Jev-Style 0.8B Decision v3

updated 7 days ago

Jev-style decisions on your laptop: a calibrated probability for every option in one pass. 0.8B, 25,600-token input, Apache-2.0.

chaoliangUNSW/Jev-Style-0.8B-Decision-v3

Text Classification • 0.8B • Updated 4 days ago • 467 • 4

Note Start here. Main model (transformers, full precision) with quick start, full results and protocols.
chaoliangUNSW/Jev-Style-0.8B-Decision-v3-GGUF

Text Classification • 0.8B • Updated 5 days ago • 2.36k • 11

Note For llama.cpp: F16, Q8_0 and Q4_K_M. Smallest download: 0.53 GB (Q4_K_M).
chaoliangUNSW/Jev-Style-0.8B-Decision-v3-MLX

Text Classification • Updated 5 days ago • 1

Note For Apple silicon: 8-bit (0.80 GB) and bf16 (1.50 GB) in one repo.
Running on Zero

Agents

1

Jev-Style v3

⚖

1

0.8B decisions, a probability for every option

Note Quick try: your own text and question, a probability for every option. No install needed.
Running on Zero

Agents

Jev-Style v3 Demo

⚖

Calibrated yes/no, choice and score answers

Note Use-case demos: triage, email, guardrails, RAG filter, moderation, routing, 51 languages, long documents, agent approval. Each tab turns the probabilities into an action and shows the rule.