CoolFace
Modelpublic

guinansu/AnyBudgetSearchAgent-Qwen3-4B-KimiLen-step240

sourceHugging Faceupdated 14d agoView on Hugging Face
0likes19downloads
2 commits on main
3dd07aa14d ago

Baseline B: Kimi k1.5 group-relative length reward (beta=0.5, mean-only advantage, tool-call count as length) from v3 step 200, GRPO step 240

guinansu
afe496914d ago

initial commit

guinansu