一作ACL 2026Findings· Findings of ACL
Step Potential Advantage Estimation: Harnessing Intermediate Confidence and Correctness for Efficient Mathematical Reasoning
Fei Wu · Zhenrong Zhang · Qikai Chang · Jianshu Zhang · Quan Liu · Jun Du
摘要
Reinforcement Learning with Verifiable Rewards (RLVR) elicits long chain-of-thought reasoning in large language models (LLMs), but outcome-based rewards lead to coarse-grained advantage estimation. While existing approaches improve RLVR via token-level entropy or sequence-level length control, th...