Hint-Guided Diversified Policy Optimization improves LLM reasoning

Arxiv CS.CL (NLP/language models)

The paper introduces a method combining hint-level guidance with diversified sampling to enhance RLVR training for LLM reasoning. Experiments show significant gains on math reasoning benchmarks.

llmreinforcement-learning

Arxiv CS.CL (NLP/language models)·Zhiyu CaoKaixin WuMingjie ZhongPeifeng LiXiaobo LiCan YeQiaoming Zhu

8 days ago