Nemotron-Cascade 2: Post-Training LLMs with Cascade RL and Multi-Domain On-Policy Distillation
Paper • 2603.19220 • Published • 70
Post training
-
-
Not Every Rubric Teaches Equally: Policy-Aware Rubric Rewards for RLVR
Paper • 2605.20164 • Published • 5 -
GoLongRL: Capability-Oriented Long Context Reinforcement Learning with Multitask Alignment
Paper • 2605.19577 • Published • 58 -
EnvFactory: Scaling Tool-Use Agents via Executable Environments Synthesis and Robust RL
Paper • 2605.18703 • Published • 49 -
Mid-Training with Self-Generated Data Improves Reinforcement Learning in Language Models
Paper • 2605.08472 • Published • 5 -
Self-Distilled Agentic Reinforcement Learning
Paper • 2605.15155 • Published • 117 -
Learning to Build the Environment: Self-Evolving Reasoning RL via Verifiable Environment Synthesis
Paper • 2605.14392 • Published • 9 -
RewardHarness: Self-Evolving Agentic Post-Training
Paper • 2605.08703 • Published • 239 -
RubricEM: Meta-RL with Rubric-guided Policy Decomposition beyond Verifiable Rewards
Paper • 2605.10899 • Published • 77 -
Towards On-Policy Data Evolution for Visual-Native Multimodal Deep Search Agents
Paper • 2605.10832 • Published • 21 -
Missing Old Logits in Asynchronous Agentic RL: Semantic Mismatch and Repair Methods for Off-Policy Correction
Paper • 2605.12070 • Published • 15 -
Reward Hacking in Rubric-Based Reinforcement Learning
Paper • 2605.12474 • Published • 4 -
AutoLLMResearch: Training Research Agents for Automating LLM Experiment Configuration -- Learning from Cheap, Optimizing Expensive
Paper • 2605.11518 • Published • 3 -
DeltaRubric: Generative Multimodal Reward Modeling via Joint Planning and Verification
Paper • 2605.09269 • Published • 4 -
DeepRefine: Agent-Compiled Knowledge Refinement via Reinforcement Learning
Paper • 2605.10488 • Published • 2 -
Flow-OPD: On-Policy Distillation for Flow Matching Models
Paper • 2605.08063 • Published • 99 -
OpenSeeker-v2: Pushing the Limits of Search Agents with Informative and High-Difficulty Trajectories
Paper • 2605.04036 • Published • 71 -
QUEST: Training Frontier Deep Research Agents with Fully Synthetic Tasks
Paper • 2605.24218 • Published • 43 -
Claw-Anything: Benchmarking Always-On Personal Assistants with Broader Access to User's Digital World
Paper • 2605.26086 • Published • 26 -
RUBRIC-ARROW: Alternating Pointwise Rubric Reward Modeling for LLM Post-training in Non-verifiable Domains
Paper • 2605.29156 • Published • 10 -
Verifiable Rewards Beyond Math and Code: Lightweight Corpus-Grounded Process Supervision for Factual Question Answering
Paper • 2605.29648 • Published • 7 -
LongTraceRL: Learning Long-Context Reasoning from Search Agent Trajectories with Rubric Rewards
Paper • 2605.31584 • Published • 40 -
GrepSeek: Training Search Agents for Direct Corpus Interaction
Paper • 2605.29307 • Published • 115 -
SAAS: Self-Aware Reinforcement Learning for Over-Search Mitigation in Agentic Search
Paper • 2605.29796 • Published • 21 -
OpenWebRL: Demystifying Online Multi-turn Reinforcement Learning for Visual Web Agents
Paper • 2606.02031 • Published • 20 -
When Does Multi-Agent RL Improve LLM Workflows? Workflow, Scale, and Policy-Sharing Tradeoffs
Paper • 2605.24202 • Published • 16 -
PaddleOCR-VL-1.6: Expanding the Frontier of Document Parsing with Under-Optimized Region Refinement and Progressive Post-Training
Paper • 2606.03264 • Published • 26 -
Reproducing, Analyzing, and Detecting Reward Hacking in Rubric-Based Reinforcement Learning
Paper • 2606.04923 • Published • 42 -
Reinforcement Learning from Rich Feedback with Distributional DAgger
Paper • 2606.05152 • Published • 4 -
Compress-Distill: Reasoning Trace Compression for Efficient Knowledge Distillation
Paper • 2606.05988 • Published • 3 -
Critic-R: Improving Agentic Search using Instruction-tuned Retrievers with Natural Language Introspective Feedback
Paper • 2606.00590 • Published • 1 -
SearchSwarm: Towards Delegation Intelligence in Agentic LLMs for Long-Horizon Deep Research
Paper • 2606.09730 • Published • 56 -
FORT-Searcher: Synthesizing Shortcut-Resistant Search Tasks for Training Deep Search Agents
Paper • 2606.12087 • Published • 50 -
EvoBrowseComp: Benchmarking Search Agents on Evolving Knowledge
Paper • 2606.13120 • Published • 4 -
Zone of Proximal Policy Optimization: Teacher in Prompts, Not Gradients
