← 研踪 PaperTrailarXiv:2606.29296 · v3
Process Advantage Signal Shaping: Advantage Construction for Process-Supervised Reinforcement Learning in LLM Reasoners
首次提交 2026-06-28 17:36
版本更新 2026-11-17 20:13
雷达观测 2026-11-18 11:33
摘要
Dense process supervision provides local evaluations, but leaves open how these scores should become policy advantages. We introduce PASS (Process Advantage Signal Shaping), an advantage-construction method for group-relative policy optimization (GRPO). PASS separately normalizes and fuses reward channels, constructs batch-shared credit units from variation in the fused score trace, and normalizes return-to-go by the remaining unit count. These operations connect feedback scale, credit resolution, and accumulation horizon in a single computation. We evaluate learned process reward model (PRM) feedback for mathematical reasoning and teacher-derived token feedback for multi-hop question answering (QA). Across seven mathematics benchmarks, PASS improves average pass@1 by 4.62 percentage points over outcome-only GRPO with approximately shorter responses. Matched comparisons with Step-GRPO and component ablations further support the choice of shared credit units and residual normalization. On multi-hop QA, PASS improves over process-signal controls across two normalizers; token-level ablations reveal an accuracy--length trade-off: token-level accumulation slightly favors pass@1, while PASS achieves higher pass@8/16 with shorter responses. The results establish the conversion from local process scores to policy advantages as a consequential design choice across distinct supervision sources.
作者备注:27 pages, 5 figures
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v3 · 2026-11-17 20:13 · 最新观测版本
Process Advantage Signal Shaping: Advantage Construction for Process-Supervised Reinforcement Learning in LLM Reasoners
Dense process supervision provides local evaluations, but leaves open how these scores should become policy advantages. We introduce PASS (Process Advantage Signal Shaping), an advantage-construction method for group-relative policy optimization (GRPO). PASS separately normalizes and fuses reward channels, constructs batch-shared credit units from variation in the fused score trace, and normalizes return-to-go by the remaining unit count. These operations connect feedback scale, credit resolution, and accumulation horizon in a single computation. We evaluate learned process reward model (PRM) feedback for mathematical reasoning and teacher-derived token feedback for multi-hop question answering (QA). Across seven mathematics benchmarks, PASS improves average pass@1 by 4.62 percentage points over outcome-only GRPO with approximately shorter responses. Matched comparisons with Step-GRPO and component ablations further support the choice of shared credit units and residual normalization. On multi-hop QA, PASS improves over process-signal controls across two normalizers; token-level ablations reveal an accuracy--length trade-off: token-level accumulation slightly favors pass@1, while PASS achieves higher pass@8/16 with shorter responses. The results establish the conversion from local process scores to policy advantages as a consequential design choice across distinct supervision sources.
查看此版本解读 →查看此版本原文 ↗