906068806,
      "grad_norm": 0.0,
      "learning_rate": 1.395197139543873e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14922
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 72.65625,
      "epoch": 0.7210572091225358,
      "grad_norm": 0.0,
      "learning_rate": 1.394955546965597e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 14923
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.9375,
      "epoch": 0.7211055276381909,
      "grad_norm": 0.0,
      "learning_rate": 1.3947139543873212e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14924
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.625,
      "epoch": 0.7211538461538461,
      "grad_norm": 0.0,
      "learning_rate": 1.394472361809045e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14925
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.4375,
      "epoch": 0.7212021646695014,
      "grad_norm": 0.0,
      "learning_rate": 1.3942307692307692e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14926
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.46875,
      "epoch": 0.7212504831851565,
      "grad_norm": 0.0,
      "learning_rate": 1.393989176652493e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14927
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 82.59375,
      "epoch": 0.7212988017008117,
      "grad_norm": 0.0,
      "learning_rate": 1.3937475840742172e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 14928
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.90625,
      "epoch": 0.721347120216467,
      "grad_norm": 0.0,
      "learning_rate": 1.3935059914959413e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14929
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.375,
      "epoch": 0.7213954387321222,
      "grad_norm": 0.0,
      "learning_rate": 1.3932643989176652e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14930
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.625,
      "epoch": 0.7214437572477773,
      "grad_norm": 0.0,
      "learning_rate": 1.3930228063393894e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14931
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.34375,
      "epoch": 0.7214920757634325,
      "grad_norm": 0.0,
      "learning_rate": 1.392781213761113e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14932
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.34375,
      "epoch": 0.7215403942790878,
      "grad_norm": 0.0,
      "learning_rate": 1.392539621182837e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14933
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.25,
      "epoch": 0.7215887127947429,
      "grad_norm": 0.0,
      "learning_rate": 1.3922980286045612e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14934
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 81.3125,
      "epoch": 0.7216370313103981,
      "grad_norm": 159.19808246824562,
      "learning_rate": 1.392056436026285e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.3255898952484131,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14935
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.09375,
      "epoch": 0.7216853498260534,
      "grad_norm": 0.0,
      "learning_rate": 1.3918148434480092e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 14936
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.6875,
      "epoch": 0.7217336683417085,
      "grad_norm": 0.0,
      "learning_rate": 1.391573250869733e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14937
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 73.9375,
      "epoch": 0.7217819868573637,
      "grad_norm": 0.0,
      "learning_rate": 1.3913316582914572e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 14938
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.4375,
      "epoch": 0.721830305373019,
      "grad_norm": 0.0,
      "learning_rate": 1.3910900657131814e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14939
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.3125,
      "epoch": 0.7218786238886742,
      "grad_norm": 0.0,
      "learning_rate": 1.3908484731349053e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14940
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.84375,
      "epoch": 0.7219269424043293,
      "grad_norm": 0.0,
      "learning_rate": 1.3906068805566294e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14941
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.625,
      "epoch": 0.7219752609199845,
      "grad_norm": 0.0,
      "learning_rate": 1.390365287978353e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14942
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.8125,
      "epoch": 0.7220235794356398,
      "grad_norm": 0.0,
      "learning_rate": 1.3901236954000771e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14943
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.5,
      "epoch": 0.7220718979512949,
      "grad_norm": 0.0,
      "learning_rate": 1.3898821028218013e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14944
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.875,
      "epoch": 0.7221202164669501,
      "grad_norm": 0.0,
      "learning_rate": 1.3896405102435251e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 14945
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.53125,
      "epoch": 0.7221685349826054,
      "grad_norm": 0.0,
      "learning_rate": 1.3893989176652493e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14946
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.625,
      "epoch": 0.7222168534982605,
      "grad_norm": 0.0,
      "learning_rate": 1.3891573250869731e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14947
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.375,
      "epoch": 0.7222651720139157,
      "grad_norm": 0.0,
      "learning_rate": 1.3889157325086973e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14948
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.71875,
      "epoch": 0.722313490529571,
      "grad_norm": 0.0,
      "learning_rate": 1.3886741399304214e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14949
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.21875,
      "epoch": 0.7223618090452262,
      "grad_norm": 0.0,
      "learning_rate": 1.3884325473521453e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14950
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.4375,
      "epoch": 0.7224101275608813,
      "grad_norm": 0.0,
      "learning_rate": 1.3881909547738694e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14951
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.1875,
      "epoch": 0.7224584460765365,
      "grad_norm": 0.0,
      "learning_rate": 1.387949362195593e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14952
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.25,
      "epoch": 0.7225067645921918,
      "grad_norm": 0.0,
      "learning_rate": 1.3877077696173172e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 14953
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.21875,
      "epoch": 0.7225550831078469,
      "grad_norm": 0.0,
      "learning_rate": 1.3874661770390413e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14954
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.84375,
      "epoch": 0.7226034016235021,
      "grad_norm": 0.0,
      "learning_rate": 1.3872245844607652e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14955
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.0,
      "epoch": 0.7226517201391573,
      "grad_norm": 0.0,
      "learning_rate": 1.3869829918824893e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 14956
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.46875,
      "epoch": 0.7227000386548125,
      "grad_norm": 0.0,
      "learning_rate": 1.3867413993042132e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14957
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.59375,
      "epoch": 0.7227483571704677,
      "grad_norm": 0.0,
      "learning_rate": 1.3864998067259373e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14958
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.59375,
      "epoch": 0.7227966756861229,
      "grad_norm": 0.0,
      "learning_rate": 1.3862582141476615e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14959
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.6875,
      "epoch": 0.7228449942017782,
      "grad_norm": 0.0,
      "learning_rate": 1.3860166215693853e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14960
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.0,
      "epoch": 0.7228933127174333,
      "grad_norm": 0.0,
      "learning_rate": 1.3857750289911095e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14961
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 71.5,
      "epoch": 0.7229416312330885,
      "grad_norm": 0.0,
      "learning_rate": 1.3855334364128333e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14962
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.34375,
      "epoch": 0.7229899497487438,
      "grad_norm": 0.0,
      "learning_rate": 1.3852918438345572e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14963
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.65625,
      "epoch": 0.7230382682643989,
      "grad_norm": 0.0,
      "learning_rate": 1.3850502512562813e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14964
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 81.96875,
      "epoch": 0.7230865867800541,
      "grad_norm": 0.0,
      "learning_rate": 1.3848086586780052e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 14965
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.8125,
      "epoch": 0.7231349052957093,
      "grad_norm": 0.0,
      "learning_rate": 1.3845670660997294e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 14966
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.71875,
      "epoch": 0.7231832238113645,
      "grad_norm": 0.0,
      "learning_rate": 1.3843254735214532e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14967
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.84375,
      "epoch": 0.7232315423270197,
      "grad_norm": 0.0,
      "learning_rate": 1.3840838809431774e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 14968
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.34375,
      "epoch": 0.7232798608426749,
      "grad_norm": 0.0,
      "learning_rate": 1.3838422883649015e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14969
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.625,
      "epoch": 0.7233281793583302,
      "grad_norm": 0.0,
      "learning_rate": 1.3836006957866254e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14970
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.75,
      "epoch": 0.7233764978739853,
      "grad_norm": 0.0,
      "learning_rate": 1.3833591032083495e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14971
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.4375,
      "epoch": 0.7234248163896405,
      "grad_norm": 0.0,
      "learning_rate": 1.3831175106300734e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14972
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.21875,
      "epoch": 0.7234731349052957,
      "grad_norm": 0.0,
      "learning_rate": 1.3828759180517973e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14973
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.28125,
      "epoch": 0.7235214534209509,
      "grad_norm": 0.0,
      "learning_rate": 1.3826343254735214e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14974
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 94.46875,
      "epoch": 0.7235697719366061,
      "grad_norm": 0.0,
      "learning_rate": 1.3823927328952453e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 14975
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.84375,
      "epoch": 0.7236180904522613,
      "grad_norm": 0.0,
      "learning_rate": 1.3821511403169694e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14976
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.28125,
      "epoch": 0.7236664089679165,
      "grad_norm": 0.0,
      "learning_rate": 1.3819095477386933e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14977
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.375,
      "epoch": 0.7237147274835717,
      "grad_norm": 0.0,
      "learning_rate": 1.3816679551604174e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14978
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.28125,
      "epoch": 0.7237630459992269,
      "grad_norm": 0.0,
      "learning_rate": 1.3814263625821415e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14979
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.875,
      "epoch": 0.7238113645148821,
      "grad_norm": 0.0,
      "learning_rate": 1.3811847700038654e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14980
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.25,
      "epoch": 0.7238596830305373,
      "grad_norm": 0.0,
      "learning_rate": 1.3809431774255895e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14981
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.53125,
      "epoch": 0.7239080015461925,
      "grad_norm": 0.0,
      "learning_rate": 1.3807015848473134e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 14982
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.40625,
      "epoch": 0.7239563200618477,
      "grad_norm": 0.0,
      "learning_rate": 1.3804599922690373e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14983
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.8125,
      "epoch": 0.7240046385775029,
      "grad_norm": 0.0,
      "learning_rate": 1.3802183996907614e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 14984
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.5,
      "epoch": 0.7240529570931581,
      "grad_norm": 0.0,
      "learning_rate": 1.3799768071124853e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 14985
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 69.46875,
      "epoch": 0.7241012756088133,
      "grad_norm": 0.0,
      "learning_rate": 1.3797352145342094e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 14986
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.28125,
      "epoch": 0.7241495941244684,
      "grad_norm": 0.0,
      "learning_rate": 1.3794936219559333e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 14987
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 85.28125,
      "epoch": 0.7241979126401237,
      "grad_norm": 0.0,
      "learning_rate": 1.3792520293776574e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 14988
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.53125,
      "epoch": 0.7242462311557789,
      "grad_norm": 0.0,
      "learning_rate": 1.3790104367993816e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 14989
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.375,
      "epoch": 0.7242945496714341,
      "grad_norm": 0.0,
      "learning_rate": 1.3787688442211054e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 14990
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.5625,
      "epoch": 0.7243428681870893,
      "grad_norm": 0.0,
      "learning_rate": 1.3785272516428296e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 14991
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.78125,
      "epoch": 0.7243911867027445,
      "grad_norm": 0.0,
      "learning_rate": 1.3782856590645535e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 14992
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.65625,
      "epoch": 0.7244395052183997,
      "grad_norm": 0.0,
      "learning_rate": 1.3780440664862776e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 14993
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 78.1875,
      "epoch": 0.7244878237340548,
      "grad_norm": 5.536047602417978,
      "learning_rate": 1.3778024739080015e-07,
      "loss": -0.0,
      "reward": 2.375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 14994
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.875,
      "epoch": 0.7245361422497101,
      "grad_norm": 0.0,
      "learning_rate": 1.3775608813297253e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14995
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.375,
      "epoch": 0.7245844607653653,
      "grad_norm": 0.0,
      "learning_rate": 1.3773192887514495e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 14996
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.25,
      "epoch": 0.7246327792810204,
      "grad_norm": 0.0,
      "learning_rate": 1.3770776961731733e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 14997
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.09375,
      "epoch": 0.7246810977966757,
      "grad_norm": 0.0,
      "learning_rate": 1.3768361035948975e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 14998
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.34375,
      "epoch": 0.7247294163123309,
      "grad_norm": 0.0,
      "learning_rate": 1.3765945110166216e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 14999
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 70.4375,
      "epoch": 0.7247777348279861,
      "grad_norm": 0.0,
      "learning_rate": 1.3763529184383455e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15000
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 95.03125,
      "epoch": 0.7248260533436413,
      "grad_norm": 132.28361821686235,
      "learning_rate": 1.3761113258600696e-07,
      "loss": 0.0,
      "reward": 1.84375,
      "reward_std": 0.3074946999549866,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15001
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.3125,
      "epoch": 0.7248743718592965,
      "grad_norm": 0.0,
      "learning_rate": 1.3758697332817935e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15002
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.1875,
      "epoch": 0.7249226903749517,
      "grad_norm": 0.0,
      "learning_rate": 1.3756281407035176e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15003
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.28125,
      "epoch": 0.7249710088906068,
      "grad_norm": 0.0,
      "learning_rate": 1.3753865481252415e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15004
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 82.59375,
      "epoch": 0.7250193274062621,
      "grad_norm": 0.0,
      "learning_rate": 1.3751449555469654e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15005
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.21875,
      "epoch": 0.7250676459219173,
      "grad_norm": 0.0,
      "learning_rate": 1.3749033629686895e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15006
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.0625,
      "epoch": 0.7251159644375724,
      "grad_norm": 0.0,
      "learning_rate": 1.3746617703904134e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15007
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.46875,
      "epoch": 0.7251642829532277,
      "grad_norm": 0.0,
      "learning_rate": 1.3744201778121375e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15008
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 93.46875,
      "epoch": 0.7252126014688829,
      "grad_norm": 0.0,
      "learning_rate": 1.3741785852338617e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15009
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 99.1875,
      "epoch": 0.7252609199845381,
      "grad_norm": 2.3330822265619116,
      "learning_rate": 1.3739369926555855e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.3014557659626007,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15010
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.125,
      "epoch": 0.7253092385001932,
      "grad_norm": 0.0,
      "learning_rate": 1.3736954000773097e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15011
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.8125,
      "epoch": 0.7253575570158485,
      "grad_norm": 0.0,
      "learning_rate": 1.3734538074990335e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15012
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.875,
      "epoch": 0.7254058755315037,
      "grad_norm": 0.0,
      "learning_rate": 1.3732122149207577e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15013
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.5,
      "epoch": 0.7254541940471588,
      "grad_norm": 0.0,
      "learning_rate": 1.3729706223424818e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15014
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.90625,
      "epoch": 0.7255025125628141,
      "grad_norm": 0.0,
      "learning_rate": 1.3727290297642054e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15015
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.125,
      "epoch": 0.7255508310784693,
      "grad_norm": 0.0,
      "learning_rate": 1.3724874371859296e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15016
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.28125,
      "epoch": 0.7255991495941244,
      "grad_norm": 0.0,
      "learning_rate": 1.3722458446076534e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15017
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.53125,
      "epoch": 0.7256474681097796,
      "grad_norm": 0.0,
      "learning_rate": 1.3720042520293776e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15018
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.125,
      "epoch": 0.7256957866254349,
      "grad_norm": 0.0,
      "learning_rate": 1.3717626594511017e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 15019
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.40625,
      "epoch": 0.7257441051410901,
      "grad_norm": 0.0,
      "learning_rate": 1.3715210668728256e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15020
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.1875,
      "epoch": 0.7257924236567452,
      "grad_norm": 22.66268244501462,
      "learning_rate": 1.3712794742945497e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15021
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 94.4375,
      "epoch": 0.7258407421724005,
      "grad_norm": 0.0,
      "learning_rate": 1.3710378817162736e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15022
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.6875,
      "epoch": 0.7258890606880557,
      "grad_norm": 0.0,
      "learning_rate": 1.3707962891379977e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15023
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.3125,
      "epoch": 0.7259373792037108,
      "grad_norm": 0.0,
      "learning_rate": 1.3705546965597218e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15024
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.03125,
      "epoch": 0.725985697719366,
      "grad_norm": 0.0,
      "learning_rate": 1.3703131039814455e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15025
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 70.65625,
      "epoch": 0.7260340162350213,
      "grad_norm": 0.0,
      "learning_rate": 1.3700715114031696e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15026
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.125,
      "epoch": 0.7260823347506764,
      "grad_norm": 0.0,
      "learning_rate": 1.3698299188248935e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15027
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.625,
      "epoch": 0.7261306532663316,
      "grad_norm": 0.0,
      "learning_rate": 1.3695883262466176e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15028
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.1875,
      "epoch": 0.7261789717819869,
      "grad_norm": 0.0,
      "learning_rate": 1.3693467336683417e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15029
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.65625,
      "epoch": 0.7262272902976421,
      "grad_norm": 0.0,
      "learning_rate": 1.3691051410900656e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15030
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.78125,
      "epoch": 0.7262756088132972,
      "grad_norm": 0.0,
      "learning_rate": 1.3688635485117897e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15031
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.0,
      "epoch": 0.7263239273289525,
      "grad_norm": 0.0,
      "learning_rate": 1.3686219559335136e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15032
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.96875,
      "epoch": 0.7263722458446077,
      "grad_norm": 0.0,
      "learning_rate": 1.3683803633552378e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15033
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.6875,
      "epoch": 0.7264205643602628,
      "grad_norm": 0.0,
      "learning_rate": 1.368138770776962e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15034
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 68.8125,
      "epoch": 0.726468882875918,
      "grad_norm": 12.358304515290413,
      "learning_rate": 1.3678971781986855e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.28622081875801086,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15035
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.34375,
      "epoch": 0.7265172013915733,
      "grad_norm": 0.0,
      "learning_rate": 1.3676555856204096e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15036
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.09375,
      "epoch": 0.7265655199072284,
      "grad_norm": 0.0,
      "learning_rate": 1.3674139930421335e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 15037
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 76.5625,
      "epoch": 0.7266138384228836,
      "grad_norm": 0.0,
      "learning_rate": 1.3671724004638576e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15038
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.71875,
      "epoch": 0.7266621569385389,
      "grad_norm": 3.618556265456178,
      "learning_rate": 1.3669308078855818e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15039
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.8125,
      "epoch": 0.7267104754541941,
      "grad_norm": 0.0,
      "learning_rate": 1.3666892153073056e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15040
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.03125,
      "epoch": 0.7267587939698492,
      "grad_norm": 0.0,
      "learning_rate": 1.3664476227290298e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.2630178928375244,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15041
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.78125,
      "epoch": 0.7268071124855044,
      "grad_norm": 0.0,
      "learning_rate": 1.3662060301507537e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15042
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.59375,
      "epoch": 0.7268554310011597,
      "grad_norm": 0.0,
      "learning_rate": 1.3659644375724778e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15043
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.84375,
      "epoch": 0.7269037495168148,
      "grad_norm": 0.0,
      "learning_rate": 1.365722844994202e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15044
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.03125,
      "epoch": 0.72695206803247,
      "grad_norm": 0.0,
      "learning_rate": 1.3654812524159255e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24251842498779297,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15045
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 71.25,
      "epoch": 0.7270003865481253,
      "grad_norm": 0.0,
      "learning_rate": 1.3652396598376497e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15046
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.375,
      "epoch": 0.7270487050637804,
      "grad_norm": 0.0,
      "learning_rate": 1.3649980672593735e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15047
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.84375,
      "epoch": 0.7270970235794356,
      "grad_norm": 0.0,
      "learning_rate": 1.3647564746810977e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15048
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.84375,
      "epoch": 0.7271453420950909,
      "grad_norm": 0.0,
      "learning_rate": 1.3645148821028218e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15049
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 75.0,
      "epoch": 0.7271936606107461,
      "grad_norm": 0.0,
      "learning_rate": 1.3642732895245457e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15050
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.75,
      "epoch": 0.7272419791264012,
      "grad_norm": 0.0,
      "learning_rate": 1.3640316969462698e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15051
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 92.53125,
      "epoch": 0.7272902976420564,
      "grad_norm": 9.028377625432253,
      "learning_rate": 1.3637901043679937e-07,
      "loss": -0.0,
      "reward": 2.203125,
      "reward_std": 0.2961486577987671,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15052
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.9375,
      "epoch": 0.7273386161577117,
      "grad_norm": 0.0,
      "learning_rate": 1.3635485117897178e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15053
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.75,
      "epoch": 0.7273869346733668,
      "grad_norm": 0.0,
      "learning_rate": 1.363306919211442e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 15054
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.5,
      "epoch": 0.727435253189022,
      "grad_norm": 0.0,
      "learning_rate": 1.3630653266331656e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15055
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.40625,
      "epoch": 0.7274835717046773,
      "grad_norm": 0.0,
      "learning_rate": 1.3628237340548897e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15056
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 84.96875,
      "epoch": 0.7275318902203324,
      "grad_norm": 0.0,
      "learning_rate": 1.3625821414766136e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15057
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.40625,
      "epoch": 0.7275802087359876,
      "grad_norm": 0.0,
      "learning_rate": 1.3623405488983377e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15058
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.40625,
      "epoch": 0.7276285272516428,
      "grad_norm": 0.0,
      "learning_rate": 1.3620989563200619e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15059
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.125,
      "epoch": 0.7276768457672981,
      "grad_norm": 0.0,
      "learning_rate": 1.3618573637417857e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15060
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 93.9375,
      "epoch": 0.7277251642829532,
      "grad_norm": 17.006714775963005,
      "learning_rate": 1.3616157711635099e-07,
      "loss": 0.0,
      "reward": 2.234375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15061
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.65625,
      "epoch": 0.7277734827986084,
      "grad_norm": 0.0,
      "learning_rate": 1.3613741785852337e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15062
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.6875,
      "epoch": 0.7278218013142637,
      "grad_norm": 0.0,
      "learning_rate": 1.361132586006958e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15063
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.84375,
      "epoch": 0.7278701198299188,
      "grad_norm": 0.0,
      "learning_rate": 1.360890993428682e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15064
