作者位次 #7/10ACL 2026Poster· Natural Language Processing
Outcome-Grounded Advantage Reshaping for Fine-Grained Credit Assignment in Mathematical Reasoning
Ziheng Li · Liu Kang · Feng Xiao · Luxi Xing · Qingyi Si · Zhuoran Li · Weikang Gong · Deqing Yang · Yanghua Xiao · Hongcheng Guo
摘要
Group Relative Policy Optimization (GRPO) has emerged as a promising critic-free reinforcement learning paradigm for reasoning tasks. However, standard GRPO employs a coarse-grained credit assignment mechanism that propagates group-level rewards uniformly to to every token in a sequence, neglecti...