TY  - RPRT
TI  - Utilizing and Calibrating Hindsight Process Rewards via Reinforcement with Mutual Information Self-Evaluation
AU  - Jiashu Yao
AU  - Heyan Huang
AU  - Zeming Liu
AU  - Yuhang Guo
PY  - 2026
UR  - https://arxiv.org/abs/2604.11611
ID  - 2604.11611
ER  -