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 75.9375,
      "epoch": 0.727918438345574,
      "grad_norm": 17.298111897280492,
      "learning_rate": 1.360649400850406e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.21875,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 15065
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.625,
      "epoch": 0.7279667568612292,
      "grad_norm": 0.0,
      "learning_rate": 1.3604078082721297e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15066
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.8125,
      "epoch": 0.7280150753768844,
      "grad_norm": 0.0,
      "learning_rate": 1.3601662156938536e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15067
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 84.25,
      "epoch": 0.7280633938925396,
      "grad_norm": 5.686103783742272,
      "learning_rate": 1.3599246231155778e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.23421898484230042,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15068
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.78125,
      "epoch": 0.7281117124081948,
      "grad_norm": 0.0,
      "learning_rate": 1.359683030537302e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15069
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.46875,
      "epoch": 0.7281600309238501,
      "grad_norm": 0.0,
      "learning_rate": 1.3594414379590258e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15070
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.03125,
      "epoch": 0.7282083494395052,
      "grad_norm": 0.0,
      "learning_rate": 1.35919984538075e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15071
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.59375,
      "epoch": 0.7282566679551604,
      "grad_norm": 0.0,
      "learning_rate": 1.3589582528024738e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15072
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.34375,
      "epoch": 0.7283049864708157,
      "grad_norm": 0.0,
      "learning_rate": 1.358716660224198e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15073
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.84375,
      "epoch": 0.7283533049864708,
      "grad_norm": 0.0,
      "learning_rate": 1.358475067645922e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15074
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.90625,
      "epoch": 0.728401623502126,
      "grad_norm": 0.0,
      "learning_rate": 1.358233475067646e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15075
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.125,
      "epoch": 0.7284499420177812,
      "grad_norm": 0.0,
      "learning_rate": 1.3579918824893698e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15076
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.9375,
      "epoch": 0.7284982605334364,
      "grad_norm": 0.0,
      "learning_rate": 1.3577502899110937e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15077
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.78125,
      "epoch": 0.7285465790490916,
      "grad_norm": 0.0,
      "learning_rate": 1.3575086973328178e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15078
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.46875,
      "epoch": 0.7285948975647468,
      "grad_norm": 0.0,
      "learning_rate": 1.357267104754542e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15079
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.46875,
      "epoch": 0.7286432160804021,
      "grad_norm": 0.0,
      "learning_rate": 1.3570255121762658e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15080
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.59375,
      "epoch": 0.7286915345960572,
      "grad_norm": 0.0,
      "learning_rate": 1.35678391959799e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 15081
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.9375,
      "epoch": 0.7287398531117124,
      "grad_norm": 0.0,
      "learning_rate": 1.3565423270197138e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15082
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.9375,
      "epoch": 0.7287881716273676,
      "grad_norm": 0.0,
      "learning_rate": 1.356300734441438e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15083
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 84.6875,
      "epoch": 0.7288364901430228,
      "grad_norm": 0.0,
      "learning_rate": 1.356059141863162e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15084
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.0625,
      "epoch": 0.728884808658678,
      "grad_norm": 0.0,
      "learning_rate": 1.355817549284886e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15085
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.59375,
      "epoch": 0.7289331271743332,
      "grad_norm": 0.0,
      "learning_rate": 1.3555759567066098e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 15086
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 80.09375,
      "epoch": 0.7289814456899883,
      "grad_norm": 0.0,
      "learning_rate": 1.3553343641283337e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15087
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.71875,
      "epoch": 0.7290297642056436,
      "grad_norm": 0.0,
      "learning_rate": 1.3550927715500578e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15088
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 86.96875,
      "epoch": 0.7290780827212988,
      "grad_norm": 11.201844142407934,
      "learning_rate": 1.354851178971782e-07,
      "loss": 0.0,
      "reward": 1.90625,
      "reward_std": 0.25282126665115356,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15089
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.15625,
      "epoch": 0.729126401236954,
      "grad_norm": 0.0,
      "learning_rate": 1.3546095863935058e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15090
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.96875,
      "epoch": 0.7291747197526092,
      "grad_norm": 0.0,
      "learning_rate": 1.35436799381523e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15091
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 73.25,
      "epoch": 0.7292230382682644,
      "grad_norm": 0.0,
      "learning_rate": 1.3541264012369539e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15092
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.3125,
      "epoch": 0.7292713567839196,
      "grad_norm": 0.0,
      "learning_rate": 1.353884808658678e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15093
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.09375,
      "epoch": 0.7293196752995748,
      "grad_norm": 0.0,
      "learning_rate": 1.353643216080402e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15094
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.625,
      "epoch": 0.72936799381523,
      "grad_norm": 0.0,
      "learning_rate": 1.353401623502126e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15095
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.125,
      "epoch": 0.7294163123308852,
      "grad_norm": 0.0,
      "learning_rate": 1.35316003092385e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15096
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 78.46875,
      "epoch": 0.7294646308465403,
      "grad_norm": 0.0,
      "learning_rate": 1.3529184383455737e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15097
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.8125,
      "epoch": 0.7295129493621956,
      "grad_norm": 0.0,
      "learning_rate": 1.352676845767298e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15098
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.3125,
      "epoch": 0.7295612678778508,
      "grad_norm": 0.0,
      "learning_rate": 1.352435253189022e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15099
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.1875,
      "epoch": 0.729609586393506,
      "grad_norm": 0.0,
      "learning_rate": 1.352193660610746e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15100
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.40625,
      "epoch": 0.7296579049091612,
      "grad_norm": 0.0,
      "learning_rate": 1.35195206803247e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15101
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.84375,
      "epoch": 0.7297062234248164,
      "grad_norm": 0.0,
      "learning_rate": 1.351710475454194e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15102
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.25,
      "epoch": 0.7297545419404716,
      "grad_norm": 0.0,
      "learning_rate": 1.351468882875918e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15103
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.5625,
      "epoch": 0.7298028604561267,
      "grad_norm": 0.0,
      "learning_rate": 1.3512272902976422e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15104
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.9375,
      "epoch": 0.729851178971782,
      "grad_norm": 0.0,
      "learning_rate": 1.350985697719366e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15105
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.96875,
      "epoch": 0.7298994974874372,
      "grad_norm": 0.0,
      "learning_rate": 1.3507441051410902e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15106
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 81.40625,
      "epoch": 0.7299478160030923,
      "grad_norm": 0.0,
      "learning_rate": 1.3505025125628138e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15107
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.09375,
      "epoch": 0.7299961345187476,
      "grad_norm": 0.0,
      "learning_rate": 1.350260919984538e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15108
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.46875,
      "epoch": 0.7300444530344028,
      "grad_norm": 0.0,
      "learning_rate": 1.350019327406262e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15109
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.90625,
      "epoch": 0.730092771550058,
      "grad_norm": 0.0,
      "learning_rate": 1.349777734827986e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15110
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.0,
      "epoch": 0.7301410900657131,
      "grad_norm": 0.0,
      "learning_rate": 1.34953614224971e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15111
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.71875,
      "epoch": 0.7301894085813684,
      "grad_norm": 0.0,
      "learning_rate": 1.349294549671434e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15112
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 106.71875,
      "epoch": 0.7302377270970236,
      "grad_norm": 11.370972660344885,
      "learning_rate": 1.349052957093158e-07,
      "loss": -0.0,
      "reward": 2.0625,
      "reward_std": 0.3061639666557312,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15113
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 73.5625,
      "epoch": 0.7302860456126787,
      "grad_norm": 0.0,
      "learning_rate": 1.3488113645148822e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 15114
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.875,
      "epoch": 0.730334364128334,
      "grad_norm": 0.0,
      "learning_rate": 1.348569771936606e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15115
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.65625,
      "epoch": 0.7303826826439892,
      "grad_norm": 0.0,
      "learning_rate": 1.3483281793583302e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15116
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 75.8125,
      "epoch": 0.7304310011596443,
      "grad_norm": 0.0,
      "learning_rate": 1.3480865867800538e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15117
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 81.65625,
      "epoch": 0.7304793196752996,
      "grad_norm": 0.0,
      "learning_rate": 1.347844994201778e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15118
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.15625,
      "epoch": 0.7305276381909548,
      "grad_norm": 0.0,
      "learning_rate": 1.347603401623502e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 15119
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.71875,
      "epoch": 0.73057595670661,
      "grad_norm": 0.0,
      "learning_rate": 1.347361809045226e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15120
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.3125,
      "epoch": 0.7306242752222651,
      "grad_norm": 0.0,
      "learning_rate": 1.34712021646695e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15121
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.3125,
      "epoch": 0.7306725937379204,
      "grad_norm": 0.0,
      "learning_rate": 1.346878623888674e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15122
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.59375,
      "epoch": 0.7307209122535756,
      "grad_norm": 0.0,
      "learning_rate": 1.346637031310398e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15123
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 101.03125,
      "epoch": 0.7307692307692307,
      "grad_norm": 0.0,
      "learning_rate": 1.3463954387321222e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15124
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.71875,
      "epoch": 0.730817549284886,
      "grad_norm": 0.0,
      "learning_rate": 1.346153846153846e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15125
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.875,
      "epoch": 0.7308658678005412,
      "grad_norm": 0.0,
      "learning_rate": 1.3459122535755702e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15126
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 82.3125,
      "epoch": 0.7309141863161963,
      "grad_norm": 0.0,
      "learning_rate": 1.3456706609972939e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15127
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.46875,
      "epoch": 0.7309625048318515,
      "grad_norm": 0.0,
      "learning_rate": 1.345429068419018e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15128
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.03125,
      "epoch": 0.7310108233475068,
      "grad_norm": 0.0,
      "learning_rate": 1.345187475840742e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15129
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.625,
      "epoch": 0.731059141863162,
      "grad_norm": 0.0,
      "learning_rate": 1.344945883262466e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15130
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.3125,
      "epoch": 0.7311074603788171,
      "grad_norm": 0.0,
      "learning_rate": 1.3447042906841901e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15131
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.9375,
      "epoch": 0.7311557788944724,
      "grad_norm": 0.0,
      "learning_rate": 1.344462698105914e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15132
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.75,
      "epoch": 0.7312040974101276,
      "grad_norm": 0.0,
      "learning_rate": 1.3442211055276381e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2198973000049591,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15133
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.125,
      "epoch": 0.7312524159257827,
      "grad_norm": 0.0,
      "learning_rate": 1.3439795129493623e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15134
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.90625,
      "epoch": 0.731300734441438,
      "grad_norm": 0.0,
      "learning_rate": 1.3437379203710862e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15135
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.25,
      "epoch": 0.7313490529570932,
      "grad_norm": 0.0,
      "learning_rate": 1.3434963277928103e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15136
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.125,
      "epoch": 0.7313973714727483,
      "grad_norm": 0.0,
      "learning_rate": 1.3432547352145342e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15137
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.4375,
      "epoch": 0.7314456899884035,
      "grad_norm": 0.0,
      "learning_rate": 1.343013142636258e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15138
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.96875,
      "epoch": 0.7314940085040588,
      "grad_norm": 19.004637321199954,
      "learning_rate": 1.3427715500579822e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15139
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.96875,
      "epoch": 0.731542327019714,
      "grad_norm": 0.0,
      "learning_rate": 1.342529957479706e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15140
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.40625,
      "epoch": 0.7315906455353691,
      "grad_norm": 0.0,
      "learning_rate": 1.3422883649014302e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 15141
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.5,
      "epoch": 0.7316389640510244,
      "grad_norm": 0.0,
      "learning_rate": 1.342046772323154e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15142
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.40625,
      "epoch": 0.7316872825666796,
      "grad_norm": 0.0,
      "learning_rate": 1.3418051797448782e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 15143
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.65625,
      "epoch": 0.7317356010823347,
      "grad_norm": 0.0,
      "learning_rate": 1.3415635871666023e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 15144
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.0,
      "epoch": 0.7317839195979899,
      "grad_norm": 0.0,
      "learning_rate": 1.3413219945883262e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 15145
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.875,
      "epoch": 0.7318322381136452,
      "grad_norm": 0.0,
      "learning_rate": 1.3410804020100503e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15146
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.6875,
      "epoch": 0.7318805566293003,
      "grad_norm": 0.0,
      "learning_rate": 1.3408388094317742e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15147
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.3125,
      "epoch": 0.7319288751449555,
      "grad_norm": 0.0,
      "learning_rate": 1.340597216853498e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15148
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.875,
      "epoch": 0.7319771936606108,
      "grad_norm": 0.0,
      "learning_rate": 1.3403556242752222e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15149
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 107.1875,
      "epoch": 0.732025512176266,
      "grad_norm": 0.0,
      "learning_rate": 1.340114031696946e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 15150
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.0625,
      "epoch": 0.7320738306919211,
      "grad_norm": 0.0,
      "learning_rate": 1.3398724391186702e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15151
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 79.21875,
      "epoch": 0.7321221492075763,
      "grad_norm": 23.744665776739264,
      "learning_rate": 1.339630846540394e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.285190224647522,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15152
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.34375,
      "epoch": 0.7321704677232316,
      "grad_norm": 0.0,
      "learning_rate": 1.3393892539621182e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15153
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.9375,
      "epoch": 0.7322187862388867,
      "grad_norm": 0.0,
      "learning_rate": 1.3391476613838424e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15154
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.09375,
      "epoch": 0.7322671047545419,
      "grad_norm": 0.0,
      "learning_rate": 1.3389060688055662e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15155
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.21875,
      "epoch": 0.7323154232701972,
      "grad_norm": 0.0,
      "learning_rate": 1.3386644762272904e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.2177756428718567,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15156
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.5625,
      "epoch": 0.7323637417858523,
      "grad_norm": 0.0,
      "learning_rate": 1.3384228836490142e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15157
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.25,
      "epoch": 0.7324120603015075,
      "grad_norm": 0.0,
      "learning_rate": 1.338181291070738e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15158
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.46875,
      "epoch": 0.7324603788171627,
      "grad_norm": 0.0,
      "learning_rate": 1.3379396984924622e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15159
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.65625,
      "epoch": 0.732508697332818,
      "grad_norm": 0.0,
      "learning_rate": 1.337698105914186e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15160
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.0,
      "epoch": 0.7325570158484731,
      "grad_norm": 0.0,
      "learning_rate": 1.3374565133359103e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.24464009702205658,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15161
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.625,
      "epoch": 0.7326053343641283,
      "grad_norm": 0.0,
      "learning_rate": 1.337214920757634e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23568855226039886,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15162
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.84375,
      "epoch": 0.7326536528797836,
      "grad_norm": 0.0,
      "learning_rate": 1.3369733281793583e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15163
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.21875,
      "epoch": 0.7327019713954387,
      "grad_norm": 0.0,
      "learning_rate": 1.3367317356010824e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15164
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.25,
      "epoch": 0.7327502899110939,
      "grad_norm": 0.0,
      "learning_rate": 1.3364901430228063e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15165
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.09375,
      "epoch": 0.7327986084267492,
      "grad_norm": 0.0,
      "learning_rate": 1.3362485504445304e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15166
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.15625,
      "epoch": 0.7328469269424043,
      "grad_norm": 0.0,
      "learning_rate": 1.3360069578662543e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15167
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.28125,
      "epoch": 0.7328952454580595,
      "grad_norm": 0.0,
      "learning_rate": 1.3357653652879784e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.260896235704422,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15168
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.4375,
      "epoch": 0.7329435639737147,
      "grad_norm": 0.0,
      "learning_rate": 1.3355237727097023e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15169
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.09375,
      "epoch": 0.73299188248937,
      "grad_norm": 0.0,
      "learning_rate": 1.3352821801314262e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15170
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.84375,
      "epoch": 0.7330402010050251,
      "grad_norm": 0.0,
      "learning_rate": 1.3350405875531503e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15171
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.59375,
      "epoch": 0.7330885195206803,
      "grad_norm": 0.0,
      "learning_rate": 1.3347989949748742e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15172
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.15625,
      "epoch": 0.7331368380363356,
      "grad_norm": 0.0,
      "learning_rate": 1.3345574023965983e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15173
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.8125,
      "epoch": 0.7331851565519907,
      "grad_norm": 0.0,
      "learning_rate": 1.3343158098183224e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15174
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.09375,
      "epoch": 0.7332334750676459,
      "grad_norm": 0.0,
      "learning_rate": 1.3340742172400463e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15175
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.53125,
      "epoch": 0.7332817935833011,
      "grad_norm": 0.0,
      "learning_rate": 1.3338326246617704e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24251842498779297,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15176
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.3125,
      "epoch": 0.7333301120989563,
      "grad_norm": 0.0,
      "learning_rate": 1.3335910320834943e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15177
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.6875,
      "epoch": 0.7333784306146115,
      "grad_norm": 0.0,
      "learning_rate": 1.3333494395052185e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15178
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 79.6875,
      "epoch": 0.7334267491302667,
      "grad_norm": 0.0,
      "learning_rate": 1.3331078469269423e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 15179
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.5,
      "epoch": 0.733475067645922,
      "grad_norm": 0.0,
      "learning_rate": 1.3328662543486662e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15180
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.0625,
      "epoch": 0.7335233861615771,
      "grad_norm": 0.0,
      "learning_rate": 1.3326246617703903e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15181
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.4375,
      "epoch": 0.7335717046772323,
      "grad_norm": 0.0,
      "learning_rate": 1.3323830691921142e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15182
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.6875,
      "epoch": 0.7336200231928875,
      "grad_norm": 0.0,
      "learning_rate": 1.3321414766138383e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15183
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.03125,
      "epoch": 0.7336683417085427,
      "grad_norm": 0.0,
      "learning_rate": 1.3318998840355625e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15184
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 78.15625,
      "epoch": 0.7337166602241979,
      "grad_norm": 0.0,
      "learning_rate": 1.3316582914572863e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15185
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 82.625,
      "epoch": 0.7337649787398531,
      "grad_norm": 0.0,
      "learning_rate": 1.3314166988790105e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15186
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 92.34375,
      "epoch": 0.7338132972555083,
      "grad_norm": 4.343762848461193,
      "learning_rate": 1.3311751063007344e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.2505248188972473,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15187
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.78125,
      "epoch": 0.7338616157711635,
      "grad_norm": 0.0,
      "learning_rate": 1.3309335137224585e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15188
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.1875,
      "epoch": 0.7339099342868187,
      "grad_norm": 0.0,
      "learning_rate": 1.3306919211441824e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15189
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 102.34375,
      "epoch": 0.733958252802474,
      "grad_norm": 9.927974132494706,
      "learning_rate": 1.3304503285659062e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.17598573863506317,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15190
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.125,
      "epoch": 0.7340065713181291,
      "grad_norm": 0.0,
      "learning_rate": 1.3302087359876304e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15191
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 91.96875,
      "epoch": 0.7340548898337843,
      "grad_norm": 0.0,
      "learning_rate": 1.3299671434093542e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 15192
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.15625,
      "epoch": 0.7341032083494395,
      "grad_norm": 0.0,
      "learning_rate": 1.3297255508310784e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15193
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.40625,
      "epoch": 0.7341515268650947,
      "grad_norm": 0.0,
      "learning_rate": 1.3294839582528025e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15194
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.46875,
      "epoch": 0.7341998453807499,
      "grad_norm": 0.0,
      "learning_rate": 1.3292423656745264e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15195
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 81.46875,
      "epoch": 0.7342481638964051,
      "grad_norm": 0.0,
      "learning_rate": 1.3290007730962505e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15196
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 104.25,
      "epoch": 0.7342964824120602,
      "grad_norm": 0.0,
      "learning_rate": 1.3287591805179744e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 15197
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.28125,
      "epoch": 0.7343448009277155,
      "grad_norm": 0.0,
      "learning_rate": 1.3285175879396985e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15198
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.78125,
      "epoch": 0.7343931194433707,
      "grad_norm": 0.0,
      "learning_rate": 1.3282759953614227e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15199
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 89.8125,
      "epoch": 0.734441437959026,
      "grad_norm": 12.948450611239414,
      "learning_rate": 1.3280344027831463e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15200
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.6875,
      "epoch": 0.7344897564746811,
      "grad_norm": 0.0,
      "learning_rate": 1.3277928102048704e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15201
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.59375,
      "epoch": 0.7345380749903363,
      "grad_norm": 0.0,
      "learning_rate": 1.3275512176265943e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 15202
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.96875,
      "epoch": 0.7345863935059915,
      "grad_norm": 0.0,
      "learning_rate": 1.3273096250483184e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.2561880350112915,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15203
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 96.875,
      "epoch": 0.7346347120216467,
      "grad_norm": 10.558490111907014,
      "learning_rate": 1.3270680324700426e-07,
      "loss": 0.0,
      "reward": 2.203125,
      "reward_std": 0.36323678493499756,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15204
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 75.84375,
