Know When to Stop: Segment-Level Credit Assignment for Reducing Overthinking

2026-07-01

AI Summary

This work addresses the inefficiency and inaccuracy in reasoning language models caused by excessive deliberation, which often leads to redundant or self-contradictory introspection. The authors propose a segment-level credit assignment mechanism that requires no additional annotations, leveraging cheap supervision signals derived from comparing intermediate reasoning steps with the ground-truth answer to dynamically assess each segment’s contribution to final correctness. Building on this, they introduce DASH (Drift-Aware Advantage SHaping), a method that reshapes the advantage function to steer the model toward productive reasoning pathways. Evaluated on challenging mathematical benchmarks such as AIME25, DASH achieves a 50.8% accuracy—substantially outperforming GRPO at 45.4%—while effectively curbing overthinking and enhancing the model’s self-correction capabilities.

Abstract

Reasoning language models frequently overthink: generating extended chains of behaviors such as hedging, approach abandonment, and self contradiction that consume tokens without improving answers. We show that these behaviors are not merely a consequence of length; even when controlling for response length, incorrect traces exhibit higher rates of unproductive self-reflection than correct ones. Addressing this requires identifying where self-reflection helps vs hurts, but obtaining these step-level annotations is costly. We observe that intermediate answer commitments within reasoning traces can provide a cheap proxy: by comparing each final answer candidate in the trace to the ground truth, we can determine whether subsequent reflection is productive without any additional supervision. Building on this insight, we propose DASH (Drift Aware advantage SHaping), which assigns segment-level credit based on whether each reasoning segment leads toward or away from correctness. On competition-level math benchmarks, DASH achieves the highest accuracy where overthinking is prevalent (AIME25: 50.8% vs. 45.4% GRPO) while reducing overthinking behaviors and achieving more productive self-correction than baselines.

Problem

• overthinking
• reasoning language models
• self-reflection
• credit assignment
• segment-level

Innovation

• credit assignment
• overthinking
• self-reflection
• reasoning traces
• DASH