PACT 论文：对齐 critic 与策略，提升 LLM 强化学习训练效果
PACT: From Credit Assignment to Critic Alignment
一篇把 LLM 强化学习里 credit assignment 讲清楚的论文，PACT 训练后在数学推理上比 GRPO 高 8.8 个点，搞 RL 后训练的值得看看推导。
论文提出 Completeness、Prefix Consistency、Neutrality 三个正则条件，证明它们唯一确定 token 级 credit 分配。基于这一理论，作者指出 On-Policy Distillation 的教师模型相当于隐式 critic，RLOO 信号在期望意义上与 token 级 credit 一致。由此提出 PACT 训练方法，采用 Actor-then-Critic 更新顺序并对 critic 训练施加重要性采样校正。在四个数学推理基准上平均准确率 72.87%，比 GRPO 和 PPO 分别高 8.80 和 13.16 个百分点，SWE-bench Verified 通过率 67.4%。
PACT: From Credit Assignment to Critic Alignment
Reinforcement learning has become a central component of large language model (LLM) post-training, yet token-level credit lacks a generally accepted mathematical definition, leaving its relationship to commonly used training signals unclear. We formulate three regularity conditions, namely Completeness, Prefix Consistency, and Neutrality, and prove that they uniquely determine token-level credit. This characterization provides a unified basis for explaining phenomena across existing algorithms and guides the development of an improved actor-critic training procedure. Through this lens, an ideal teacher in On-Policy Distillation (OPD) acts as an implicit critic, yielding an expected policy gradient proportional to that induced by token-level credit. Response-level REINFORCE Leave-One-Out (RLOO) signals match the expected policy-gradient contribution of token-level credit despite their coarser granularity. We further establish approximate credit sparsity under bounded outcome rewards and show how intermediate critic errors in Generalized Advantage Estimation (GAE) can become comparable to the underlying credit. These motivate Policy Aligned Critic Training (PACT), which adopts an Actor-then-Critic update order to apply importance sampling correction to critic training and better align the critic with the updated policy. In agentic mathematical reasoning, PACT achieves 72.87% average accuracy across four benchmarks, outperforming GRPO and PPO by 8.80 and 13.16 percentage points, respectively. On SWE-bench Verified, PACT achieves a pass rate of 67.4%, outperforming PPO, GRPO, and SAO by 2.4, 2.0, and 3.8 percentage points, respectively.