      "epoch": 0.7346830305373019,
      "grad_norm": 0.0,
      "learning_rate": 1.3268264398917664e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15205
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.5625,
      "epoch": 0.7347313490529571,
      "grad_norm": 0.0,
      "learning_rate": 1.3265848473134906e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15206
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.0625,
      "epoch": 0.7347796675686122,
      "grad_norm": 0.0,
      "learning_rate": 1.3263432547352144e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15207
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.125,
      "epoch": 0.7348279860842675,
      "grad_norm": 0.0,
      "learning_rate": 1.3261016621569386e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15208
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 79.71875,
      "epoch": 0.7348763045999227,
      "grad_norm": 0.0,
      "learning_rate": 1.3258600695786627e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15209
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.4375,
      "epoch": 0.7349246231155779,
      "grad_norm": 0.0,
      "learning_rate": 1.3256184770003863e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15210
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.875,
      "epoch": 0.7349729416312331,
      "grad_norm": 0.0,
      "learning_rate": 1.3253768844221105e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15211
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.5625,
      "epoch": 0.7350212601468883,
      "grad_norm": 0.0,
      "learning_rate": 1.3251352918438343e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15212
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.59375,
      "epoch": 0.7350695786625435,
      "grad_norm": 0.0,
      "learning_rate": 1.3248936992655585e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15213
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.3125,
      "epoch": 0.7351178971781986,
      "grad_norm": 0.0,
      "learning_rate": 1.3246521066872826e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15214
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.15625,
      "epoch": 0.7351662156938539,
      "grad_norm": 0.0,
      "learning_rate": 1.3244105141090065e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15215
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.0625,
      "epoch": 0.7352145342095091,
      "grad_norm": 0.0,
      "learning_rate": 1.3241689215307306e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15216
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.59375,
      "epoch": 0.7352628527251642,
      "grad_norm": 0.0,
      "learning_rate": 1.3239273289524545e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15217
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.1875,
      "epoch": 0.7353111712408195,
      "grad_norm": 0.0,
      "learning_rate": 1.3236857363741786e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15218
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 77.375,
      "epoch": 0.7353594897564747,
      "grad_norm": 0.0,
      "learning_rate": 1.3234441437959027e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15219
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.6875,
      "epoch": 0.7354078082721299,
      "grad_norm": 0.0,
      "learning_rate": 1.3232025512176264e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15220
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.03125,
      "epoch": 0.735456126787785,
      "grad_norm": 0.0,
      "learning_rate": 1.3229609586393505e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15221
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.59375,
      "epoch": 0.7355044453034403,
      "grad_norm": 0.0,
      "learning_rate": 1.3227193660610744e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15222
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 76.5,
      "epoch": 0.7355527638190955,
      "grad_norm": 0.0,
      "learning_rate": 1.3224777734827985e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15223
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 109.78125,
      "epoch": 0.7356010823347506,
      "grad_norm": 138.54965772811974,
      "learning_rate": 1.3222361809045226e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.2993341088294983,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15224
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.5,
      "epoch": 0.7356494008504059,
      "grad_norm": 0.0,
      "learning_rate": 1.3219945883262465e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15225
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.71875,
      "epoch": 0.7356977193660611,
      "grad_norm": 0.0,
      "learning_rate": 1.3217529957479706e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15226
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 78.75,
      "epoch": 0.7357460378817162,
      "grad_norm": 0.0,
      "learning_rate": 1.3215114031696945e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15227
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.75,
      "epoch": 0.7357943563973715,
      "grad_norm": 0.0,
      "learning_rate": 1.3212698105914187e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15228
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.5,
      "epoch": 0.7358426749130267,
      "grad_norm": 0.0,
      "learning_rate": 1.3210282180131428e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15229
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.25,
      "epoch": 0.7358909934286819,
      "grad_norm": 0.0,
      "learning_rate": 1.3207866254348664e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 15230
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.28125,
      "epoch": 0.735939311944337,
      "grad_norm": 0.0,
      "learning_rate": 1.3205450328565905e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15231
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 80.53125,
      "epoch": 0.7359876304599923,
      "grad_norm": 0.0,
      "learning_rate": 1.3203034402783144e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15232
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.78125,
      "epoch": 0.7360359489756475,
      "grad_norm": 0.0,
      "learning_rate": 1.3200618477000385e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15233
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.28125,
      "epoch": 0.7360842674913026,
      "grad_norm": 0.0,
      "learning_rate": 1.3198202551217627e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15234
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 90.28125,
      "epoch": 0.7361325860069579,
      "grad_norm": 0.0,
      "learning_rate": 1.3195786625434865e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15235
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 78.21875,
      "epoch": 0.7361809045226131,
      "grad_norm": 0.0,
      "learning_rate": 1.3193370699652107e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15236
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.6875,
      "epoch": 0.7362292230382682,
      "grad_norm": 0.0,
      "learning_rate": 1.3190954773869346e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15237
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.125,
      "epoch": 0.7362775415539234,
      "grad_norm": 0.0,
      "learning_rate": 1.3188538848086587e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15238
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.625,
      "epoch": 0.7363258600695787,
      "grad_norm": 0.0,
      "learning_rate": 1.3186122922303828e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15239
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.5,
      "epoch": 0.7363741785852339,
      "grad_norm": 0.0,
      "learning_rate": 1.3183706996521067e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15240
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.96875,
      "epoch": 0.736422497100889,
      "grad_norm": 0.0,
      "learning_rate": 1.3181291070738306e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15241
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 105.75,
      "epoch": 0.7364708156165443,
      "grad_norm": 0.0,
      "learning_rate": 1.3178875144955544e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 15242
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.40625,
      "epoch": 0.7365191341321995,
      "grad_norm": 0.0,
      "learning_rate": 1.3176459219172786e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15243
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.4375,
      "epoch": 0.7365674526478546,
      "grad_norm": 0.0,
      "learning_rate": 1.3174043293390027e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15244
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.28125,
      "epoch": 0.7366157711635098,
      "grad_norm": 0.0,
      "learning_rate": 1.3171627367607266e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15245
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 100.21875,
      "epoch": 0.7366640896791651,
      "grad_norm": 0.0,
      "learning_rate": 1.3169211441824507e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15246
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.96875,
      "epoch": 0.7367124081948202,
      "grad_norm": 0.0,
      "learning_rate": 1.3166795516041746e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15247
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.40625,
      "epoch": 0.7367607267104754,
      "grad_norm": 0.0,
      "learning_rate": 1.3164379590258987e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15248
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 79.875,
      "epoch": 0.7368090452261307,
      "grad_norm": 0.0,
      "learning_rate": 1.3161963664476229e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15249
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 90.84375,
      "epoch": 0.7368573637417859,
      "grad_norm": 0.0,
      "learning_rate": 1.3159547738693467e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15250
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.9375,
      "epoch": 0.736905682257441,
      "grad_norm": 0.0,
      "learning_rate": 1.3157131812910706e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15251
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 83.6875,
      "epoch": 0.7369540007730963,
      "grad_norm": 0.0,
      "learning_rate": 1.3154715887127945e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15252
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.9375,
      "epoch": 0.7370023192887515,
      "grad_norm": 0.0,
      "learning_rate": 1.3152299961345186e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15253
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 94.3125,
      "epoch": 0.7370506378044066,
      "grad_norm": 3.5449171332341707,
      "learning_rate": 1.3149884035562428e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15254
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 87.71875,
      "epoch": 0.7370989563200618,
      "grad_norm": 0.0,
      "learning_rate": 1.3147468109779666e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15255
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.65625,
      "epoch": 0.7371472748357171,
      "grad_norm": 0.0,
      "learning_rate": 1.3145052183996908e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15256
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.4375,
      "epoch": 0.7371955933513722,
      "grad_norm": 0.0,
      "learning_rate": 1.3142636258214146e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15257
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 76.5,
      "epoch": 0.7372439118670274,
      "grad_norm": 0.0,
      "learning_rate": 1.3140220332431388e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 15258
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.1875,
      "epoch": 0.7372922303826827,
      "grad_norm": 0.0,
      "learning_rate": 1.313780440664863e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15259
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.65625,
      "epoch": 0.7373405488983379,
      "grad_norm": 0.0,
      "learning_rate": 1.3135388480865868e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15260
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 77.21875,
      "epoch": 0.737388867413993,
      "grad_norm": 0.0,
      "learning_rate": 1.3132972555083106e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15261
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.65625,
      "epoch": 0.7374371859296482,
      "grad_norm": 0.0,
      "learning_rate": 1.3130556629300345e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15262
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 80.3125,
      "epoch": 0.7374855044453035,
      "grad_norm": 0.0,
      "learning_rate": 1.3128140703517587e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15263
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.15625,
      "epoch": 0.7375338229609586,
      "grad_norm": 0.0,
      "learning_rate": 1.3125724777734828e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15264
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.28125,
      "epoch": 0.7375821414766138,
      "grad_norm": 0.0,
      "learning_rate": 1.3123308851952067e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15265
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.3125,
      "epoch": 0.7376304599922691,
      "grad_norm": 0.0,
      "learning_rate": 1.3120892926169308e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15266
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.375,
      "epoch": 0.7376787785079242,
      "grad_norm": 0.0,
      "learning_rate": 1.3118477000386547e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15267
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.0625,
      "epoch": 0.7377270970235794,
      "grad_norm": 0.0,
      "learning_rate": 1.3116061074603788e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15268
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 79.125,
      "epoch": 0.7377754155392346,
      "grad_norm": 0.0,
      "learning_rate": 1.311364514882103e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15269
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 112.9375,
      "epoch": 0.7378237340548899,
      "grad_norm": 6.672818320791891,
      "learning_rate": 1.3111229223038268e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15270
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.21875,
      "epoch": 0.737872052570545,
      "grad_norm": 0.0,
      "learning_rate": 1.310881329725551e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15271
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.15625,
      "epoch": 0.7379203710862002,
      "grad_norm": 0.0,
      "learning_rate": 1.3106397371472746e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15272
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 90.5625,
      "epoch": 0.7379686896018555,
      "grad_norm": 0.0,
      "learning_rate": 1.3103981445689987e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 15273
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.21875,
      "epoch": 0.7380170081175106,
      "grad_norm": 0.0,
      "learning_rate": 1.3101565519907228e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15274
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 80.3125,
      "epoch": 0.7380653266331658,
      "grad_norm": 0.0,
      "learning_rate": 1.3099149594124467e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15275
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.3125,
      "epoch": 0.738113645148821,
      "grad_norm": 0.0,
      "learning_rate": 1.3096733668341708e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15276
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 86.65625,
      "epoch": 0.7381619636644762,
      "grad_norm": 9.697508353415804,
      "learning_rate": 1.3094317742558947e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.22738909721374512,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15277
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.125,
      "epoch": 0.7382102821801314,
      "grad_norm": 0.0,
      "learning_rate": 1.3091901816776188e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15278
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.0625,
      "epoch": 0.7382586006957866,
      "grad_norm": 0.0,
      "learning_rate": 1.308948589099343e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15279
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 111.3125,
      "epoch": 0.7383069192114419,
      "grad_norm": 4.364695690694747,
      "learning_rate": 1.3087069965210669e-07,
      "loss": -0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15280
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.15625,
      "epoch": 0.738355237727097,
      "grad_norm": 0.0,
      "learning_rate": 1.308465403942791e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15281
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 75.46875,
      "epoch": 0.7384035562427522,
      "grad_norm": 10.583634297888908,
      "learning_rate": 1.3082238113645146e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.3133465051651001,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 0.96875,
      "step": 15282
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.4375,
      "epoch": 0.7384518747584075,
      "grad_norm": 0.0,
      "learning_rate": 1.3079822187862387e-07,
      "loss": 0.0,
      "reward": 2.296875,
      "reward_std": 0.260896235704422,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15283
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.9375,
      "epoch": 0.7385001932740626,
      "grad_norm": 0.0,
      "learning_rate": 1.307740626207963e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15284
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 85.25,
      "epoch": 0.7385485117897178,
      "grad_norm": 0.0,
      "learning_rate": 1.3074990336296867e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15285
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.5,
      "epoch": 0.738596830305373,
      "grad_norm": 0.0,
      "learning_rate": 1.307257441051411e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15286
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.21875,
      "epoch": 0.7386451488210282,
      "grad_norm": 0.0,
      "learning_rate": 1.3070158484731348e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15287
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.96875,
      "epoch": 0.7386934673366834,
      "grad_norm": 0.0,
      "learning_rate": 1.306774255894859e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15288
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 129.78125,
      "epoch": 0.7387417858523386,
      "grad_norm": 0.0,
      "learning_rate": 1.306532663316583e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15289
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.46875,
      "epoch": 0.7387901043679939,
      "grad_norm": 0.0,
      "learning_rate": 1.306291070738307e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15290
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.84375,
      "epoch": 0.738838422883649,
      "grad_norm": 0.0,
      "learning_rate": 1.306049478160031e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15291
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 74.875,
      "epoch": 0.7388867413993042,
      "grad_norm": 0.0,
      "learning_rate": 1.3058078855817546e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15292
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.84375,
      "epoch": 0.7389350599149594,
      "grad_norm": 0.0,
      "learning_rate": 1.3055662930034788e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15293
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 83.6875,
      "epoch": 0.7389833784306146,
      "grad_norm": 0.0,
      "learning_rate": 1.305324700425203e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15294
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.5625,
      "epoch": 0.7390316969462698,
      "grad_norm": 0.0,
      "learning_rate": 1.3050831078469268e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15295
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 77.78125,
      "epoch": 0.739080015461925,
      "grad_norm": 0.0,
      "learning_rate": 1.304841515268651e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15296
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.28125,
      "epoch": 0.7391283339775802,
      "grad_norm": 0.0,
      "learning_rate": 1.3045999226903748e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15297
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.1875,
      "epoch": 0.7391766524932354,
      "grad_norm": 0.0,
      "learning_rate": 1.304358330112099e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15298
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.3125,
      "epoch": 0.7392249710088906,
      "grad_norm": 16.602597978715796,
      "learning_rate": 1.304116737533823e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12387890368700027,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15299
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.6875,
      "epoch": 0.7392732895245459,
      "grad_norm": 0.0,
      "learning_rate": 1.303875144955547e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15300
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.90625,
      "epoch": 0.739321608040201,
      "grad_norm": 0.0,
      "learning_rate": 1.303633552377271e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15301
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.84375,
      "epoch": 0.7393699265558562,
      "grad_norm": 0.0,
      "learning_rate": 1.3033919597989947e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15302
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.875,
      "epoch": 0.7394182450715114,
      "grad_norm": 0.0,
      "learning_rate": 1.3031503672207188e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15303
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 117.65625,
      "epoch": 0.7394665635871666,
      "grad_norm": 6.302188011149146,
      "learning_rate": 1.302908774642443e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.3104073107242584,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15304
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.125,
      "epoch": 0.7395148821028218,
      "grad_norm": 0.0,
      "learning_rate": 1.3026671820641668e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15305
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.1875,
      "epoch": 0.739563200618477,
      "grad_norm": 0.0,
      "learning_rate": 1.302425589485891e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15306
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 104.9375,
      "epoch": 0.7396115191341321,
      "grad_norm": 5.9341944939161015,
      "learning_rate": 1.3021839969076148e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15307
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 85.65625,
      "epoch": 0.7396598376497874,
      "grad_norm": 0.0,
      "learning_rate": 1.301942404329339e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15308
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.09375,
      "epoch": 0.7397081561654426,
      "grad_norm": 0.0,
      "learning_rate": 1.301700811751063e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15309
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.9375,
      "epoch": 0.7397564746810978,
      "grad_norm": 0.0,
      "learning_rate": 1.301459219172787e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15310
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.5,
      "epoch": 0.739804793196753,
      "grad_norm": 0.0,
      "learning_rate": 1.301217626594511e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15311
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.5625,
      "epoch": 0.7398531117124082,
      "grad_norm": 0.0,
      "learning_rate": 1.3009760340162347e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15312
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.3125,
      "epoch": 0.7399014302280634,
      "grad_norm": 0.0,
      "learning_rate": 1.3007344414379589e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15313
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.4375,
      "epoch": 0.7399497487437185,
      "grad_norm": 0.0,
      "learning_rate": 1.300492848859683e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15314
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.6875,
      "epoch": 0.7399980672593738,
      "grad_norm": 0.0,
      "learning_rate": 1.3002512562814069e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15315
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 99.34375,
      "epoch": 0.740046385775029,
      "grad_norm": 0.0,
      "learning_rate": 1.300009663703131e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15316
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.96875,
      "epoch": 0.7400947042906841,
      "grad_norm": 0.0,
      "learning_rate": 1.299768071124855e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15317
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.0625,
      "epoch": 0.7401430228063394,
      "grad_norm": 0.0,
      "learning_rate": 1.299526478546579e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15318
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.03125,
      "epoch": 0.7401913413219946,
      "grad_norm": 0.0,
      "learning_rate": 1.2992848859683031e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15319
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.90625,
      "epoch": 0.7402396598376498,
      "grad_norm": 0.0,
      "learning_rate": 1.299043293390027e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15320
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.34375,
      "epoch": 0.740287978353305,
      "grad_norm": 0.0,
      "learning_rate": 1.2988017008117511e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15321
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.0625,
      "epoch": 0.7403362968689602,
      "grad_norm": 0.0,
      "learning_rate": 1.298560108233475e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15322
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.78125,
      "epoch": 0.7403846153846154,
      "grad_norm": 0.0,
      "learning_rate": 1.298318515655199e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15323
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 83.40625,
      "epoch": 0.7404329339002705,
      "grad_norm": 0.0,
      "learning_rate": 1.298076923076923e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15324
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.90625,
      "epoch": 0.7404812524159258,
      "grad_norm": 0.0,
      "learning_rate": 1.297835330498647e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15325
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.625,
      "epoch": 0.740529570931581,
      "grad_norm": 0.0,
      "learning_rate": 1.297593737920371e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887325048446655,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15326
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.875,
      "epoch": 0.7405778894472361,
      "grad_norm": 0.0,
      "learning_rate": 1.297352145342095e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15327
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.5625,
      "epoch": 0.7406262079628914,
      "grad_norm": 0.0,
      "learning_rate": 1.297110552763819e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15328
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 88.34375,
      "epoch": 0.7406745264785466,
      "grad_norm": 0.0,
      "learning_rate": 1.2968689601855432e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15329
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.46875,
      "epoch": 0.7407228449942018,
      "grad_norm": 0.0,
      "learning_rate": 1.296627367607267e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15330
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.46875,
      "epoch": 0.740771163509857,
      "grad_norm": 0.0,
      "learning_rate": 1.2963857750289912e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15331
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 120.46875,
      "epoch": 0.7408194820255122,
      "grad_norm": 0.0,
      "learning_rate": 1.296144182450715e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15332
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.875,
      "epoch": 0.7408678005411674,
      "grad_norm": 0.0,
      "learning_rate": 1.295902589872439e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15333
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 85.875,
      "epoch": 0.7409161190568225,
      "grad_norm": 0.0,
      "learning_rate": 1.295660997294163e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15334
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.4375,
      "epoch": 0.7409644375724778,
      "grad_norm": 0.0,
      "learning_rate": 1.295419404715887e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15335
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 95.9375,
      "epoch": 0.741012756088133,
      "grad_norm": 0.0,
      "learning_rate": 1.295177812137611e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15336
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.25,
      "epoch": 0.7410610746037881,
      "grad_norm": 0.0,
      "learning_rate": 1.294936219559335e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15337
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.46875,
      "epoch": 0.7411093931194433,
      "grad_norm": 0.0,
      "learning_rate": 1.294694626981059e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15338
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.65625,
      "epoch": 0.7411577116350986,
      "grad_norm": 0.0,
      "learning_rate": 1.2944530344027832e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15339
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.3125,
      "epoch": 0.7412060301507538,
      "grad_norm": 0.0,
      "learning_rate": 1.294211441824507e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15340
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 101.4375,
      "epoch": 0.7412543486664089,
      "grad_norm": 0.0,
      "learning_rate": 1.2939698492462312e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15341
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 100.6875,
      "epoch": 0.7413026671820642,
      "grad_norm": 0.0,
      "learning_rate": 1.293728256667955e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15342
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.125,
      "epoch": 0.7413509856977194,
      "grad_norm": 0.0,
      "learning_rate": 1.293486664089679e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15343
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 118.625,
      "epoch": 0.7413993042133745,
      "grad_norm": 8.962965672518994,
      "learning_rate": 1.293245071511403e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.4343159794807434,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15344
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.78125,
      "epoch": 0.7414476227290298,
      "grad_norm": 0.0,
      "learning_rate": 1.293003478933127e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15345
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.84375,
      "epoch": 0.741495941244685,
      "grad_norm": 0.0,
      "learning_rate": 1.292761886354851e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15346
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.34375,
      "epoch": 0.7415442597603401,
      "grad_norm": 0.0,
