Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Reducing Credit Assignment Variance via Counterfactual Reasoning Paths

About

Reinforcement learning for multi-step reasoning with large language models (LLMs) typically relies on sparse terminal rewards, which creates a poorly conditioned credit-assignment problem: the final feedback is propagated uniformly across all intermediate decisions. This leads to high gradient variance, unstable training, and many ineffective updates, ultimately limiting sustained model improvement. We propose a counterfactual-comparison framework for credit assignment. For each input, the framework samples multiple reasoning trajectories and treats their differences as implicit approximations to alternative decisions. This yields an implicit process-level advantage estimator that converts sparse terminal rewards into step-sensitive learning signals. Building on this framework, we introduce Implicit Behavior Policy Optimization (IBPO), which substantially improves training stability and the performance ceiling on mathematical and code-reasoning benchmarks. Our results point to a promising direction for unlocking the reasoning potential of LLMs.

Fei Ding, Yongkang Zhang, Youwei Wang, Zijian Zeng• 2026

Related benchmarks

TaskDatasetResultRank
Mathematical ReasoningHMMT25
Accuracy (%)81.1
115
Code ReasoningLiveCodeBench
Accuracy76
90
Mathematical ReasoningAIME 25
Accuracy (%)94.4
14
Showing 3 of 3 rows

Other info

Follow for update