bay-yearick-lab/grpo-standard-deviation-identity

Exact finite-group identity behind GRPO reward standardization, unifying GRPO / Dr. GRPO / DAPO for RLVR and LLM reasoning. Paper + code.

语言：TeX · GitHub 星标：4 · Fork：0

话题：dapo · dr-grpo · grpo · llm · policy-gradient · reasoning · reinforcement-learning · reward-shaping

在 GitHub 上查看原仓库