      "learning_rate": 1.292520293776575e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15347
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.78125,
      "epoch": 0.7415925782759953,
      "grad_norm": 0.0,
      "learning_rate": 1.292278701198299e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15348
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.25,
      "epoch": 0.7416408967916506,
      "grad_norm": 0.0,
      "learning_rate": 1.2920371086200233e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15349
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.75,
      "epoch": 0.7416892153073058,
      "grad_norm": 0.0,
      "learning_rate": 1.291795516041747e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15350
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 113.75,
      "epoch": 0.7417375338229609,
      "grad_norm": 0.0,
      "learning_rate": 1.2915539234634713e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15351
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.84375,
      "epoch": 0.7417858523386162,
      "grad_norm": 0.0,
      "learning_rate": 1.2913123308851951e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15352
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.90625,
      "epoch": 0.7418341708542714,
      "grad_norm": 0.0,
      "learning_rate": 1.2910707383069193e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15353
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 113.9375,
      "epoch": 0.7418824893699265,
      "grad_norm": 0.0,
      "learning_rate": 1.2908291457286431e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 15354
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 123.3125,
      "epoch": 0.7419308078855817,
      "grad_norm": 7.268830599999811,
      "learning_rate": 1.290587553150367e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15355
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.3125,
      "epoch": 0.741979126401237,
      "grad_norm": 0.0,
      "learning_rate": 1.2903459605720912e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15356
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 87.5,
      "epoch": 0.7420274449168921,
      "grad_norm": 0.0,
      "learning_rate": 1.290104367993815e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15357
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 116.21875,
      "epoch": 0.7420757634325473,
      "grad_norm": 5.47307617465689,
      "learning_rate": 1.2898627754155392e-07,
      "loss": -0.0,
      "reward": 1.390625,
      "reward_std": 0.21528217196464539,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15358
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 122.25,
      "epoch": 0.7421240819482026,
      "grad_norm": 0.0,
      "learning_rate": 1.2896211828372633e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15359
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.75,
      "epoch": 0.7421724004638578,
      "grad_norm": 0.0,
      "learning_rate": 1.2893795902589872e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15360
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.4375,
      "epoch": 0.7422207189795129,
      "grad_norm": 0.0,
      "learning_rate": 1.2891379976807113e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15361
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 119.875,
      "epoch": 0.7422690374951681,
      "grad_norm": 0.0,
      "learning_rate": 1.2888964051024352e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.24039676785469055,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15362
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.96875,
      "epoch": 0.7423173560108234,
      "grad_norm": 0.0,
      "learning_rate": 1.2886548125241593e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15363
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.65625,
      "epoch": 0.7423656745264785,
      "grad_norm": 0.0,
      "learning_rate": 1.2884132199458832e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15364
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.5,
      "epoch": 0.7424139930421337,
      "grad_norm": 16.093389133326532,
      "learning_rate": 1.288171627367607e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15365
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.0625,
      "epoch": 0.742462311557789,
      "grad_norm": 0.0,
      "learning_rate": 1.2879300347893312e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15366
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.375,
      "epoch": 0.7425106300734441,
      "grad_norm": 0.0,
      "learning_rate": 1.287688442211055e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15367
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 111.15625,
      "epoch": 0.7425589485890993,
      "grad_norm": 0.0,
      "learning_rate": 1.2874468496327792e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15368
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 112.8125,
      "epoch": 0.7426072671047546,
      "grad_norm": 0.0,
      "learning_rate": 1.2872052570545033e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 15369
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.3125,
      "epoch": 0.7426555856204098,
      "grad_norm": 0.0,
      "learning_rate": 1.2869636644762272e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15370
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.375,
      "epoch": 0.7427039041360649,
      "grad_norm": 0.0,
      "learning_rate": 1.2867220718979513e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15371
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.6875,
      "epoch": 0.7427522226517201,
      "grad_norm": 0.0,
      "learning_rate": 1.2864804793196752e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15372
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.1875,
      "epoch": 0.7428005411673754,
      "grad_norm": 5.145403478397317,
      "learning_rate": 1.2862388867413994e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15373
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 124.03125,
      "epoch": 0.7428488596830305,
      "grad_norm": 0.0,
      "learning_rate": 1.2859972941631235e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.24511480331420898,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15374
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.59375,
      "epoch": 0.7428971781986857,
      "grad_norm": 0.0,
      "learning_rate": 1.285755701584847e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15375
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.0625,
      "epoch": 0.742945496714341,
      "grad_norm": 0.0,
      "learning_rate": 1.2855141090065712e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15376
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.21875,
      "epoch": 0.7429938152299961,
      "grad_norm": 0.0,
      "learning_rate": 1.285272516428295e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15377
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 99.625,
      "epoch": 0.7430421337456513,
      "grad_norm": 0.0,
      "learning_rate": 1.2850309238500192e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15378
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 107.5,
      "epoch": 0.7430904522613065,
      "grad_norm": 4.575625928266825,
      "learning_rate": 1.2847893312717434e-07,
      "loss": -0.0,
      "reward": 1.84375,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15379
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 117.4375,
      "epoch": 0.7431387707769618,
      "grad_norm": 0.0,
      "learning_rate": 1.2845477386934672e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15380
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 135.3125,
      "epoch": 0.7431870892926169,
      "grad_norm": 0.0,
      "learning_rate": 1.2843061461151914e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15381
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.65625,
      "epoch": 0.7432354078082721,
      "grad_norm": 0.0,
      "learning_rate": 1.2840645535369153e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15382
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 103.3125,
      "epoch": 0.7432837263239274,
      "grad_norm": 0.0,
      "learning_rate": 1.2838229609586394e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 15383
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.21875,
      "epoch": 0.7433320448395825,
      "grad_norm": 0.0,
      "learning_rate": 1.2835813683803635e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15384
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.65625,
      "epoch": 0.7433803633552377,
      "grad_norm": 0.0,
      "learning_rate": 1.2833397758020871e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15385
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 129.1875,
      "epoch": 0.743428681870893,
      "grad_norm": 0.0,
      "learning_rate": 1.2830981832238113e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15386
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.46875,
      "epoch": 0.7434770003865481,
      "grad_norm": 0.0,
      "learning_rate": 1.2828565906455351e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15387
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.0,
      "epoch": 0.7435253189022033,
      "grad_norm": 0.0,
      "learning_rate": 1.2826149980672593e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15388
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.75,
      "epoch": 0.7435736374178585,
      "grad_norm": 0.0,
      "learning_rate": 1.2823734054889834e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15389
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.28125,
      "epoch": 0.7436219559335138,
      "grad_norm": 0.0,
      "learning_rate": 1.2821318129107073e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15390
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.375,
      "epoch": 0.7436702744491689,
      "grad_norm": 0.0,
      "learning_rate": 1.2818902203324314e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15391
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.4375,
      "epoch": 0.7437185929648241,
      "grad_norm": 0.0,
      "learning_rate": 1.2816486277541553e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15392
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 84.96875,
      "epoch": 0.7437669114804794,
      "grad_norm": 0.0,
      "learning_rate": 1.2814070351758794e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15393
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 84.8125,
      "epoch": 0.7438152299961345,
      "grad_norm": 0.0,
      "learning_rate": 1.2811654425976036e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15394
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 194.0,
      "epoch": 0.7438635485117897,
      "grad_norm": 0.0,
      "learning_rate": 1.2809238500193272e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.2777610421180725,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 0.96875,
      "step": 15395
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.5,
      "epoch": 0.7439118670274449,
      "grad_norm": 0.0,
      "learning_rate": 1.2806822574410513e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15396
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.84375,
      "epoch": 0.7439601855431001,
      "grad_norm": 0.0,
      "learning_rate": 1.2804406648627752e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15397
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.6875,
      "epoch": 0.7440085040587553,
      "grad_norm": 0.0,
      "learning_rate": 1.2801990722844993e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15398
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.90625,
      "epoch": 0.7440568225744105,
      "grad_norm": 0.0,
      "learning_rate": 1.2799574797062235e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15399
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.84375,
      "epoch": 0.7441051410900658,
      "grad_norm": 0.0,
      "learning_rate": 1.2797158871279473e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15400
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 124.90625,
      "epoch": 0.7441534596057209,
      "grad_norm": 0.0,
      "learning_rate": 1.2794742945496715e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15401
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.28125,
      "epoch": 0.7442017781213761,
      "grad_norm": 0.0,
      "learning_rate": 1.2792327019713953e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15402
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.53125,
      "epoch": 0.7442500966370313,
      "grad_norm": 0.0,
      "learning_rate": 1.2789911093931195e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15403
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.96875,
      "epoch": 0.7442984151526865,
      "grad_norm": 0.0,
      "learning_rate": 1.2787495168148436e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15404
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.71875,
      "epoch": 0.7443467336683417,
      "grad_norm": 0.0,
      "learning_rate": 1.2785079242365672e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15405
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.34375,
      "epoch": 0.7443950521839969,
      "grad_norm": 0.0,
      "learning_rate": 1.2782663316582914e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15406
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 107.59375,
      "epoch": 0.744443370699652,
      "grad_norm": 0.0,
      "learning_rate": 1.2780247390800152e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 15407
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 99.78125,
      "epoch": 0.7444916892153073,
      "grad_norm": 0.0,
      "learning_rate": 1.2777831465017394e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 15408
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.21875,
      "epoch": 0.7445400077309625,
      "grad_norm": 0.0,
      "learning_rate": 1.2775415539234635e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15409
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.53125,
      "epoch": 0.7445883262466177,
      "grad_norm": 0.0,
      "learning_rate": 1.2772999613451874e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15410
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.28125,
      "epoch": 0.7446366447622729,
      "grad_norm": 0.0,
      "learning_rate": 1.2770583687669115e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15411
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.03125,
      "epoch": 0.7446849632779281,
      "grad_norm": 0.0,
      "learning_rate": 1.2768167761886354e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15412
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.21875,
      "epoch": 0.7447332817935833,
      "grad_norm": 0.0,
      "learning_rate": 1.2765751836103595e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15413
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.03125,
      "epoch": 0.7447816003092385,
      "grad_norm": 0.0,
      "learning_rate": 1.2763335910320836e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15414
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.9375,
      "epoch": 0.7448299188248937,
      "grad_norm": 0.0,
      "learning_rate": 1.2760919984538073e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15415
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.6875,
      "epoch": 0.7448782373405489,
      "grad_norm": 0.0,
      "learning_rate": 1.2758504058755314e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15416
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 85.75,
      "epoch": 0.744926555856204,
      "grad_norm": 0.0,
      "learning_rate": 1.2756088132972553e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15417
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.78125,
      "epoch": 0.7449748743718593,
      "grad_norm": 0.0,
      "learning_rate": 1.2753672207189794e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15418
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.75,
      "epoch": 0.7450231928875145,
      "grad_norm": 0.0,
      "learning_rate": 1.2751256281407035e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15419
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 91.0625,
      "epoch": 0.7450715114031697,
      "grad_norm": 36.518248096424536,
      "learning_rate": 1.2748840355624274e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15420
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.59375,
      "epoch": 0.7451198299188249,
      "grad_norm": 0.0,
      "learning_rate": 1.2746424429841515e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15421
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.9375,
      "epoch": 0.7451681484344801,
      "grad_norm": 0.0,
      "learning_rate": 1.2744008504058754e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15422
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 89.625,
      "epoch": 0.7452164669501353,
      "grad_norm": 0.0,
      "learning_rate": 1.2741592578275996e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 15423
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.59375,
      "epoch": 0.7452647854657904,
      "grad_norm": 0.0,
      "learning_rate": 1.2739176652493237e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15424
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.65625,
      "epoch": 0.7453131039814457,
      "grad_norm": 0.0,
      "learning_rate": 1.2736760726710476e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15425
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 95.8125,
      "epoch": 0.7453614224971009,
      "grad_norm": 0.0,
      "learning_rate": 1.2734344800927714e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.20044592022895813,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 0.96875,
      "step": 15426
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.65625,
      "epoch": 0.745409741012756,
      "grad_norm": 0.0,
      "learning_rate": 1.2731928875144953e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15427
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.21875,
      "epoch": 0.7454580595284113,
      "grad_norm": 0.0,
      "learning_rate": 1.2729512949362194e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15428
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 131.59375,
      "epoch": 0.7455063780440665,
      "grad_norm": 0.0,
      "learning_rate": 1.2727097023579436e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15429
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.25,
      "epoch": 0.7455546965597217,
      "grad_norm": 0.0,
      "learning_rate": 1.2724681097796674e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15430
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.40625,
      "epoch": 0.7456030150753769,
      "grad_norm": 0.0,
      "learning_rate": 1.2722265172013916e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15431
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.21875,
      "epoch": 0.7456513335910321,
      "grad_norm": 0.0,
      "learning_rate": 1.2719849246231155e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15432
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.5,
      "epoch": 0.7456996521066873,
      "grad_norm": 0.0,
      "learning_rate": 1.2717433320448396e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15433
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 86.125,
      "epoch": 0.7457479706223424,
      "grad_norm": 0.0,
      "learning_rate": 1.2715017394665637e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 15434
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 101.46875,
      "epoch": 0.7457962891379977,
      "grad_norm": 28.87433194022234,
      "learning_rate": 1.2712601468882876e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.2731459438800812,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15435
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.03125,
      "epoch": 0.7458446076536529,
      "grad_norm": 0.0,
      "learning_rate": 1.2710185543100115e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15436
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 96.46875,
      "epoch": 0.745892926169308,
      "grad_norm": 0.0,
      "learning_rate": 1.2707769617317353e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15437
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 103.46875,
      "epoch": 0.7459412446849633,
      "grad_norm": 5.864328793944272,
      "learning_rate": 1.2705353691534595e-07,
      "loss": -0.0,
      "reward": 2.125,
      "reward_std": 0.27102428674697876,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15438
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 128.5,
      "epoch": 0.7459895632006185,
      "grad_norm": 0.0,
      "learning_rate": 1.2702937765751836e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.2382849156856537,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15439
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 125.5,
      "epoch": 0.7460378817162737,
      "grad_norm": 0.0,
      "learning_rate": 1.2700521839969075e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 15440
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.28125,
      "epoch": 0.7460862002319288,
      "grad_norm": 0.0,
      "learning_rate": 1.2698105914186316e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15441
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.25,
      "epoch": 0.7461345187475841,
      "grad_norm": 0.0,
      "learning_rate": 1.2695689988403555e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15442
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.40625,
      "epoch": 0.7461828372632393,
      "grad_norm": 0.0,
      "learning_rate": 1.2693274062620796e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 15443
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.53125,
      "epoch": 0.7462311557788944,
      "grad_norm": 0.0,
      "learning_rate": 1.2690858136838038e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15444
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.5625,
      "epoch": 0.7462794742945497,
      "grad_norm": 0.0,
      "learning_rate": 1.2688442211055276e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15445
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.875,
      "epoch": 0.7463277928102049,
      "grad_norm": 0.0,
      "learning_rate": 1.2686026285272515e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15446
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.9375,
      "epoch": 0.74637611132586,
      "grad_norm": 0.0,
      "learning_rate": 1.2683610359489754e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15447
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.0625,
      "epoch": 0.7464244298415152,
      "grad_norm": 0.0,
      "learning_rate": 1.2681194433706995e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15448
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.375,
      "epoch": 0.7464727483571705,
      "grad_norm": 0.0,
      "learning_rate": 1.2678778507924237e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15449
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.4375,
      "epoch": 0.7465210668728257,
      "grad_norm": 0.0,
      "learning_rate": 1.2676362582141475e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15450
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.625,
      "epoch": 0.7465693853884808,
      "grad_norm": 0.0,
      "learning_rate": 1.2673946656358717e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15451
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 111.09375,
      "epoch": 0.7466177039041361,
      "grad_norm": 0.0,
      "learning_rate": 1.2671530730575955e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15452
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.03125,
      "epoch": 0.7466660224197913,
      "grad_norm": 0.0,
      "learning_rate": 1.2669114804793197e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15453
    },
    {
      "all_correct": 0.0,
      "all_wrong": 0.75,
      "completion_length": 112.53125,
      "epoch": 0.7467143409354464,
      "grad_norm": 8.871403940115597,
      "learning_rate": 1.2666698879010438e-07,
      "loss": 0.0,
      "reward": 1.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.03125,
      "rewards/bonus_rewards": 0.015625,
      "rewards/format_reward": 1.0,
      "step": 15454
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 112.90625,
      "epoch": 0.7467626594511017,
      "grad_norm": 3.5397613139384023,
      "learning_rate": 1.2664282953227677e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15455
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.21875,
      "epoch": 0.7468109779667569,
      "grad_norm": 0.0,
      "learning_rate": 1.2661867027444918e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15456
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.1875,
      "epoch": 0.746859296482412,
      "grad_norm": 0.0,
      "learning_rate": 1.2659451101662154e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15457
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 114.28125,
      "epoch": 0.7469076149980672,
      "grad_norm": 0.0,
      "learning_rate": 1.2657035175879396e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15458
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 87.59375,
      "epoch": 0.7469559335137225,
      "grad_norm": 17.89713692131075,
      "learning_rate": 1.2654619250096637e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15459
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.84375,
      "epoch": 0.7470042520293777,
      "grad_norm": 0.0,
      "learning_rate": 1.2652203324313876e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15460
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.96875,
      "epoch": 0.7470525705450328,
      "grad_norm": 0.0,
      "learning_rate": 1.2649787398531117e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15461
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.09375,
      "epoch": 0.7471008890606881,
      "grad_norm": 0.0,
      "learning_rate": 1.2647371472748356e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15462
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.9375,
      "epoch": 0.7471492075763433,
      "grad_norm": 0.0,
      "learning_rate": 1.2644955546965597e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15463
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.875,
      "epoch": 0.7471975260919984,
      "grad_norm": 0.0,
      "learning_rate": 1.2642539621182838e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15464
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 131.9375,
      "epoch": 0.7472458446076536,
      "grad_norm": 0.0,
      "learning_rate": 1.2640123695400077e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15465
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.0625,
      "epoch": 0.7472941631233089,
      "grad_norm": 0.0,
      "learning_rate": 1.2637707769617319e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15466
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.46875,
      "epoch": 0.747342481638964,
      "grad_norm": 0.0,
      "learning_rate": 1.2635291843834555e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15467
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.96875,
      "epoch": 0.7473908001546192,
      "grad_norm": 0.0,
      "learning_rate": 1.2632875918051796e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15468
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.03125,
      "epoch": 0.7474391186702745,
      "grad_norm": 0.0,
      "learning_rate": 1.2630459992269037e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15469
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.125,
      "epoch": 0.7474874371859297,
      "grad_norm": 0.0,
      "learning_rate": 1.2628044066486276e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15470
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 108.1875,
      "epoch": 0.7475357557015848,
      "grad_norm": 37.62237353623092,
      "learning_rate": 1.2625628140703517e-07,
      "loss": -0.0,
      "reward": 1.890625,
      "reward_std": 0.2461063265800476,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15471
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 98.96875,
      "epoch": 0.74758407421724,
      "grad_norm": 0.0,
      "learning_rate": 1.2623212214920756e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 15472
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 100.03125,
      "epoch": 0.7476323927328953,
      "grad_norm": 0.0,
      "learning_rate": 1.2620796289137997e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15473
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.71875,
      "epoch": 0.7476807112485504,
      "grad_norm": 0.0,
      "learning_rate": 1.261838036335524e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15474
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.15625,
      "epoch": 0.7477290297642056,
      "grad_norm": 0.0,
      "learning_rate": 1.2615964437572478e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15475
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.9375,
      "epoch": 0.7477773482798609,
      "grad_norm": 0.0,
      "learning_rate": 1.261354851178972e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15476
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 123.65625,
      "epoch": 0.747825666795516,
      "grad_norm": 0.0,
      "learning_rate": 1.2611132586006955e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15477
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.65625,
      "epoch": 0.7478739853111712,
      "grad_norm": 0.0,
      "learning_rate": 1.2608716660224196e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15478
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 86.53125,
      "epoch": 0.7479223038268265,
      "grad_norm": 0.0,
      "learning_rate": 1.2606300734441438e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15479
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 112.3125,
      "epoch": 0.7479706223424817,
      "grad_norm": 5.46840477043066,
      "learning_rate": 1.2603884808658676e-07,
      "loss": 0.0,
      "reward": 1.765625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15480
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.375,
      "epoch": 0.7480189408581368,
      "grad_norm": 0.0,
      "learning_rate": 1.2601468882875918e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15481
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.09375,
      "epoch": 0.748067259373792,
      "grad_norm": 0.0,
      "learning_rate": 1.2599052957093156e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15482
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.25,
      "epoch": 0.7481155778894473,
      "grad_norm": 0.0,
      "learning_rate": 1.2596637031310398e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15483
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.25,
      "epoch": 0.7481638964051024,
      "grad_norm": 0.0,
      "learning_rate": 1.259422110552764e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 15484
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.3125,
      "epoch": 0.7482122149207576,
      "grad_norm": 0.0,
      "learning_rate": 1.2591805179744878e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15485
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 100.96875,
      "epoch": 0.7482605334364129,
      "grad_norm": 21.949239338381666,
      "learning_rate": 1.258938925396212e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15486
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 115.09375,
      "epoch": 0.748308851952068,
      "grad_norm": 0.0,
