AI RESEARCH
Hidden States Know Where Reasoning Diverges: Credit Assignment via Span-Level Wasserstein Distance
arXiv CS.LG
•
ArXi:2604.23318v1 Announce Type: cross Group Relative Policy Optimization (GRPO) performs coarse-grained credit assignment in reinforcement learning with verifiable rewards (RLVR) by assigning the same advantage to all tokens in a rollout. Process reward models can provide finer-grained supervision, but they require step-level annotation or additional reward modeling. We show that hidden-state distributions contain a useful signal for local reasoning quality that can be extracted using only outcome-level correctness labels available in.