Graph-Based Chain-of-Thought Pruning for Reducing Redundant Reflections in Reasoning LLMs

TL;DR

This paper introduces a graph-based framework to optimize chain-of-thought reasoning in large language models, significantly reducing redundant reflections and improving efficiency without sacrificing accuracy.

Contribution

It proposes converting linear reasoning traces into DAGs and applying dual pruning strategies, combined with a three-stage training pipeline, to enhance reasoning efficiency.

Findings

Reduces reasoning tokens by 42% on average.

Maintains or improves reasoning accuracy.

Effectively eliminates redundant reflection patterns.

Abstract

Extending CoT through RL has been widely used to enhance the reasoning capabilities of LLMs. However, due to the sparsity of reward signals, it can also induce undesirable thinking patterns such as overthinking, i.e., generating redundant intermediate reasoning content. In this work, we argue that a major source of such redundancy is inefficient reflection, which often manifests in two problematic patterns: Indiscriminate Reflection, where the model performs broad, low-impact checks throughout reasoning, and Repetitive Reflection, where it repeatedly re-verifies an already established conclusion. To address this, we introduce a graph-based CoT optimization framework. Specifically, we convert each linear CoT into a directed acyclic graph (DAG) with explicit dependency edges, and design a dual pruning strategy: branch-level pruning removes weakly contributing reflection branches, while depth-level pruning eliminates late-stage re-verification. We distill this behavior via a three-stage pipeline: (1) SFT to initialize the policy on pruned concise traces, (2) DPO to prefer correct but less redundant trajectories, and (3) GRPO with length penalty to jointly optimize answer correctness and efficiency. Experiments show that our approach reduces the average reasoning tokens by 42\% while maintaining or improving accuracy.