      "learning_rate": 1.2586973328179355e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15487
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.84375,
      "epoch": 0.7483571704677232,
      "grad_norm": 0.0,
      "learning_rate": 1.2584557402396597e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15488
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.21875,
      "epoch": 0.7484054889833784,
      "grad_norm": 0.0,
      "learning_rate": 1.2582141476613838e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15489
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.03125,
      "epoch": 0.7484538074990337,
      "grad_norm": 0.0,
      "learning_rate": 1.2579725550831077e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15490
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.1875,
      "epoch": 0.7485021260146888,
      "grad_norm": 0.0,
      "learning_rate": 1.2577309625048318e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15491
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.59375,
      "epoch": 0.748550444530344,
      "grad_norm": 0.0,
      "learning_rate": 1.2574893699265557e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15492
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.3125,
      "epoch": 0.7485987630459993,
      "grad_norm": 0.0,
      "learning_rate": 1.2572477773482798e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15493
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.59375,
      "epoch": 0.7486470815616544,
      "grad_norm": 0.0,
      "learning_rate": 1.257006184770004e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15494
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.78125,
      "epoch": 0.7486954000773096,
      "grad_norm": 0.0,
      "learning_rate": 1.2567645921917278e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15495
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.78125,
      "epoch": 0.7487437185929648,
      "grad_norm": 0.0,
      "learning_rate": 1.256522999613452e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15496
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.84375,
      "epoch": 0.74879203710862,
      "grad_norm": 0.0,
      "learning_rate": 1.2562814070351758e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15497
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 89.1875,
      "epoch": 0.7488403556242752,
      "grad_norm": 0.0,
      "learning_rate": 1.2560398144568997e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15498
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.59375,
      "epoch": 0.7488886741399304,
      "grad_norm": 0.0,
      "learning_rate": 1.2557982218786238e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15499
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 102.75,
      "epoch": 0.7489369926555857,
      "grad_norm": 4.459077687933015,
      "learning_rate": 1.2555566293003477e-07,
      "loss": 0.0,
      "reward": 2.171875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15500
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 114.28125,
      "epoch": 0.7489853111712408,
      "grad_norm": 12.18514061307951,
      "learning_rate": 1.2553150367220719e-07,
      "loss": 0.0,
      "reward": 1.859375,
      "reward_std": 0.15580293536186218,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15501
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.75,
      "epoch": 0.749033629686896,
      "grad_norm": 0.0,
      "learning_rate": 1.2550734441437957e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15502
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 103.15625,
      "epoch": 0.7490819482025513,
      "grad_norm": 13.747876788879715,
      "learning_rate": 1.2548318515655199e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15503
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.78125,
      "epoch": 0.7491302667182064,
      "grad_norm": 0.0,
      "learning_rate": 1.254590258987244e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15504
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 86.90625,
      "epoch": 0.7491785852338616,
      "grad_norm": 0.0,
      "learning_rate": 1.254348666408968e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15505
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.09375,
      "epoch": 0.7492269037495168,
      "grad_norm": 0.0,
      "learning_rate": 1.254107073830692e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15506
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.46875,
      "epoch": 0.749275222265172,
      "grad_norm": 0.0,
      "learning_rate": 1.253865481252416e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15507
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 159.71875,
      "epoch": 0.7493235407808272,
      "grad_norm": 0.0,
      "learning_rate": 1.2536238886741398e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 0.96875,
      "step": 15508
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 111.1875,
      "epoch": 0.7493718592964824,
      "grad_norm": 0.0,
      "learning_rate": 1.253382296095864e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15509
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.34375,
      "epoch": 0.7494201778121377,
      "grad_norm": 0.0,
      "learning_rate": 1.2531407035175878e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15510
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.65625,
      "epoch": 0.7494684963277928,
      "grad_norm": 0.0,
      "learning_rate": 1.252899110939312e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15511
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.8125,
      "epoch": 0.749516814843448,
      "grad_norm": 0.0,
      "learning_rate": 1.2526575183610358e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15512
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 121.90625,
      "epoch": 0.7495651333591032,
      "grad_norm": 17.898050726924673,
      "learning_rate": 1.25241592578276e-07,
      "loss": -0.0,
      "reward": 1.53125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15513
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.0625,
      "epoch": 0.7496134518747584,
      "grad_norm": 0.0,
      "learning_rate": 1.252174333204484e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15514
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.125,
      "epoch": 0.7496617703904136,
      "grad_norm": 0.0,
      "learning_rate": 1.251932740626208e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15515
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.84375,
      "epoch": 0.7497100889060688,
      "grad_norm": 0.0,
      "learning_rate": 1.251691148047932e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15516
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 138.78125,
      "epoch": 0.749758407421724,
      "grad_norm": 11.688792476367489,
      "learning_rate": 1.251449555469656e-07,
      "loss": 0.0,
      "reward": 1.8125,
      "reward_std": 0.2925041913986206,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15517
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.40625,
      "epoch": 0.7498067259373792,
      "grad_norm": 0.0,
      "learning_rate": 1.2512079628913798e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15518
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 86.75,
      "epoch": 0.7498550444530344,
      "grad_norm": 0.0,
      "learning_rate": 1.250966370313104e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15519
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 88.9375,
      "epoch": 0.7499033629686896,
      "grad_norm": 0.0,
      "learning_rate": 1.2507247777348278e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15520
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.6875,
      "epoch": 0.7499516814843448,
      "grad_norm": 0.0,
      "learning_rate": 1.250483185156552e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15521
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.875,
      "epoch": 0.75,
      "grad_norm": 0.0,
      "learning_rate": 1.2502415925782758e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15522
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.1875,
      "epoch": 0.7500483185156552,
      "grad_norm": 0.0,
      "learning_rate": 1.25e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15523
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.46875,
      "epoch": 0.7500966370313104,
      "grad_norm": 0.0,
      "learning_rate": 1.2497584074217238e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15524
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.84375,
      "epoch": 0.7501449555469656,
      "grad_norm": 0.0,
      "learning_rate": 1.249516814843448e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15525
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.125,
      "epoch": 0.7501932740626208,
      "grad_norm": 0.0,
      "learning_rate": 1.249275222265172e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15526
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.96875,
      "epoch": 0.750241592578276,
      "grad_norm": 0.0,
      "learning_rate": 1.249033629686896e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15527
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.53125,
      "epoch": 0.7502899110939312,
      "grad_norm": 0.0,
      "learning_rate": 1.24879203710862e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15528
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.34375,
      "epoch": 0.7503382296095864,
      "grad_norm": 0.0,
      "learning_rate": 1.248550444530344e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15529
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.46875,
      "epoch": 0.7503865481252416,
      "grad_norm": 0.0,
      "learning_rate": 1.2483088519520678e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15530
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.75,
      "epoch": 0.7504348666408968,
      "grad_norm": 0.0,
      "learning_rate": 1.248067259373792e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15531
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.78125,
      "epoch": 0.750483185156552,
      "grad_norm": 0.0,
      "learning_rate": 1.247825666795516e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15532
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 106.21875,
      "epoch": 0.7505315036722072,
      "grad_norm": 0.0,
      "learning_rate": 1.24758407421724e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15533
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.96875,
      "epoch": 0.7505798221878623,
      "grad_norm": 0.0,
      "learning_rate": 1.2473424816389639e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15534
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.71875,
      "epoch": 0.7506281407035176,
      "grad_norm": 0.0,
      "learning_rate": 1.247100889060688e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15535
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.65625,
      "epoch": 0.7506764592191728,
      "grad_norm": 0.0,
      "learning_rate": 1.246859296482412e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15536
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.1875,
      "epoch": 0.750724777734828,
      "grad_norm": 0.0,
      "learning_rate": 1.246617703904136e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15537
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.34375,
      "epoch": 0.7507730962504832,
      "grad_norm": 0.0,
      "learning_rate": 1.2463761113258601e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15538
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.84375,
      "epoch": 0.7508214147661384,
      "grad_norm": 0.0,
      "learning_rate": 1.246134518747584e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15539
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.4375,
      "epoch": 0.7508697332817936,
      "grad_norm": 0.0,
      "learning_rate": 1.245892926169308e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15540
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.28125,
      "epoch": 0.7509180517974487,
      "grad_norm": 0.0,
      "learning_rate": 1.245651333591032e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15541
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 117.75,
      "epoch": 0.750966370313104,
      "grad_norm": 8.21667943604607,
      "learning_rate": 1.2454097410127562e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15542
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.75,
      "epoch": 0.7510146888287592,
      "grad_norm": 0.0,
      "learning_rate": 1.24516814843448e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15543
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.90625,
      "epoch": 0.7510630073444143,
      "grad_norm": 0.0,
      "learning_rate": 1.244926555856204e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15544
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.03125,
      "epoch": 0.7511113258600696,
      "grad_norm": 0.0,
      "learning_rate": 1.244684963277928e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15545
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 139.375,
      "epoch": 0.7511596443757248,
      "grad_norm": 0.0,
      "learning_rate": 1.2444433706996522e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15546
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 92.75,
      "epoch": 0.75120796289138,
      "grad_norm": 0.0,
      "learning_rate": 1.244201778121376e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15547
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.59375,
      "epoch": 0.7512562814070352,
      "grad_norm": 0.0,
      "learning_rate": 1.2439601855431002e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15548
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 88.96875,
      "epoch": 0.7513045999226904,
      "grad_norm": 0.0,
      "learning_rate": 1.243718592964824e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15549
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.0,
      "epoch": 0.7513529184383456,
      "grad_norm": 0.0,
      "learning_rate": 1.243477000386548e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15550
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 123.96875,
      "epoch": 0.7514012369540007,
      "grad_norm": 0.0,
      "learning_rate": 1.243235407808272e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15551
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.46875,
      "epoch": 0.751449555469656,
      "grad_norm": 0.0,
      "learning_rate": 1.2429938152299962e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 15552
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.15625,
      "epoch": 0.7514978739853112,
      "grad_norm": 0.0,
      "learning_rate": 1.24275222265172e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1552036553621292,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15553
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.75,
      "epoch": 0.7515461925009663,
      "grad_norm": 0.0,
      "learning_rate": 1.242510630073444e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15554
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.25,
      "epoch": 0.7515945110166216,
      "grad_norm": 0.0,
      "learning_rate": 1.242269037495168e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15555
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 97.4375,
      "epoch": 0.7516428295322768,
      "grad_norm": 15.590088441854197,
      "learning_rate": 1.2420274449168922e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.2970154285430908,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15556
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.9375,
      "epoch": 0.751691148047932,
      "grad_norm": 0.0,
      "learning_rate": 1.241785852338616e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15557
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.84375,
      "epoch": 0.7517394665635871,
      "grad_norm": 0.0,
      "learning_rate": 1.2415442597603402e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15558
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.03125,
      "epoch": 0.7517877850792424,
      "grad_norm": 0.0,
      "learning_rate": 1.241302667182064e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15559
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.875,
      "epoch": 0.7518361035948976,
      "grad_norm": 0.0,
      "learning_rate": 1.241061074603788e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15560
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.75,
      "epoch": 0.7518844221105527,
      "grad_norm": 0.0,
      "learning_rate": 1.240819482025512e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15561
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.40625,
      "epoch": 0.751932740626208,
      "grad_norm": 0.0,
      "learning_rate": 1.2405778894472362e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15562
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 173.625,
      "epoch": 0.7519810591418632,
      "grad_norm": 80.79902495891595,
      "learning_rate": 1.24033629686896e-07,
      "loss": -0.0,
      "reward": 2.09375,
      "reward_std": 0.5334908962249756,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 0.96875,
      "step": 15563
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.75,
      "epoch": 0.7520293776575183,
      "grad_norm": 0.0,
      "learning_rate": 1.2400947042906842e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15564
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.1875,
      "epoch": 0.7520776961731735,
      "grad_norm": 0.0,
      "learning_rate": 1.239853111712408e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15565
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.90625,
      "epoch": 0.7521260146888288,
      "grad_norm": 0.0,
      "learning_rate": 1.2396115191341322e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15566
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.09375,
      "epoch": 0.752174333204484,
      "grad_norm": 0.0,
      "learning_rate": 1.239369926555856e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15567
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.875,
      "epoch": 0.7522226517201391,
      "grad_norm": 0.0,
      "learning_rate": 1.2391283339775803e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15568
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 117.25,
      "epoch": 0.7522709702357944,
      "grad_norm": 5.3755326872839335,
      "learning_rate": 1.238886741399304e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15569
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.09375,
      "epoch": 0.7523192887514496,
      "grad_norm": 0.0,
      "learning_rate": 1.238645148821028e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15570
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 137.75,
      "epoch": 0.7523676072671047,
      "grad_norm": 0.0,
      "learning_rate": 1.238403556242752e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15571
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.71875,
      "epoch": 0.75241592578276,
      "grad_norm": 0.0,
      "learning_rate": 1.2381619636644763e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15572
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.4375,
      "epoch": 0.7524642442984152,
      "grad_norm": 0.0,
      "learning_rate": 1.2379203710862001e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15573
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.96875,
      "epoch": 0.7525125628140703,
      "grad_norm": 0.0,
      "learning_rate": 1.2376787785079243e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15574
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.15625,
      "epoch": 0.7525608813297255,
      "grad_norm": 0.0,
      "learning_rate": 1.2374371859296481e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.21778544783592224,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15575
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.59375,
      "epoch": 0.7526091998453808,
      "grad_norm": 0.0,
      "learning_rate": 1.2371955933513723e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15576
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 120.46875,
      "epoch": 0.752657518361036,
      "grad_norm": 0.0,
      "learning_rate": 1.2369540007730962e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15577
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 120.28125,
      "epoch": 0.7527058368766911,
      "grad_norm": 0.0,
      "learning_rate": 1.2367124081948203e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15578
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 123.53125,
      "epoch": 0.7527541553923464,
      "grad_norm": 0.0,
      "learning_rate": 1.2364708156165442e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15579
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.78125,
      "epoch": 0.7528024739080016,
      "grad_norm": 0.0,
      "learning_rate": 1.236229223038268e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15580
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.125,
      "epoch": 0.7528507924236567,
      "grad_norm": 0.0,
      "learning_rate": 1.2359876304599922e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15581
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.40625,
      "epoch": 0.7528991109393119,
      "grad_norm": 0.0,
      "learning_rate": 1.2357460378817163e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15582
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.53125,
      "epoch": 0.7529474294549672,
      "grad_norm": 0.0,
      "learning_rate": 1.2355044453034402e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15583
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.0,
      "epoch": 0.7529957479706223,
      "grad_norm": 0.0,
      "learning_rate": 1.2352628527251643e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15584
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.0625,
      "epoch": 0.7530440664862775,
      "grad_norm": 0.0,
      "learning_rate": 1.2350212601468882e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15585
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.78125,
      "epoch": 0.7530923850019328,
      "grad_norm": 0.0,
      "learning_rate": 1.2347796675686123e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15586
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.1875,
      "epoch": 0.753140703517588,
      "grad_norm": 0.0,
      "learning_rate": 1.2345380749903362e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15587
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.25,
      "epoch": 0.7531890220332431,
      "grad_norm": 0.0,
      "learning_rate": 1.2342964824120603e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15588
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 125.125,
      "epoch": 0.7532373405488983,
      "grad_norm": 0.0,
      "learning_rate": 1.2340548898337842e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15589
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.53125,
      "epoch": 0.7532856590645536,
      "grad_norm": 0.0,
      "learning_rate": 1.233813297255508e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15590
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.21875,
      "epoch": 0.7533339775802087,
      "grad_norm": 0.0,
      "learning_rate": 1.2335717046772322e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15591
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 115.5625,
      "epoch": 0.7533822960958639,
      "grad_norm": 0.0,
      "learning_rate": 1.2333301120989563e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15592
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 129.75,
      "epoch": 0.7534306146115192,
      "grad_norm": 0.0,
      "learning_rate": 1.2330885195206802e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15593
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.875,
      "epoch": 0.7534789331271743,
      "grad_norm": 0.0,
      "learning_rate": 1.2328469269424044e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15594
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.65625,
      "epoch": 0.7535272516428295,
      "grad_norm": 0.0,
      "learning_rate": 1.2326053343641282e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.22461533546447754,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15595
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 132.8125,
      "epoch": 0.7535755701584848,
      "grad_norm": 0.0,
      "learning_rate": 1.2323637417858524e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15596
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 115.09375,
      "epoch": 0.75362388867414,
      "grad_norm": 16.23578917559714,
      "learning_rate": 1.2321221492075762e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15597
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 123.34375,
      "epoch": 0.7536722071897951,
      "grad_norm": 0.0,
      "learning_rate": 1.2318805566293004e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15598
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.6875,
      "epoch": 0.7537205257054503,
      "grad_norm": 0.0,
      "learning_rate": 1.2316389640510242e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15599
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.78125,
      "epoch": 0.7537688442211056,
      "grad_norm": 0.0,
      "learning_rate": 1.2313973714727484e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15600
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.59375,
      "epoch": 0.7538171627367607,
      "grad_norm": 0.0,
      "learning_rate": 1.2311557788944723e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15601
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.21875,
      "epoch": 0.7538654812524159,
      "grad_norm": 0.0,
      "learning_rate": 1.2309141863161964e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15602
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.65625,
      "epoch": 0.7539137997680712,
      "grad_norm": 0.0,
      "learning_rate": 1.2306725937379203e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15603
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 111.59375,
      "epoch": 0.7539621182837263,
      "grad_norm": 0.0,
      "learning_rate": 1.2304310011596444e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15604
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 129.09375,
      "epoch": 0.7540104367993815,
      "grad_norm": 0.0,
      "learning_rate": 1.2301894085813683e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15605
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.5,
      "epoch": 0.7540587553150367,
      "grad_norm": 0.0,
      "learning_rate": 1.2299478160030924e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15606
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.09375,
      "epoch": 0.754107073830692,
      "grad_norm": 0.0,
      "learning_rate": 1.2297062234248163e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15607
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.3125,
      "epoch": 0.7541553923463471,
      "grad_norm": 0.0,
      "learning_rate": 1.2294646308465404e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15608
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 138.3125,
      "epoch": 0.7542037108620023,
      "grad_norm": 4.579009186152704,
      "learning_rate": 1.2292230382682643e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15609
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 132.5,
      "epoch": 0.7542520293776576,
      "grad_norm": 18.5443574598323,
      "learning_rate": 1.2289814456899884e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.3234682083129883,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15610
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.6875,
      "epoch": 0.7543003478933127,
      "grad_norm": 0.0,
      "learning_rate": 1.2287398531117123e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15611
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 133.875,
      "epoch": 0.7543486664089679,
      "grad_norm": 0.0,
      "learning_rate": 1.2284982605334364e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15612
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.5625,
      "epoch": 0.7543969849246231,
      "grad_norm": 0.0,
      "learning_rate": 1.2282566679551603e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15613
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 112.875,
      "epoch": 0.7544453034402783,
      "grad_norm": 0.0,
      "learning_rate": 1.2280150753768844e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15614
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.0625,
      "epoch": 0.7544936219559335,
      "grad_norm": 0.0,
      "learning_rate": 1.2277734827986083e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15615
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.625,
      "epoch": 0.7545419404715887,
      "grad_norm": 0.0,
      "learning_rate": 1.2275318902203324e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15616
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.625,
      "epoch": 0.754590258987244,
      "grad_norm": 0.0,
      "learning_rate": 1.2272902976420563e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15617
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.78125,
      "epoch": 0.7546385775028991,
      "grad_norm": 0.0,
      "learning_rate": 1.2270487050637804e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15618
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.125,
      "epoch": 0.7546868960185543,
      "grad_norm": 0.0,
      "learning_rate": 1.2268071124855043e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15619
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 141.0625,
      "epoch": 0.7547352145342096,
      "grad_norm": 21.06097791337788,
      "learning_rate": 1.2265655199072285e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.360832542181015,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15620
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.09375,
      "epoch": 0.7547835330498647,
      "grad_norm": 0.0,
      "learning_rate": 1.2263239273289523e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15621
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 118.15625,
      "epoch": 0.7548318515655199,
      "grad_norm": 0.0,
      "learning_rate": 1.2260823347506765e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15622
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 118.28125,
      "epoch": 0.7548801700811751,
      "grad_norm": 0.0,
      "learning_rate": 1.2258407421724003e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15623
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 129.125,
      "epoch": 0.7549284885968303,
      "grad_norm": 19.05638059154577,
      "learning_rate": 1.2255991495941245e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.3061639666557312,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15624
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.65625,
      "epoch": 0.7549768071124855,
      "grad_norm": 0.0,
      "learning_rate": 1.2253575570158483e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15625
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.53125,
      "epoch": 0.7550251256281407,
      "grad_norm": 0.0,
      "learning_rate": 1.2251159644375725e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15626
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 124.53125,
      "epoch": 0.755073444143796,
      "grad_norm": 0.0,
      "learning_rate": 1.2248743718592964e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15627
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 117.375,
      "epoch": 0.7551217626594511,
      "grad_norm": 28.35328519545127,