Paper • 2606.18216 • Published • 64 -
Dr-DCI: Scaling Direct Corpus Interaction via Dynamic Workspace Expansion
Paper • 2606.14885 • Published • 12 -
OpenThoughts-Agent: Data Recipes for Agentic Models
Paper • 2606.24855 • Published • 49 -
Holistic Data Scheduler for LLM Pre-training via Multi-Objective Reinforcement Learning
Paper • 2606.24133 • Published • 12 -
OPID: On-Policy Skill Distillation for Agentic Reinforcement Learning
Paper • 2606.26790 • Published • 59 -
Why Multi-Step Tool-Use Reinforcement Learning Collapses and How Supervisory Signals Fix It
Paper • 2606.26027 • Published • 19 -
Neglected Free Lunch from Post-training: Progress Advantage for LLM Agents
Paper • 2606.26080 • Published • 13 -
GBC: Gradient-Based Connections for Optimizing Multi-Agent Systems
Paper • 2606.28187 • Published • 13 -
Single-Rollout Asynchronous Optimization for Agentic Reinforcement Learning
Paper • 2607.07508 • Published • 33 -
Long-Horizon-Terminal-Bench: Testing the Limits of Agents on Long-Horizon Terminal Tasks with Dense Reward-Based Grading
Paper • 2607.08964 • Published • 78 -
Trust Region Policy Distillation
Paper • 2607.04751 • Published • 34 -
Principled Analysis of Deep Reinforcement Learning Evaluation and Design Paradigms
Paper • 2607.07769 • Published • 9 -
SEED: Self-Evolving On-Policy Distillation for Agentic Reinforcement Learning
Paper • 2607.14777 • Published • 105 -
LongStraw: Long-Context RL Beyond 2M Tokens under a Fixed GPU Budget
Paper • 2607.14952 • Published • 94 -
Demystifying On-Policy Distillation: Roles, Pathologies, and Regulations
Paper • 2607.13399 • Published • 18 -
SearchOS-V1: Towards Robust Open-Domain Information-Seeking Agent Collaboration
Paper • 2607.15257 • Published • 70 -
DeepSearch-World: Self-Distillation for Deep Search Agents in a Verifiable Environment
Paper • 2607.07820 • Published • 94 -
Molt: A Scalable PyTorch-Native Training Framework for Agentic Reinforcement Learning
Paper • 2607.21653 • Published • 33 -
SkillRise: Agentic Reinforcement Learning for Cross-Task Skill Evolution
Paper • 2607.26784 • Published • 28 -
Enhancing Rubric-based RL via Self-Distillation
Paper • 2607.18082 • Published • 25 -
From RLVR to RLSVR: Task Transformation Induces Self-Verifiable Rewards for Open-Ended LLM Self-Improvement
Paper • 2607.23802 • Published • 96 -
Weak-to-Strong On-Policy Distillation
Paper • 2607.26246 • Published • 58 -
Agent Lightning v1.0: Towards Harnessed Agentic RL
Paper • 2608.17528 • Published • 35 -
Agentic ESOpt: Fine-Tuning Long-Horizon LLM Agents with Minimal GPU Requirements
Paper • 2608.17310 • Published • 111 -
LEGO-RL: Harness-Native Reinforcement Learning for Coding Agents
Paper • 2608.17393 • Published • 27 -
Prime Agent: A Self-Improving RLM Harness
Paper • 2608.23552 • Published • 57 -
From Production Traffic to Post-Training: Building a Self-Hosted LLM That Covers the Corporate Request Mix
Paper • 2609.01572 • Published • 36 -
Rethinking On-Policy Distillation of Large Language Models II: One Training Example
Paper • 2609.04172 • Published • 105 -
Knowing When Not to Reuse: Conditional Experience Transfer in Autonomous LLM Post-Training
Paper • 2608.26730 • Published • 153 -
DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training
Paper • 2609.04094 • Published • 27 -
Small Language Models as Judges for Rubric-Based Reinforcement Learning
Paper • 2608.30005 • Published • 3 -
Does On-Policy Distillation Really Distill? From Noisy Teacher to Self-Improvement
Paper • 2608.31046 • Published • 99 -
RISE: Recursive Improvement via Self-Extrapolating Policy Distillation
Paper • 2609.05295 • Published • 19 -
Iris: Climbing to the Search Frontier
Paper • 2609.04304 • Published • 72 -
Revisiting Complete Reasoning Traces for Post-Training
Paper • 2609.07103 • Published • 22 -
Miles v0.1: Production-Level Post-Training
Paper • 2609.08368 • Published • 61 -
Online Draft Co-Training for Speculative Decoding in Large-Scale, Long-Context RL Post-Training
Paper • 2609.07108 • Published • 34 -
RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learning
Paper • 2609.20784 • Published • 59 -
MiMo-V2.6: Scaling Reinforcement Learning Towards Self-Improvement
Paper • 2610.11959 • Published • 92 -
More Methods, Better Agents? A Reality Check on Self-Distillation in Agentic Reinforcement Learning
Paper • 2610.19789 • Published • 82