作者位次 #4/10ACL 2026Poster· Natural Language Processing
A ttn PO : Attention-Guided Process Supervision for Efficient Reasoning
Shuaiyi Nie · Dingsiyu · Wenyuan Zhang · Linhao Yu · Tianmeng Yang · Yao Chen · Weichong Yin · YU SUN · hua wu · Tingwen Liu
摘要
Large reasoning models trained with reinforcement learning and verifiable rewards (RLVR) achieve strong performance on complex reasoning tasks, yet often overthink, generating redundant reasoning without performance gains. Existing trajectory-level length penalties often fail to effectively short...