      "learning_rate": 1.2246327792810205e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.36394569277763367,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15628
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.8125,
      "epoch": 0.7551700811751063,
      "grad_norm": 0.0,
      "learning_rate": 1.2243911867027444e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15629
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.5,
      "epoch": 0.7552183996907615,
      "grad_norm": 0.0,
      "learning_rate": 1.2241495941244685e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15630
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.40625,
      "epoch": 0.7552667182064167,
      "grad_norm": 0.0,
      "learning_rate": 1.2239080015461926e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15631
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 100.21875,
      "epoch": 0.7553150367220719,
      "grad_norm": 0.0,
      "learning_rate": 1.2236664089679165e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 15632
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 111.6875,
      "epoch": 0.7553633552377271,
      "grad_norm": 0.0,
      "learning_rate": 1.2234248163896404e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15633
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.75,
      "epoch": 0.7554116737533823,
      "grad_norm": 0.0,
      "learning_rate": 1.2231832238113645e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15634
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 116.875,
      "epoch": 0.7554599922690375,
      "grad_norm": 0.0,
      "learning_rate": 1.2229416312330884e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 15635
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.8125,
      "epoch": 0.7555083107846927,
      "grad_norm": 0.0,
      "learning_rate": 1.2227000386548125e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15636
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.84375,
      "epoch": 0.755556629300348,
      "grad_norm": 0.0,
      "learning_rate": 1.2224584460765364e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15637
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.9375,
      "epoch": 0.7556049478160031,
      "grad_norm": 0.0,
      "learning_rate": 1.2222168534982605e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1462521106004715,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15638
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.09375,
      "epoch": 0.7556532663316583,
      "grad_norm": 0.0,
      "learning_rate": 1.2219752609199844e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15639
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 142.59375,
      "epoch": 0.7557015848473135,
      "grad_norm": 0.0,
      "learning_rate": 1.2217336683417085e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15640
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.5,
      "epoch": 0.7557499033629687,
      "grad_norm": 0.0,
      "learning_rate": 1.2214920757634327e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15641
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 104.21875,
      "epoch": 0.7557982218786239,
      "grad_norm": 0.0,
      "learning_rate": 1.2212504831851565e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15642
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 129.25,
      "epoch": 0.7558465403942791,
      "grad_norm": 0.0,
      "learning_rate": 1.2210088906068804e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15643
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.15625,
      "epoch": 0.7558948589099342,
      "grad_norm": 0.0,
      "learning_rate": 1.2207672980286046e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15644
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.71875,
      "epoch": 0.7559431774255895,
      "grad_norm": 0.0,
      "learning_rate": 1.2205257054503284e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15645
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.1875,
      "epoch": 0.7559914959412447,
      "grad_norm": 0.0,
      "learning_rate": 1.2202841128720526e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15646
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.03125,
      "epoch": 0.7560398144568999,
      "grad_norm": 0.0,
      "learning_rate": 1.2200425202937764e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15647
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.9375,
      "epoch": 0.7560881329725551,
      "grad_norm": 0.0,
      "learning_rate": 1.2198009277155006e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15648
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 126.3125,
      "epoch": 0.7561364514882103,
      "grad_norm": 0.0,
      "learning_rate": 1.2195593351372244e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15649
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.59375,
      "epoch": 0.7561847700038655,
      "grad_norm": 0.0,
      "learning_rate": 1.2193177425589486e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15650
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 124.40625,
      "epoch": 0.7562330885195206,
      "grad_norm": 8.627113488097345,
      "learning_rate": 1.2190761499806727e-07,
      "loss": 0.0,
      "reward": 2.21875,
      "reward_std": 0.3498838543891907,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15651
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.09375,
      "epoch": 0.7562814070351759,
      "grad_norm": 0.0,
      "learning_rate": 1.2188345574023966e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15652
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.40625,
      "epoch": 0.7563297255508311,
      "grad_norm": 0.0,
      "learning_rate": 1.2185929648241205e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15653
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.53125,
      "epoch": 0.7563780440664862,
      "grad_norm": 0.0,
      "learning_rate": 1.2183513722458446e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15654
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 125.4375,
      "epoch": 0.7564263625821415,
      "grad_norm": 0.0,
      "learning_rate": 1.2181097796675685e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15655
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.6875,
      "epoch": 0.7564746810977967,
      "grad_norm": 0.0,
      "learning_rate": 1.2178681870892926e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15656
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.84375,
      "epoch": 0.7565229996134519,
      "grad_norm": 0.0,
      "learning_rate": 1.2176265945110165e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15657
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.34375,
      "epoch": 0.756571318129107,
      "grad_norm": 0.0,
      "learning_rate": 1.2173850019327406e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15658
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.625,
      "epoch": 0.7566196366447623,
      "grad_norm": 0.0,
      "learning_rate": 1.2171434093544645e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15659
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.96875,
      "epoch": 0.7566679551604175,
      "grad_norm": 0.0,
      "learning_rate": 1.2169018167761886e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15660
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.0625,
      "epoch": 0.7567162736760726,
      "grad_norm": 0.0,
      "learning_rate": 1.2166602241979128e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15661
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.875,
      "epoch": 0.7567645921917279,
      "grad_norm": 0.0,
      "learning_rate": 1.2164186316196366e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15662
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 91.5,
      "epoch": 0.7568129107073831,
      "grad_norm": 0.0,
      "learning_rate": 1.2161770390413605e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15663
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.34375,
      "epoch": 0.7568612292230382,
      "grad_norm": 0.0,
      "learning_rate": 1.2159354464630846e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15664
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 100.53125,
      "epoch": 0.7569095477386935,
      "grad_norm": 0.0,
      "learning_rate": 1.2156938538848085e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15665
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 123.90625,
      "epoch": 0.7569578662543487,
      "grad_norm": 0.0,
      "learning_rate": 1.2154522613065326e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15666
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.09375,
      "epoch": 0.7570061847700039,
      "grad_norm": 0.0,
      "learning_rate": 1.2152106687282568e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15667
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 114.0625,
      "epoch": 0.757054503285659,
      "grad_norm": 12.210088478020394,
      "learning_rate": 1.2149690761499806e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.3072218894958496,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15668
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.0,
      "epoch": 0.7571028218013143,
      "grad_norm": 0.0,
      "learning_rate": 1.2147274835717045e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15669
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 121.3125,
      "epoch": 0.7571511403169695,
      "grad_norm": 6.426791479653457,
      "learning_rate": 1.2144858909934287e-07,
      "loss": 0.0,
      "reward": 2.1875,
      "reward_std": 0.28566449880599976,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15670
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.40625,
      "epoch": 0.7571994588326246,
      "grad_norm": 0.0,
      "learning_rate": 1.2142442984151528e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15671
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 88.625,
      "epoch": 0.7572477773482799,
      "grad_norm": 0.0,
      "learning_rate": 1.2140027058368767e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15672
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 124.1875,
      "epoch": 0.7572960958639351,
      "grad_norm": 0.0,
      "learning_rate": 1.2137611132586005e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15673
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 103.21875,
      "epoch": 0.7573444143795902,
      "grad_norm": 0.0,
      "learning_rate": 1.2135195206803247e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15674
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 126.8125,
      "epoch": 0.7573927328952454,
      "grad_norm": 0.0,
      "learning_rate": 1.2132779281020485e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15675
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.90625,
      "epoch": 0.7574410514109007,
      "grad_norm": 0.0,
      "learning_rate": 1.2130363355237727e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15676
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 114.96875,
      "epoch": 0.7574893699265559,
      "grad_norm": 0.0,
      "learning_rate": 1.2127947429454968e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15677
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.53125,
      "epoch": 0.757537688442211,
      "grad_norm": 0.0,
      "learning_rate": 1.2125531503672207e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15678
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.3125,
      "epoch": 0.7575860069578663,
      "grad_norm": 0.0,
      "learning_rate": 1.2123115577889446e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15679
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 91.75,
      "epoch": 0.7576343254735215,
      "grad_norm": 18.476019576833036,
      "learning_rate": 1.2120699652106687e-07,
      "loss": -0.0,
      "reward": 1.78125,
      "reward_std": 0.2505247890949249,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15680
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 119.6875,
      "epoch": 0.7576826439891766,
      "grad_norm": 31.674754045726587,
      "learning_rate": 1.2118283726323928e-07,
      "loss": -0.0,
      "reward": 1.40625,
      "reward_std": 0.21316051483154297,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 15681
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.9375,
      "epoch": 0.7577309625048319,
      "grad_norm": 0.0,
      "learning_rate": 1.2115867800541167e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15682
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.875,
      "epoch": 0.7577792810204871,
      "grad_norm": 0.0,
      "learning_rate": 1.2113451874758406e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15683
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.8125,
      "epoch": 0.7578275995361422,
      "grad_norm": 0.0,
      "learning_rate": 1.2111035948975647e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15684
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 101.4375,
      "epoch": 0.7578759180517974,
      "grad_norm": 0.0,
      "learning_rate": 1.2108620023192886e-07,
      "loss": 0.0,
      "reward": 1.296875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.046875,
      "rewards/format_reward": 1.0,
      "step": 15685
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.59375,
      "epoch": 0.7579242365674527,
      "grad_norm": 0.0,
      "learning_rate": 1.2106204097410127e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15686
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 121.5,
      "epoch": 0.7579725550831079,
      "grad_norm": 0.0,
      "learning_rate": 1.2103788171627369e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 15687
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.90625,
      "epoch": 0.758020873598763,
      "grad_norm": 0.0,
      "learning_rate": 1.2101372245844607e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15688
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 121.0625,
      "epoch": 0.7580691921144183,
      "grad_norm": 0.0,
      "learning_rate": 1.2098956320061846e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15689
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.71875,
      "epoch": 0.7581175106300735,
      "grad_norm": 0.0,
      "learning_rate": 1.2096540394279087e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15690
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 124.8125,
      "epoch": 0.7581658291457286,
      "grad_norm": 0.0,
      "learning_rate": 1.209412446849633e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15691
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 128.09375,
      "epoch": 0.7582141476613838,
      "grad_norm": 0.0,
      "learning_rate": 1.2091708542713567e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15692
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.6875,
      "epoch": 0.7582624661770391,
      "grad_norm": 0.0,
      "learning_rate": 1.2089292616930806e-07,
      "loss": 0.0,
      "reward": 1.984375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15693
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 107.59375,
      "epoch": 0.7583107846926942,
      "grad_norm": 0.0,
      "learning_rate": 1.2086876691148047e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15694
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 119.15625,
      "epoch": 0.7583591032083494,
      "grad_norm": 0.0,
      "learning_rate": 1.2084460765365286e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15695
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 116.34375,
      "epoch": 0.7584074217240047,
      "grad_norm": 11.402126922129057,
      "learning_rate": 1.2082044839582528e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.17782479524612427,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15696
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.125,
      "epoch": 0.7584557402396599,
      "grad_norm": 0.0,
      "learning_rate": 1.207962891379977e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15697
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 108.875,
      "epoch": 0.758504058755315,
      "grad_norm": 3.8681039001876534,
      "learning_rate": 1.2077212988017008e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15698
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.625,
      "epoch": 0.7585523772709702,
      "grad_norm": 0.0,
      "learning_rate": 1.2074797062234246e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15699
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 98.34375,
      "epoch": 0.7586006957866255,
      "grad_norm": 0.0,
      "learning_rate": 1.2072381136451488e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15700
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.875,
      "epoch": 0.7586490143022806,
      "grad_norm": 0.0,
      "learning_rate": 1.206996521066873e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15701
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.3125,
      "epoch": 0.7586973328179358,
      "grad_norm": 0.0,
      "learning_rate": 1.2067549284885968e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15702
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.65625,
      "epoch": 0.7587456513335911,
      "grad_norm": 0.0,
      "learning_rate": 1.206513335910321e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15703
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 89.5625,
      "epoch": 0.7587939698492462,
      "grad_norm": 0.0,
      "learning_rate": 1.2062717433320448e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15704
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.6875,
      "epoch": 0.7588422883649014,
      "grad_norm": 0.0,
      "learning_rate": 1.2060301507537687e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15705
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.46875,
      "epoch": 0.7588906068805567,
      "grad_norm": 0.0,
      "learning_rate": 1.2057885581754928e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15706
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 119.25,
      "epoch": 0.7589389253962119,
      "grad_norm": 0.0,
      "learning_rate": 1.205546965597217e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15707
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 128.84375,
      "epoch": 0.758987243911867,
      "grad_norm": 0.0,
      "learning_rate": 1.2053053730189408e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15708
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.71875,
      "epoch": 0.7590355624275222,
      "grad_norm": 0.0,
      "learning_rate": 1.2050637804406647e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15709
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.125,
      "epoch": 0.7590838809431775,
      "grad_norm": 0.0,
      "learning_rate": 1.2048221878623888e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15710
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.4375,
      "epoch": 0.7591321994588326,
      "grad_norm": 0.0,
      "learning_rate": 1.204580595284113e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15711
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.96875,
      "epoch": 0.7591805179744878,
      "grad_norm": 0.0,
      "learning_rate": 1.2043390027058368e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15712
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 107.53125,
      "epoch": 0.759228836490143,
      "grad_norm": 0.0,
      "learning_rate": 1.204097410127561e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15713
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 105.875,
      "epoch": 0.7592771550057982,
      "grad_norm": 0.0,
      "learning_rate": 1.2038558175492848e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15714
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.5,
      "epoch": 0.7593254735214534,
      "grad_norm": 0.0,
      "learning_rate": 1.2036142249710087e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15715
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 106.96875,
      "epoch": 0.7593737920371086,
      "grad_norm": 39.70660746479936,
      "learning_rate": 1.2033726323927328e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15716
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.8125,
      "epoch": 0.7594221105527639,
      "grad_norm": 0.0,
      "learning_rate": 1.203131039814457e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15717
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 104.1875,
      "epoch": 0.759470429068419,
      "grad_norm": 0.0,
      "learning_rate": 1.2028894472361808e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15718
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.0,
      "epoch": 0.7595187475840742,
      "grad_norm": 0.0,
      "learning_rate": 1.2026478546579047e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15719
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.28125,
      "epoch": 0.7595670660997295,
      "grad_norm": 0.0,
      "learning_rate": 1.2024062620796289e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15720
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.0,
      "epoch": 0.7596153846153846,
      "grad_norm": 0.0,
      "learning_rate": 1.202164669501353e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15721
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.4375,
      "epoch": 0.7596637031310398,
      "grad_norm": 0.0,
      "learning_rate": 1.2019230769230769e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15722
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 130.875,
      "epoch": 0.759712021646695,
      "grad_norm": 0.0,
      "learning_rate": 1.201681484344801e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.22673699259757996,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15723
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.78125,
      "epoch": 0.7597603401623502,
      "grad_norm": 0.0,
      "learning_rate": 1.2014398917665249e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15724
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 106.90625,
      "epoch": 0.7598086586780054,
      "grad_norm": 0.0,
      "learning_rate": 1.2011982991882487e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15725
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.0,
      "epoch": 0.7598569771936606,
      "grad_norm": 0.0,
      "learning_rate": 1.200956706609973e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15726
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 91.5625,
      "epoch": 0.7599052957093159,
      "grad_norm": 0.0,
      "learning_rate": 1.200715114031697e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15727
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.375,
      "epoch": 0.759953614224971,
      "grad_norm": 0.0,
      "learning_rate": 1.200473521453421e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15728
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.28125,
      "epoch": 0.7600019327406262,
      "grad_norm": 0.0,
      "learning_rate": 1.2002319288751448e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15729
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 100.84375,
      "epoch": 0.7600502512562815,
      "grad_norm": 21.024647441862424,
      "learning_rate": 1.199990336296869e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.2657213509082794,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15730
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.96875,
      "epoch": 0.7600985697719366,
      "grad_norm": 0.0,
      "learning_rate": 1.199748743718593e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15731
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.625,
      "epoch": 0.7601468882875918,
      "grad_norm": 0.0,
      "learning_rate": 1.199507151140317e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15732
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.125,
      "epoch": 0.760195206803247,
      "grad_norm": 0.0,
      "learning_rate": 1.199265558562041e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15733
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 92.6875,
      "epoch": 0.7602435253189022,
      "grad_norm": 0.0,
      "learning_rate": 1.199023965983765e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15734
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 128.875,
      "epoch": 0.7602918438345574,
      "grad_norm": 0.0,
      "learning_rate": 1.1987823734054888e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15735
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.46875,
      "epoch": 0.7603401623502126,
      "grad_norm": 0.0,
      "learning_rate": 1.198540780827213e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15736
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.28125,
      "epoch": 0.7603884808658679,
      "grad_norm": 0.0,
      "learning_rate": 1.198299188248937e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15737
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.53125,
      "epoch": 0.760436799381523,
      "grad_norm": 0.0,
      "learning_rate": 1.198057595670661e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1735912710428238,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15738
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.75,
      "epoch": 0.7604851178971782,
      "grad_norm": 0.0,
      "learning_rate": 1.1978160030923848e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15739
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 111.8125,
      "epoch": 0.7605334364128334,
      "grad_norm": 0.0,
      "learning_rate": 1.197574410514109e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15740
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 109.78125,
      "epoch": 0.7605817549284886,
      "grad_norm": 5.1482066934850605,
      "learning_rate": 1.197332817935833e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15741
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 89.96875,
      "epoch": 0.7606300734441438,
      "grad_norm": 0.0,
      "learning_rate": 1.197091225357557e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15742
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.40625,
      "epoch": 0.760678391959799,
      "grad_norm": 0.0,
      "learning_rate": 1.196849632779281e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15743
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 123.0,
      "epoch": 0.7607267104754541,
      "grad_norm": 0.0,
      "learning_rate": 1.196608040201005e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15744
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.0625,
      "epoch": 0.7607750289911094,
      "grad_norm": 0.0,
      "learning_rate": 1.1963664476227288e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15745
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.0,
      "epoch": 0.7608233475067646,
      "grad_norm": 0.0,
      "learning_rate": 1.196124855044453e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15746
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 107.53125,
      "epoch": 0.7608716660224198,
      "grad_norm": 5.3540515365980745,
      "learning_rate": 1.195883262466177e-07,
      "loss": -0.0,
      "reward": 2.25,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15747
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.40625,
      "epoch": 0.760919984538075,
      "grad_norm": 0.0,
      "learning_rate": 1.195641669887901e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15748
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 122.25,
      "epoch": 0.7609683030537302,
      "grad_norm": 0.0,
      "learning_rate": 1.195400077309625e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 15749
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 109.84375,
      "epoch": 0.7610166215693854,
      "grad_norm": 22.87377129419471,
      "learning_rate": 1.195158484731349e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.2962459325790405,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15750
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.1875,
      "epoch": 0.7610649400850406,
      "grad_norm": 0.0,
      "learning_rate": 1.194916892153073e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15751
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.8125,
      "epoch": 0.7611132586006958,
      "grad_norm": 0.0,
      "learning_rate": 1.194675299574797e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15752
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 90.1875,
      "epoch": 0.761161577116351,
      "grad_norm": 0.0,
      "learning_rate": 1.194433706996521e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15753
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 110.0,
      "epoch": 0.7612098956320061,
      "grad_norm": 0.0,
      "learning_rate": 1.194192114418245e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15754
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 108.375,
      "epoch": 0.7612582141476614,
      "grad_norm": 5.29333748096989,
      "learning_rate": 1.1939505218399689e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.17358146607875824,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15755
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 106.0,
      "epoch": 0.7613065326633166,
      "grad_norm": 0.0,
      "learning_rate": 1.193708929261693e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 15756
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.5,
      "epoch": 0.7613548511789718,
      "grad_norm": 0.0,
      "learning_rate": 1.193467336683417e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15757
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.46875,
      "epoch": 0.761403169694627,
      "grad_norm": 0.0,
      "learning_rate": 1.193225744105141e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15758
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.875,
      "epoch": 0.7614514882102822,
      "grad_norm": 0.0,
      "learning_rate": 1.1929841515268651e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15759
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.875,
      "epoch": 0.7614998067259374,
      "grad_norm": 0.0,
      "learning_rate": 1.192742558948589e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15760
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.53125,
      "epoch": 0.7615481252415925,
      "grad_norm": 0.0,
      "learning_rate": 1.1925009663703131e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15761
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.1875,
      "epoch": 0.7615964437572478,
      "grad_norm": 0.0,
      "learning_rate": 1.192259373792037e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15762
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.03125,
      "epoch": 0.761644762272903,
      "grad_norm": 0.0,
      "learning_rate": 1.192017781213761e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15763
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.09375,
      "epoch": 0.7616930807885581,
      "grad_norm": 0.0,
      "learning_rate": 1.191776188635485e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15764
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 105.59375,
      "epoch": 0.7617413993042134,
      "grad_norm": 2.612579878167978,
      "learning_rate": 1.191534596057209e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15765
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.875,
      "epoch": 0.7617897178198686,
      "grad_norm": 0.0,
      "learning_rate": 1.1912930034789332e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15766
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 96.4375,
      "epoch": 0.7618380363355238,
      "grad_norm": 0.0,
      "learning_rate": 1.1910514109006572e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15767
