PRPO: Aligning Process Reward with Outcome Reward in Policy Optimization
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866912870272335872 |
|---|---|
| author | Ding, Ruiyi Lv, Yongxuan Meng, Xianhui Song, Jiahe Wang, Chao Jiang, Chen Cheng, Yuan |
| author_facet | Ding, Ruiyi Lv, Yongxuan Meng, Xianhui Song, Jiahe Wang, Chao Jiang, Chen Cheng, Yuan |
| contents | Policy optimization for large language models often suffers from sparse reward signals in multi-step reasoning tasks. Critic-free methods like GRPO assign a single normalized outcome reward to all tokens, providing limited guidance for intermediate reasoning . While Process Reward Models (PRMs) offer dense feedback, they risk premature collapse when used alone, as early low-reward tokens can drive policies toward truncated outputs. We introduce Process Relative Policy Optimization (PRPO), which combines outcome reliability with process-level guidance in a critic-free framework. PRPO segments reasoning sequences based on semantic clues, normalizes PRM scores into token-level advantages, and aligns their distribution with outcome advantages through location-parameter shift. On MATH500, PRPO improves Qwen2.5-Math-1.5B accuracy from 61.2% to 64.4% over GRPO using only eight rollouts and no value network, demonstrating efficient fine-grained credit assignment within critic-free optimization. Code is available at: https://github.com/SchumiDing/srpocode |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_07182 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | PRPO: Aligning Process Reward with Outcome Reward in Policy Optimization Ding, Ruiyi Lv, Yongxuan Meng, Xianhui Song, Jiahe Wang, Chao Jiang, Chen Cheng, Yuan Machine Learning Artificial Intelligence I.2.6; I.2.7 Policy optimization for large language models often suffers from sparse reward signals in multi-step reasoning tasks. Critic-free methods like GRPO assign a single normalized outcome reward to all tokens, providing limited guidance for intermediate reasoning . While Process Reward Models (PRMs) offer dense feedback, they risk premature collapse when used alone, as early low-reward tokens can drive policies toward truncated outputs. We introduce Process Relative Policy Optimization (PRPO), which combines outcome reliability with process-level guidance in a critic-free framework. PRPO segments reasoning sequences based on semantic clues, normalizes PRM scores into token-level advantages, and aligns their distribution with outcome advantages through location-parameter shift. On MATH500, PRPO improves Qwen2.5-Math-1.5B accuracy from 61.2% to 64.4% over GRPO using only eight rollouts and no value network, demonstrating efficient fine-grained credit assignment within critic-free optimization. Code is available at: https://github.com/SchumiDing/srpocode |
| title | PRPO: Aligning Process Reward with Outcome Reward in Policy Optimization |
| topic | Machine Learning Artificial Intelligence I.2.6; I.2.7 |
| url | https://arxiv.org/abs/2601.07182 |