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.1875,
      "epoch": 0.761886354851179,
      "grad_norm": 0.0,
      "learning_rate": 1.190809818322381e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15768
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 123.625,
      "epoch": 0.7619346733668342,
      "grad_norm": 0.0,
      "learning_rate": 1.190568225744105e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15769
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.96875,
      "epoch": 0.7619829918824894,
      "grad_norm": 0.0,
      "learning_rate": 1.190326633165829e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15770
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.6875,
      "epoch": 0.7620313103981445,
      "grad_norm": 0.0,
      "learning_rate": 1.1900850405875532e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15771
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.8125,
      "epoch": 0.7620796289137998,
      "grad_norm": 0.0,
      "learning_rate": 1.1898434480092772e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15772
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.8125,
      "epoch": 0.762127947429455,
      "grad_norm": 0.0,
      "learning_rate": 1.189601855431001e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15773
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 125.875,
      "epoch": 0.7621762659451101,
      "grad_norm": 0.0,
      "learning_rate": 1.189360262852725e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15774
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.625,
      "epoch": 0.7622245844607654,
      "grad_norm": 0.0,
      "learning_rate": 1.1891186702744491e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15775
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.4375,
      "epoch": 0.7622729029764206,
      "grad_norm": 0.0,
      "learning_rate": 1.1888770776961732e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15776
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 87.65625,
      "epoch": 0.7623212214920758,
      "grad_norm": 0.0,
      "learning_rate": 1.1886354851178972e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15777
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 95.78125,
      "epoch": 0.7623695400077309,
      "grad_norm": 0.0,
      "learning_rate": 1.1883938925396211e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15778
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.0625,
      "epoch": 0.7624178585233862,
      "grad_norm": 0.0,
      "learning_rate": 1.1881522999613451e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15779
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 114.0625,
      "epoch": 0.7624661770390414,
      "grad_norm": 9.302739619506077,
      "learning_rate": 1.1879107073830691e-07,
      "loss": -0.0,
      "reward": 2.265625,
      "reward_std": 0.30250388383865356,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15780
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.84375,
      "epoch": 0.7625144955546965,
      "grad_norm": 0.0,
      "learning_rate": 1.1876691148047932e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15781
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.65625,
      "epoch": 0.7625628140703518,
      "grad_norm": 0.0,
      "learning_rate": 1.1874275222265172e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15782
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.6875,
      "epoch": 0.762611132586007,
      "grad_norm": 0.0,
      "learning_rate": 1.1871859296482411e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15783
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 93.25,
      "epoch": 0.7626594511016621,
      "grad_norm": 0.0,
      "learning_rate": 1.1869443370699651e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15784
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.875,
      "epoch": 0.7627077696173173,
      "grad_norm": 0.0,
      "learning_rate": 1.1867027444916891e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15785
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.0625,
      "epoch": 0.7627560881329726,
      "grad_norm": 0.0,
      "learning_rate": 1.1864611519134132e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15786
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.03125,
      "epoch": 0.7628044066486278,
      "grad_norm": 0.0,
      "learning_rate": 1.1862195593351372e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15787
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.3125,
      "epoch": 0.7628527251642829,
      "grad_norm": 0.0,
      "learning_rate": 1.1859779667568611e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15788
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.5,
      "epoch": 0.7629010436799382,
      "grad_norm": 0.0,
      "learning_rate": 1.1857363741785851e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570313811302185,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15789
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 110.25,
      "epoch": 0.7629493621955934,
      "grad_norm": 26.562577165266795,
      "learning_rate": 1.1854947816003091e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.4495011568069458,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15790
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.96875,
      "epoch": 0.7629976807112485,
      "grad_norm": 0.0,
      "learning_rate": 1.1852531890220333e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15791
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 102.5,
      "epoch": 0.7630459992269037,
      "grad_norm": 15.466791490741317,
      "learning_rate": 1.1850115964437573e-07,
      "loss": -0.0,
      "reward": 2.375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15792
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 111.75,
      "epoch": 0.763094317742559,
      "grad_norm": 5.9827027897840575,
      "learning_rate": 1.1847700038654813e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.22835706174373627,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15793
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.1875,
      "epoch": 0.7631426362582141,
      "grad_norm": 0.0,
      "learning_rate": 1.1845284112872051e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15794
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 94.1875,
      "epoch": 0.7631909547738693,
      "grad_norm": 0.0,
      "learning_rate": 1.1842868187089291e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.15992169082164764,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15795
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 113.78125,
      "epoch": 0.7632392732895246,
      "grad_norm": 8.394450889524016,
      "learning_rate": 1.1840452261306533e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.19266104698181152,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15796
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.6875,
      "epoch": 0.7632875918051798,
      "grad_norm": 0.0,
      "learning_rate": 1.1838036335523773e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15797
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.9375,
      "epoch": 0.7633359103208349,
      "grad_norm": 0.0,
      "learning_rate": 1.1835620409741013e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15798
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.84375,
      "epoch": 0.7633842288364902,
      "grad_norm": 0.0,
      "learning_rate": 1.1833204483958252e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15799
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 122.09375,
      "epoch": 0.7634325473521454,
      "grad_norm": 0.0,
      "learning_rate": 1.1830788558175492e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15800
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.0625,
      "epoch": 0.7634808658678005,
      "grad_norm": 0.0,
      "learning_rate": 1.1828372632392733e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15801
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 112.5625,
      "epoch": 0.7635291843834557,
      "grad_norm": 6.214367331884471,
      "learning_rate": 1.1825956706609973e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15802
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 116.40625,
      "epoch": 0.763577502899111,
      "grad_norm": 0.0,
      "learning_rate": 1.1823540780827213e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15803
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 97.6875,
      "epoch": 0.7636258214147661,
      "grad_norm": 0.0,
      "learning_rate": 1.1821124855044452e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15804
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.78125,
      "epoch": 0.7636741399304213,
      "grad_norm": 0.0,
      "learning_rate": 1.1818708929261692e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15805
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 119.375,
      "epoch": 0.7637224584460766,
      "grad_norm": 0.0,
      "learning_rate": 1.1816293003478933e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15806
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 105.75,
      "epoch": 0.7637707769617318,
      "grad_norm": 0.0,
      "learning_rate": 1.1813877077696173e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15807
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 97.375,
      "epoch": 0.7638190954773869,
      "grad_norm": 8.046349581072954,
      "learning_rate": 1.1811461151913413e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15808
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.90625,
      "epoch": 0.7638674139930421,
      "grad_norm": 0.0,
      "learning_rate": 1.1809045226130652e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15809
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 101.96875,
      "epoch": 0.7639157325086974,
      "grad_norm": 7.021578618853698,
      "learning_rate": 1.1806629300347892e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15810
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.46875,
      "epoch": 0.7639640510243525,
      "grad_norm": 0.0,
      "learning_rate": 1.1804213374565133e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15811
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.0,
      "epoch": 0.7640123695400077,
      "grad_norm": 0.0,
      "learning_rate": 1.1801797448782373e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15812
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.6875,
      "epoch": 0.764060688055663,
      "grad_norm": 0.0,
      "learning_rate": 1.1799381522999613e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15813
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 109.90625,
      "epoch": 0.7641090065713181,
      "grad_norm": 3.338534408960114,
      "learning_rate": 1.1796965597216852e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15814
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.75,
      "epoch": 0.7641573250869733,
      "grad_norm": 0.0,
      "learning_rate": 1.1794549671434092e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15815
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 105.6875,
      "epoch": 0.7642056436026285,
      "grad_norm": 0.0,
      "learning_rate": 1.1792133745651334e-07,
      "loss": 0.0,
      "reward": 1.3125,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.0625,
      "rewards/format_reward": 1.0,
      "step": 15816
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.90625,
      "epoch": 0.7642539621182838,
      "grad_norm": 0.0,
      "learning_rate": 1.1789717819868574e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15817
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 105.90625,
      "epoch": 0.7643022806339389,
      "grad_norm": 0.0,
      "learning_rate": 1.1787301894085814e-07,
      "loss": 0.0,
      "reward": 2.5,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.5,
      "rewards/format_reward": 1.0,
      "step": 15818
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 125.25,
      "epoch": 0.7643505991495941,
      "grad_norm": 0.0,
      "learning_rate": 1.1784885968303052e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 15819
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.15625,
      "epoch": 0.7643989176652494,
      "grad_norm": 0.0,
      "learning_rate": 1.1782470042520292e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15820
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 112.65625,
      "epoch": 0.7644472361809045,
      "grad_norm": 0.0,
      "learning_rate": 1.1780054116737534e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15821
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 112.90625,
      "epoch": 0.7644955546965597,
      "grad_norm": 8.396478023152259,
      "learning_rate": 1.1777638190954774e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15822
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 91.78125,
      "epoch": 0.764543873212215,
      "grad_norm": 0.0,
      "learning_rate": 1.1775222265172014e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15823
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 111.1875,
      "epoch": 0.7645921917278701,
      "grad_norm": 0.0,
      "learning_rate": 1.1772806339389253e-07,
      "loss": 0.0,
      "reward": 2.46875,
      "reward_std": 0.0883883461356163,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.46875,
      "rewards/format_reward": 1.0,
      "step": 15824
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 114.1875,
      "epoch": 0.7646405102435253,
      "grad_norm": 0.0,
      "learning_rate": 1.1770390413606493e-07,
      "loss": 0.0,
      "reward": 1.609375,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 15825
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 108.65625,
      "epoch": 0.7646888287591805,
      "grad_norm": 4.823875775012871,
      "learning_rate": 1.1767974487823734e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15826
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.59375,
      "epoch": 0.7647371472748358,
      "grad_norm": 0.0,
      "learning_rate": 1.1765558562040974e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15827
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 112.53125,
      "epoch": 0.7647854657904909,
      "grad_norm": 0.0,
      "learning_rate": 1.1763142636258214e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15828
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.875,
      "epoch": 0.7648337843061461,
      "grad_norm": 0.0,
      "learning_rate": 1.1760726710475453e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15829
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 102.625,
      "epoch": 0.7648821028218014,
      "grad_norm": 0.0,
      "learning_rate": 1.1758310784692693e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15830
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 120.625,
      "epoch": 0.7649304213374565,
      "grad_norm": 3.247535034573062,
      "learning_rate": 1.1755894858909934e-07,
      "loss": 0.0,
      "reward": 1.78125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15831
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 118.46875,
      "epoch": 0.7649787398531117,
      "grad_norm": 0.0,
      "learning_rate": 1.1753478933127174e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15832
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.78125,
      "epoch": 0.7650270583687669,
      "grad_norm": 0.0,
      "learning_rate": 1.1751063007344414e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.189372718334198,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15833
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.09375,
      "epoch": 0.7650753768844221,
      "grad_norm": 0.0,
      "learning_rate": 1.1748647081561654e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15834
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 129.03125,
      "epoch": 0.7651236954000773,
      "grad_norm": 2.8002577536931526,
      "learning_rate": 1.1746231155778893e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15835
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.5625,
      "epoch": 0.7651720139157325,
      "grad_norm": 0.0,
      "learning_rate": 1.1743815229996134e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15836
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 118.71875,
      "epoch": 0.7652203324313878,
      "grad_norm": 7.050440337199744,
      "learning_rate": 1.1741399304213374e-07,
      "loss": -0.0,
      "reward": 1.765625,
      "reward_std": 0.2820974886417389,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15837
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.71875,
      "epoch": 0.7652686509470429,
      "grad_norm": 0.0,
      "learning_rate": 1.1738983378430614e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15838
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.0,
      "epoch": 0.7653169694626981,
      "grad_norm": 0.0,
      "learning_rate": 1.1736567452647855e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1825428307056427,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15839
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.375,
      "epoch": 0.7653652879783533,
      "grad_norm": 0.0,
      "learning_rate": 1.1734151526865093e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15840
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 113.375,
      "epoch": 0.7654136064940085,
      "grad_norm": 4.994824070321802,
      "learning_rate": 1.1731735601082335e-07,
      "loss": -0.0,
      "reward": 2.015625,
      "reward_std": 0.19044627249240875,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15841
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 120.15625,
      "epoch": 0.7654619250096637,
      "grad_norm": 0.0,
      "learning_rate": 1.1729319675299575e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15842
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.3125,
      "epoch": 0.7655102435253189,
      "grad_norm": 0.0,
      "learning_rate": 1.1726903749516815e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15843
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 99.59375,
      "epoch": 0.765558562040974,
      "grad_norm": 0.0,
      "learning_rate": 1.1724487823734055e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1767766922712326,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 0.96875,
      "step": 15844
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 131.96875,
      "epoch": 0.7656068805566293,
      "grad_norm": 0.0,
      "learning_rate": 1.1722071897951293e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15845
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 109.65625,
      "epoch": 0.7656551990722845,
      "grad_norm": 0.0,
      "learning_rate": 1.1719655972168535e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15846
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 106.1875,
      "epoch": 0.7657035175879398,
      "grad_norm": 4.228498990891,
      "learning_rate": 1.1717240046385775e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16952534019947052,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15847
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 126.375,
      "epoch": 0.7657518361035949,
      "grad_norm": 10.236254003160052,
      "learning_rate": 1.1714824120603015e-07,
      "loss": -0.0,
      "reward": 1.953125,
      "reward_std": 0.17598573863506317,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15848
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.3125,
      "epoch": 0.7658001546192501,
      "grad_norm": 0.0,
      "learning_rate": 1.1712408194820255e-07,
      "loss": 0.0,
      "reward": 2.484375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.484375,
      "rewards/format_reward": 1.0,
      "step": 15849
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.84375,
      "epoch": 0.7658484731349053,
      "grad_norm": 0.0,
      "learning_rate": 1.1709992269037494e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15850
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.875,
      "epoch": 0.7658967916505605,
      "grad_norm": 0.0,
      "learning_rate": 1.1707576343254735e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15851
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 97.34375,
      "epoch": 0.7659451101662157,
      "grad_norm": 0.0,
      "learning_rate": 1.1705160417471975e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15852
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.25,
      "epoch": 0.7659934286818709,
      "grad_norm": 0.0,
      "learning_rate": 1.1702744491689215e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15853
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 110.46875,
      "epoch": 0.766041747197526,
      "grad_norm": 0.0,
      "learning_rate": 1.1700328565906455e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15854
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.875,
      "epoch": 0.7660900657131813,
      "grad_norm": 0.0,
      "learning_rate": 1.1697912640123694e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15855
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 96.09375,
      "epoch": 0.7661383842288365,
      "grad_norm": 0.0,
      "learning_rate": 1.1695496714340935e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15856
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 107.15625,
      "epoch": 0.7661867027444917,
      "grad_norm": 7.32485731680739,
      "learning_rate": 1.1693080788558175e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15857
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 113.9375,
      "epoch": 0.7662350212601469,
      "grad_norm": 0.0,
      "learning_rate": 1.1690664862775415e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15858
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 127.5,
      "epoch": 0.7662833397758021,
      "grad_norm": 0.0,
      "learning_rate": 1.1688248936992655e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15859
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 129.375,
      "epoch": 0.7663316582914573,
      "grad_norm": 6.747932787639775,
      "learning_rate": 1.1685833011209894e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.30935919284820557,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15860
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 119.40625,
      "epoch": 0.7663799768071125,
      "grad_norm": 26.197213346936046,
      "learning_rate": 1.1683417085427135e-07,
      "loss": -0.0,
      "reward": 1.4375,
      "reward_std": 0.19266104698181152,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15861
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 93.21875,
      "epoch": 0.7664282953227677,
      "grad_norm": 9.494633552262705,
      "learning_rate": 1.1681001159644375e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.2630178928375244,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15862
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 109.125,
      "epoch": 0.7664766138384229,
      "grad_norm": 0.0,
      "learning_rate": 1.1678585233861615e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15863
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.84375,
      "epoch": 0.766524932354078,
      "grad_norm": 0.0,
      "learning_rate": 1.1676169308078855e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15864
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 98.375,
      "epoch": 0.7665732508697333,
      "grad_norm": 0.0,
      "learning_rate": 1.1673753382296094e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.1530819982290268,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15865
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 114.71875,
      "epoch": 0.7666215693853885,
      "grad_norm": 3.4001738081048436,
      "learning_rate": 1.1671337456513336e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15866
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 117.875,
      "epoch": 0.7666698879010437,
      "grad_norm": 3.6097406539114285,
      "learning_rate": 1.1668921530730576e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15867
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 117.25,
      "epoch": 0.7667182064166989,
      "grad_norm": 7.148187103556772,
      "learning_rate": 1.1666505604947816e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.25069957971572876,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15868
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 105.09375,
      "epoch": 0.7667665249323541,
      "grad_norm": 5.181249022887514,
      "learning_rate": 1.1664089679165056e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.21316051483154297,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15869
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.125,
      "epoch": 0.7668148434480093,
      "grad_norm": 0.0,
      "learning_rate": 1.1661673753382296e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15870
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.0625,
      "epoch": 0.7668631619636644,
      "grad_norm": 0.0,
      "learning_rate": 1.1659257827599536e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1804211586713791,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15871
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 101.71875,
      "epoch": 0.7669114804793197,
      "grad_norm": 0.0,
      "learning_rate": 1.1656841901816776e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15872
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.75,
      "epoch": 0.7669597989949749,
      "grad_norm": 0.0,
      "learning_rate": 1.1654425976034016e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15873
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 109.59375,
      "epoch": 0.76700811751063,
      "grad_norm": 6.036855593045294,
      "learning_rate": 1.1652010050251256e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.30856335163116455,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15874
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.28125,
      "epoch": 0.7670564360262853,
      "grad_norm": 0.0,
      "learning_rate": 1.1649594124468496e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15875
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.75,
      "epoch": 0.7671047545419405,
      "grad_norm": 0.0,
      "learning_rate": 1.1647178198685736e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15876
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.78125,
      "epoch": 0.7671530730575957,
      "grad_norm": 0.0,
      "learning_rate": 1.1644762272902976e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15877
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.625,
      "epoch": 0.7672013915732508,
      "grad_norm": 0.0,
      "learning_rate": 1.1642346347120216e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15878
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.4375,
      "epoch": 0.7672497100889061,
      "grad_norm": 0.0,
      "learning_rate": 1.1639930421337456e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.1872510462999344,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15879
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 106.46875,
      "epoch": 0.7672980286045613,
      "grad_norm": 5.1917314684976965,
      "learning_rate": 1.1637514495554696e-07,
      "loss": -0.0,
      "reward": 1.5625,
      "reward_std": 0.22049658000469208,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15880
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 117.5625,
      "epoch": 0.7673463471202164,
      "grad_norm": 5.718033130338568,
      "learning_rate": 1.1635098569771936e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.23568855226039886,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15881
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 112.5625,
      "epoch": 0.7673946656358717,
      "grad_norm": 0.0,
      "learning_rate": 1.1632682643989176e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.16675157845020294,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15882
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 113.375,
      "epoch": 0.7674429841515269,
      "grad_norm": 6.465286870617337,
      "learning_rate": 1.1630266718206416e-07,
      "loss": -0.0,
      "reward": 2.046875,
      "reward_std": 0.329858660697937,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15883
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 101.8125,
      "epoch": 0.767491302667182,
      "grad_norm": 0.0,
      "learning_rate": 1.1627850792423656e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15884
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 117.6875,
      "epoch": 0.7675396211828373,
      "grad_norm": 4.034932700510494,
      "learning_rate": 1.1625434866640896e-07,
      "loss": -0.0,
      "reward": 1.421875,
      "reward_std": 0.20633062720298767,
      "rewards/accuracy_reward": 0.3125,
      "rewards/bonus_rewards": 0.109375,
      "rewards/format_reward": 1.0,
      "step": 15885
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 117.3125,
      "epoch": 0.7675879396984925,
      "grad_norm": 6.80044600381442,
      "learning_rate": 1.1623018940858136e-07,
      "loss": -0.0,
      "reward": 2.1875,
      "reward_std": 0.3029785752296448,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15886
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.875,
      "epoch": 0.7676362582141477,
      "grad_norm": 0.0,
      "learning_rate": 1.1620603015075376e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.16887323558330536,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15887
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.96875,
      "epoch": 0.7676845767298028,
      "grad_norm": 0.0,
      "learning_rate": 1.1618187089292616e-07,
      "loss": 0.0,
      "reward": 2.375,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15888
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 121.15625,
      "epoch": 0.7677328952454581,
      "grad_norm": 17.66393333813293,
      "learning_rate": 1.1615771163509856e-07,
      "loss": 0.0,
      "reward": 1.59375,
      "reward_std": 0.3945732116699219,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15889
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.125,
      "epoch": 0.7677812137611133,
      "grad_norm": 0.0,
      "learning_rate": 1.1613355237727097e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.19408093392848969,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15890
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 109.1875,
      "epoch": 0.7678295322767684,
      "grad_norm": 9.387041821312978,
      "learning_rate": 1.1610939311944337e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15891
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 114.0,
      "epoch": 0.7678778507924237,
      "grad_norm": 7.4661909676739615,
      "learning_rate": 1.1608523386161577e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15892
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 111.9375,
      "epoch": 0.7679261693080789,
      "grad_norm": 5.695959043757926,
      "learning_rate": 1.1606107460378817e-07,
      "loss": -0.0,
      "reward": 1.8125,
      "reward_std": 0.3197515308856964,
      "rewards/accuracy_reward": 0.625,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15893
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 111.90625,
      "epoch": 0.767974487823734,
      "grad_norm": 9.492065200912,
      "learning_rate": 1.1603691534596057e-07,
      "loss": -0.0,
      "reward": 2.171875,
      "reward_std": 0.3540124297142029,
      "rewards/accuracy_reward": 0.875,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15894
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 102.1875,
      "epoch": 0.7680228063393892,
      "grad_norm": 5.050580961774443,
      "learning_rate": 1.1601275608813297e-07,
      "loss": -0.0,
      "reward": 2.34375,
      "reward_std": 0.24359199404716492,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15895
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 118.5625,
      "epoch": 0.7680711248550445,
      "grad_norm": 4.9864115450472815,
      "learning_rate": 1.1598859683030537e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15896
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 112.90625,
      "epoch": 0.7681194433706997,
      "grad_norm": 5.85589500536388,
      "learning_rate": 1.1596443757247777e-07,
      "loss": 0.0,
      "reward": 2.3125,
      "reward_std": 0.27439430356025696,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15897
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 100.75,
      "epoch": 0.7681677618863548,
      "grad_norm": 6.253085974828295,
      "learning_rate": 1.1594027831465017e-07,
      "loss": 0.0,
      "reward": 2.140625,
      "reward_std": 0.2804819941520691,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15898
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.71875,
      "epoch": 0.7682160804020101,
      "grad_norm": 0.0,
      "learning_rate": 1.1591611905682257e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15899
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 107.1875,
      "epoch": 0.7682643989176653,
      "grad_norm": 4.371438185062381,
      "learning_rate": 1.1589195979899497e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.32195523381233215,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15900
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 100.03125,
      "epoch": 0.7683127174333204,
      "grad_norm": 0.0,
      "learning_rate": 1.1586780054116738e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15901
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 119.3125,
      "epoch": 0.7683610359489756,
      "grad_norm": 26.864105615070507,
      "learning_rate": 1.1584364128333977e-07,
      "loss": -0.0,
      "reward": 2.140625,
      "reward_std": 0.39202046394348145,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15902
    },
    {
      "all_correct": 0.0,
      "all_wrong": 1.0,
      "completion_length": 118.625,
      "epoch": 0.7684093544646309,
      "grad_norm": 0.0,
      "learning_rate": 1.1581948202551217e-07,
      "loss": 0.0,
      "reward": 1.0,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.0,
      "rewards/bonus_rewards": 0.0,
      "rewards/format_reward": 1.0,
      "step": 15903
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.0,
      "completion_length": 131.875,
      "epoch": 0.768457672980286,
      "grad_norm": 17.19781923916387,
      "learning_rate": 1.1579532276768457e-07,
      "loss": -0.0,
      "reward": 1.984375,
      "reward_std": 0.5891344547271729,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15904
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 120.0625,
      "epoch": 0.7685059914959412,
      "grad_norm": 6.3012564609396104,
      "learning_rate": 1.1577116350985697e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15905
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 104.40625,
      "epoch": 0.7685543100115965,
      "grad_norm": 8.22004353603884,
      "learning_rate": 1.1574700425202938e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15906
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 91.96875,
      "epoch": 0.7686026285272517,
      "grad_norm": 40.43229430067537,
      "learning_rate": 1.1572284499420177e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15907
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 115.4375,
      "epoch": 0.7686509470429068,
      "grad_norm": 7.935627510232085,
      "learning_rate": 1.1569868573637417e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.40983521938323975,
      "rewards/accuracy_reward": 0.8125,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15908
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 115.09375,
      "epoch": 0.768699265558562,
      "grad_norm": 4.572840908713934,
      "learning_rate": 1.1567452647854657e-07,
      "loss": 0.0,
      "reward": 2.25,
      "reward_std": 0.2873118817806244,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15909
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 104.75,
      "epoch": 0.7687475840742173,
      "grad_norm": 4.427497522732693,
      "learning_rate": 1.1565036722071897e-07,
      "loss": -0.0,
      "reward": 1.609375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15910
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.0625,
      "epoch": 0.7687959025898724,
      "grad_norm": 0.0,
      "learning_rate": 1.1562620796289139e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15911
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 111.75,
      "epoch": 0.7688442211055276,
      "grad_norm": 13.882803141542423,
      "learning_rate": 1.1560204870506377e-07,
      "loss": -0.0,
      "reward": 1.703125,
      "reward_std": 0.3457430303096771,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15912
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 110.96875,
      "epoch": 0.7688925396211829,
      "grad_norm": 5.020146813187969,
      "learning_rate": 1.1557788944723617e-07,
      "loss": 0.0,
      "reward": 1.921875,
      "reward_std": 0.2226182520389557,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15913
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.53125,
      "epoch": 0.768940858136838,
      "grad_norm": 0.0,
      "learning_rate": 1.1555373018940857e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.11100947856903076,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15914
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 93.96875,
      "epoch": 0.7689891766524932,
      "grad_norm": 21.71472479438641,
      "learning_rate": 1.1552957093158097e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15915
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 119.96875,
      "epoch": 0.7690374951681485,
      "grad_norm": 3.8677698297009493,
      "learning_rate": 1.1550541167375339e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15916
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.375,
      "epoch": 0.7690858136838037,
      "grad_norm": 0.0,
      "learning_rate": 1.1548125241592578e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15917
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 114.0,
      "epoch": 0.7691341321994588,
      "grad_norm": 0.0,
      "learning_rate": 1.1545709315809818e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15918
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 102.21875,
      "epoch": 0.769182450715114,
      "grad_norm": 0.0,
      "learning_rate": 1.1543293390027058e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15919
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.4375,
      "epoch": 0.7692307692307693,
      "grad_norm": 0.0,
      "learning_rate": 1.1540877464244298e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15920
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 116.09375,
      "epoch": 0.7692790877464244,
      "grad_norm": 14.579811479104324,
      "learning_rate": 1.1538461538461539e-07,
      "loss": -0.0,
      "reward": 1.9375,
      "reward_std": 0.2587745785713196,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15921
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 127.90625,
      "epoch": 0.7693274062620796,
      "grad_norm": 0.0,
      "learning_rate": 1.1536045612678778e-07,
      "loss": 0.0,
      "reward": 1.96875,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15922
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 113.09375,
      "epoch": 0.7693757247777349,
      "grad_norm": 0.0,
      "learning_rate": 1.1533629686896018e-07,
      "loss": 0.0,
      "reward": 1.34375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.09375,
      "rewards/format_reward": 1.0,
      "step": 15923
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.96875,
      "epoch": 0.76942404329339,
      "grad_norm": 0.0,
      "learning_rate": 1.1531213761113258e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15924
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 101.1875,
      "epoch": 0.7694723618090452,
      "grad_norm": 0.0,
      "learning_rate": 1.1528797835330498e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15925
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 93.59375,
      "epoch": 0.7695206803247004,
      "grad_norm": 0.0,
      "learning_rate": 1.1526381909547739e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15926
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 117.375,
      "epoch": 0.7695689988403557,
      "grad_norm": 0.0,
      "learning_rate": 1.1523965983764978e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15927
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 109.9375,
      "epoch": 0.7696173173560108,
      "grad_norm": 3.5361387452303097,
      "learning_rate": 1.1521550057982218e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15928
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 116.5,
      "epoch": 0.769665635871666,
      "grad_norm": 0.0,
      "learning_rate": 1.1519134132199458e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15929
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 114.90625,
      "epoch": 0.7697139543873213,
      "grad_norm": 0.0,
      "learning_rate": 1.1516718206416698e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15930
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 121.1875,
      "epoch": 0.7697622729029764,
      "grad_norm": 0.0,
      "learning_rate": 1.151430228063394e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15931
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 104.40625,
      "epoch": 0.7698105914186316,
      "grad_norm": 0.0,
      "learning_rate": 1.1511886354851178e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15932
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 108.28125,
      "epoch": 0.7698589099342869,
      "grad_norm": 0.0,
      "learning_rate": 1.1509470429068418e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15933
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 118.59375,
      "epoch": 0.769907228449942,
      "grad_norm": 5.5442952375214265,
      "learning_rate": 1.1507054503285658e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.26409149169921875,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15934
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 102.4375,
      "epoch": 0.7699555469655972,
      "grad_norm": 0.0,
      "learning_rate": 1.1504638577502898e-07,
      "loss": 0.0,
      "reward": 1.9375,
      "reward_std": 0.1962026059627533,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15935
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 113.5625,
      "epoch": 0.7700038654812524,
      "grad_norm": 17.52793472793996,
      "learning_rate": 1.150222265172014e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15936
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.625,
      "epoch": 0.7700521839969077,
      "grad_norm": 0.0,
      "learning_rate": 1.149980672593738e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15937
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 117.375,
      "epoch": 0.7701005025125628,
      "grad_norm": 0.0,
      "learning_rate": 1.1497390800154618e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15938
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 104.6875,
      "epoch": 0.770148821028218,
      "grad_norm": 0.0,
      "learning_rate": 1.1494974874371858e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15939
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 124.125,
      "epoch": 0.7701971395438733,
      "grad_norm": 0.0,
      "learning_rate": 1.1492558948589098e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15940
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 115.9375,
      "epoch": 0.7702454580595284,
      "grad_norm": 0.0,
      "learning_rate": 1.149014302280634e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15941
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 99.5625,
      "epoch": 0.7702937765751836,
      "grad_norm": 3.6989855544921717,
      "learning_rate": 1.148772709702358e-07,
      "loss": 0.0,
      "reward": 1.734375,
      "reward_std": 0.2414703369140625,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15942
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 90.84375,
      "epoch": 0.7703420950908388,
      "grad_norm": 0.0,
      "learning_rate": 1.1485311171240819e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.10205793380737305,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15943
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 107.78125,
      "epoch": 0.770390413606494,
      "grad_norm": 0.0,
      "learning_rate": 1.1482895245458059e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15944
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 116.03125,
      "epoch": 0.7704387321221492,
      "grad_norm": 3.179400204241764,
      "learning_rate": 1.1480479319675299e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15945
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 96.375,
      "epoch": 0.7704870506378044,
      "grad_norm": 0.0,
      "learning_rate": 1.147806339389254e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15946
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 112.4375,
      "epoch": 0.7705353691534597,
      "grad_norm": 13.711384585213187,
      "learning_rate": 1.147564746810978e-07,
      "loss": 0.0,
      "reward": 2.265625,
      "reward_std": 0.31112992763519287,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.328125,
      "rewards/format_reward": 1.0,
      "step": 15947
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 113.40625,
      "epoch": 0.7705836876691148,
      "grad_norm": 5.2793798902971245,
      "learning_rate": 1.1473231542327019e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.23144522309303284,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15948
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 97.40625,
      "epoch": 0.77063200618477,
      "grad_norm": 0.0,
      "learning_rate": 1.1470815616544259e-07,
      "loss": 0.0,
      "reward": 2.390625,
      "reward_std": 0.1315089464187622,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15949
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 99.375,
      "epoch": 0.7706803247004252,
      "grad_norm": 0.0,
      "learning_rate": 1.1468399690761499e-07,
      "loss": 0.0,
      "reward": 2.359375,
      "reward_std": 0.1893727034330368,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15950
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 117.1875,
      "epoch": 0.7707286432160804,
      "grad_norm": 7.5114864270044865,
      "learning_rate": 1.146598376497874e-07,
      "loss": -0.0,
      "reward": 1.828125,
      "reward_std": 0.3457430303096771,
      "rewards/accuracy_reward": 0.59375,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15951
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 102.84375,
      "epoch": 0.7707769617317356,
      "grad_norm": 12.283700054821667,
      "learning_rate": 1.146356783919598e-07,
      "loss": 0.0,
      "reward": 1.828125,
      "reward_std": 0.14846687018871307,
      "rewards/accuracy_reward": 0.5625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15952
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 121.8125,
      "epoch": 0.7708252802473908,
      "grad_norm": 11.69138386308071,
      "learning_rate": 1.1461151913413219e-07,
      "loss": -0.0,
      "reward": 1.65625,
      "reward_std": 0.4401872754096985,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15953
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 114.3125,
      "epoch": 0.770873598763046,
      "grad_norm": 2.67719113200921,
      "learning_rate": 1.1458735987630459e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.19149437546730042,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15954
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.84375,
      "epoch": 0.7709219172787012,
      "grad_norm": 0.0,
      "learning_rate": 1.1456320061847699e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15955
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 116.5,
      "epoch": 0.7709702357943564,
      "grad_norm": 18.981326170454746,
      "learning_rate": 1.145390413606494e-07,
      "loss": -0.0,
      "reward": 2.15625,
      "reward_std": 0.3823235034942627,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15956
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 106.09375,
      "epoch": 0.7710185543100117,
      "grad_norm": 10.329107193345473,
      "learning_rate": 1.145148821028218e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.18600594997406006,
      "rewards/accuracy_reward": 0.90625,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15957
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 117.53125,
      "epoch": 0.7710668728256668,
      "grad_norm": 4.084209688056807,
      "learning_rate": 1.1449072284499419e-07,
      "loss": 0.0,
      "reward": 1.359375,
      "reward_std": 0.19939783215522766,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15958
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 109.03125,
      "epoch": 0.771115191341322,
      "grad_norm": 5.688101385202921,
      "learning_rate": 1.1446656358716659e-07,
      "loss": 0.0,
      "reward": 2.28125,
      "reward_std": 0.3304150104522705,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15959
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 121.34375,
      "epoch": 0.7711635098569772,
      "grad_norm": 3.979352125412115,
      "learning_rate": 1.1444240432933899e-07,
      "loss": 0.0,
      "reward": 2.078125,
      "reward_std": 0.26196980476379395,
      "rewards/accuracy_reward": 0.78125,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15960
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 95.5,
      "epoch": 0.7712118283726324,
      "grad_norm": 0.0,
      "learning_rate": 1.144182450715114e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.06681530922651291,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15961
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 108.03125,
      "epoch": 0.7712601468882876,
      "grad_norm": 0.0,
      "learning_rate": 1.1439408581368381e-07,
      "loss": 0.0,
      "reward": 1.6875,
      "reward_std": 0.10888782143592834,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15962
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 125.09375,
      "epoch": 0.7713084654039428,
      "grad_norm": 7.745810502561791,
      "learning_rate": 1.143699265558562e-07,
      "loss": 0.0,
      "reward": 1.625,
      "reward_std": 0.36278265714645386,
      "rewards/accuracy_reward": 0.4375,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15963
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 113.78125,
      "epoch": 0.7713567839195979,
      "grad_norm": 0.0,
      "learning_rate": 1.143457672980286e-07,
      "loss": 0.0,
      "reward": 1.71875,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15964
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 111.9375,
      "epoch": 0.7714051024352532,
      "grad_norm": 3.7036915586983277,
      "learning_rate": 1.14321608040201e-07,
      "loss": 0.0,
      "reward": 1.421875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.28125,
      "rewards/bonus_rewards": 0.140625,
      "rewards/format_reward": 1.0,
      "step": 15965
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 109.0625,
      "epoch": 0.7714534209509084,
      "grad_norm": 8.534574512266147,
      "learning_rate": 1.1429744878237341e-07,
      "loss": 0.0,
      "reward": 2.15625,
      "reward_std": 0.3014557361602783,
      "rewards/accuracy_reward": 0.84375,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15966
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 122.53125,
      "epoch": 0.7715017394665636,
      "grad_norm": 0.0,
      "learning_rate": 1.1427328952454581e-07,
      "loss": 0.0,
      "reward": 2.046875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.296875,
      "rewards/format_reward": 1.0,
      "step": 15967
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 127.46875,
      "epoch": 0.7715500579822188,
      "grad_norm": 5.096867935018937,
      "learning_rate": 1.142491302667182e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15968
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 111.59375,
      "epoch": 0.771598376497874,
      "grad_norm": 5.452951164723941,
      "learning_rate": 1.142249710088906e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.19727616012096405,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.21875,
      "rewards/format_reward": 1.0,
      "step": 15969
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 97.375,
      "epoch": 0.7716466950135292,
      "grad_norm": 0.0,
      "learning_rate": 1.14200811751063e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15970
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 106.375,
      "epoch": 0.7716950135291843,
      "grad_norm": 0.0,
      "learning_rate": 1.1417665249323541e-07,
      "loss": 0.0,
      "reward": 2.125,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15971
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 94.1875,
      "epoch": 0.7717433320448396,
      "grad_norm": 0.0,
      "learning_rate": 1.1415249323540781e-07,
      "loss": 0.0,
      "reward": 2.03125,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.28125,
      "rewards/format_reward": 1.0,
      "step": 15972
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 98.5,
      "epoch": 0.7717916505604948,
      "grad_norm": 0.0,
      "learning_rate": 1.1412833397758021e-07,
      "loss": 0.0,
      "reward": 2.109375,
      "reward_std": 0.04419417306780815,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.359375,
      "rewards/format_reward": 1.0,
      "step": 15973
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 108.5,
      "epoch": 0.7718399690761499,
      "grad_norm": 32.39235217990139,
      "learning_rate": 1.141041747197526e-07,
      "loss": -0.0,
      "reward": 1.96875,
      "reward_std": 0.49022963643074036,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15974
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 113.875,
      "epoch": 0.7718882875918052,
      "grad_norm": 0.0,
      "learning_rate": 1.14080015461925e-07,
      "loss": 0.0,
      "reward": 2.421875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.421875,
      "rewards/format_reward": 1.0,
      "step": 15975
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 92.1875,
      "epoch": 0.7719366061074604,
      "grad_norm": 0.0,
      "learning_rate": 1.1405585620409741e-07,
      "loss": 0.0,
      "reward": 2.40625,
      "reward_std": 0.1246790662407875,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.40625,
      "rewards/format_reward": 1.0,
      "step": 15976
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 111.59375,
      "epoch": 0.7719849246231156,
      "grad_norm": 0.0,
      "learning_rate": 1.1403169694626981e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.13363061845302582,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15977
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 122.8125,
      "epoch": 0.7720332431387708,
      "grad_norm": 47.119967036013,
      "learning_rate": 1.1400753768844221e-07,
      "loss": -0.0,
      "reward": 1.71875,
      "reward_std": 0.42243799567222595,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.1875,
      "rewards/format_reward": 1.0,
      "step": 15978
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 106.125,
      "epoch": 0.772081561654426,
      "grad_norm": 0.0,
      "learning_rate": 1.139833784306146e-07,
      "loss": 0.0,
      "reward": 1.703125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.203125,
      "rewards/format_reward": 1.0,
      "step": 15979
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.0625,
      "epoch": 0.7721298801700812,
      "grad_norm": 0.0,
      "learning_rate": 1.13959219172787e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.1962025910615921,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15980
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 103.625,
      "epoch": 0.7721781986857363,
      "grad_norm": 0.0,
      "learning_rate": 1.1393505991495941e-07,
      "loss": 0.0,
      "reward": 2.09375,
      "reward_std": 0.0578637570142746,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.34375,
      "rewards/format_reward": 1.0,
      "step": 15981
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 115.625,
      "epoch": 0.7722265172013916,
      "grad_norm": 0.0,
      "learning_rate": 1.1391090065713181e-07,
      "loss": 0.0,
      "reward": 1.328125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.078125,
      "rewards/format_reward": 1.0,
      "step": 15982
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 109.96875,
      "epoch": 0.7722748357170468,
      "grad_norm": 0.0,
      "learning_rate": 1.1388674139930421e-07,
      "loss": 0.0,
      "reward": 2.015625,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15983
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 103.4375,
      "epoch": 0.7723231542327019,
      "grad_norm": 0.0,
      "learning_rate": 1.138625821414766e-07,
      "loss": 0.0,
      "reward": 2.4375,
      "reward_std": 0.1462520956993103,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.4375,
      "rewards/format_reward": 1.0,
      "step": 15984
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.25,
      "completion_length": 111.6875,
      "epoch": 0.7723714727483572,
      "grad_norm": 11.535792704096623,
      "learning_rate": 1.13838422883649e-07,
      "loss": 0.0,
      "reward": 1.890625,
      "reward_std": 0.3748715817928314,
      "rewards/accuracy_reward": 0.65625,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 0.96875,
      "step": 15985
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 93.90625,
      "epoch": 0.7724197912640124,
      "grad_norm": 0.0,
      "learning_rate": 1.1381426362582142e-07,
      "loss": 0.0,
      "reward": 1.75,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15986
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 105.21875,
      "epoch": 0.7724681097796676,
      "grad_norm": 0.0,
      "learning_rate": 1.1379010436799382e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15987
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 116.0,
      "epoch": 0.7725164282953227,
      "grad_norm": 0.0,
      "learning_rate": 1.1376594511016622e-07,
      "loss": 0.0,
      "reward": 1.671875,
      "reward_std": 0.12255740165710449,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15988
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.5,
      "completion_length": 110.71875,
      "epoch": 0.772564746810978,
      "grad_norm": 0.0,
      "learning_rate": 1.137417858523386e-07,
      "loss": 0.0,
      "reward": 1.65625,
      "reward_std": 0.1293872892856598,
      "rewards/accuracy_reward": 0.5,
      "rewards/bonus_rewards": 0.15625,
      "rewards/format_reward": 1.0,
      "step": 15989
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 99.625,
      "epoch": 0.7726130653266332,
      "grad_norm": 0.0,
      "learning_rate": 1.13717626594511e-07,
      "loss": 0.0,
      "reward": 2.0,
      "reward_std": 0.17570312321186066,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.25,
      "rewards/format_reward": 1.0,
      "step": 15990
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 110.15625,
      "epoch": 0.7726613838422883,
      "grad_norm": 5.269112005890601,
      "learning_rate": 1.1369346733668342e-07,
      "loss": 0.0,
      "reward": 2.328125,
      "reward_std": 0.22835706174373627,
      "rewards/accuracy_reward": 0.9375,
      "rewards/bonus_rewards": 0.390625,
      "rewards/format_reward": 1.0,
      "step": 15991
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.75,
      "completion_length": 106.34375,
      "epoch": 0.7727097023579436,
      "grad_norm": 0.0,
      "learning_rate": 1.1366930807885582e-07,
      "loss": 0.0,
      "reward": 1.375,
      "reward_std": 0.0,
      "rewards/accuracy_reward": 0.25,
      "rewards/bonus_rewards": 0.125,
      "rewards/format_reward": 1.0,
      "step": 15992
    },
    {
      "all_correct": 1.0,
      "all_wrong": 0.0,
      "completion_length": 108.9375,
      "epoch": 0.7727580208735988,
      "grad_norm": 0.0,
      "learning_rate": 1.1364514882102822e-07,
      "loss": 0.0,
      "reward": 2.453125,
      "reward_std": 0.0646936446428299,
      "rewards/accuracy_reward": 1.0,
      "rewards/bonus_rewards": 0.453125,
      "rewards/format_reward": 1.0,
      "step": 15993
    },
    {
      "all_correct": 0.25,
      "all_wrong": 0.5,
      "completion_length": 123.0625,
      "epoch": 0.7728063393892539,
      "grad_norm": 24.953209306258607,
      "learning_rate": 1.136209895632006e-07,
      "loss": 0.0,
      "reward": 1.640625,
      "reward_std": 0.1983242630958557,
      "rewards/accuracy_reward": 0.46875,
      "rewards/bonus_rewards": 0.171875,
      "rewards/format_reward": 1.0,
      "step": 15994
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 111.96875,
      "epoch": 0.7728546579049091,
      "grad_norm": 39.1354438189142,
      "learning_rate": 1.13596830305373e-07,
      "loss": -0.0,
      "reward": 2.0,
      "reward_std": 0.16366341710090637,
      "rewards/accuracy_reward": 0.6875,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15995
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.0,
      "completion_length": 122.25,
      "epoch": 0.7729029764205644,
      "grad_norm": 4.988683779297053,
      "learning_rate": 1.1357267104754542e-07,
      "loss": 0.0,
      "reward": 1.953125,
      "reward_std": 0.4336168169975281,
      "rewards/accuracy_reward": 0.71875,
      "rewards/bonus_rewards": 0.234375,
      "rewards/format_reward": 1.0,
      "step": 15996
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 104.84375,
      "epoch": 0.7729512949362196,
      "grad_norm": 4.079559644568478,
      "learning_rate": 1.1354851178971782e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 15997
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.25,
      "completion_length": 100.75,
      "epoch": 0.7729996134518747,
      "grad_norm": 0.0,
      "learning_rate": 1.1352435253189022e-07,
      "loss": 0.0,
      "reward": 2.0625,
      "reward_std": 0.1157275140285492,
      "rewards/accuracy_reward": 0.75,
      "rewards/bonus_rewards": 0.3125,
      "rewards/format_reward": 1.0,
      "step": 15998
    },
    {
      "all_correct": 0.75,
      "all_wrong": 0.0,
      "completion_length": 125.375,
      "epoch": 0.77304793196753,
      "grad_norm": 11.27004284494836,
      "learning_rate": 1.1350019327406261e-07,
      "loss": 0.0,
      "reward": 2.34375,
      "reward_std": 0.25513991713523865,
      "rewards/accuracy_reward": 0.96875,
      "rewards/bonus_rewards": 0.375,
      "rewards/format_reward": 1.0,
      "step": 15999
    },
    {
      "all_correct": 0.5,
      "all_wrong": 0.25,
      "completion_length": 108.53125,
      "epoch": 0.7730962504831852,
      "grad_norm": 2.8478905236885748,
      "learning_rate": 1.1347603401623501e-07,
      "loss": 0.0,
      "reward": 1.796875,
      "reward_std": 0.13258251547813416,
      "rewards/accuracy_reward": 0.53125,
      "rewards/bonus_rewards": 0.265625,
      "rewards/format_reward": 1.0,
      "step": 16000
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 20696,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}