VeriThinker: Learning to Verify Makes Reasoning Model Efficient

TL;DR

VeriThinker introduces a verification-based fine-tuning method for large reasoning models, significantly reducing reasoning chain length and inference costs while maintaining or improving accuracy, including zero-shot generalization.

Contribution

The paper proposes a novel CoT compression approach by fine-tuning LRMs through an auxiliary verification task, avoiding synthetic data generation.

Findings

Reduces reasoning tokens by over 40% on MATH500 and AIME25 datasets.

Achieves slight accuracy improvements while decreasing inference costs.

Demonstrates zero-shot generalization to speculative reasoning tasks.

Abstract

Large Reasoning Models (LRMs) excel at complex tasks using Chain-of-Thought (CoT) reasoning. However, their tendency to overthinking leads to unnecessarily lengthy reasoning chains, dramatically increasing inference costs. To mitigate this issue, we introduce VeriThinker, a novel approach for CoT compression. Unlike conventional methods that fine-tune LRMs directly on the original reasoning task using synthetic concise CoT data, we innovatively fine-tune the model solely through an auxiliary verification task. By training LRMs to accurately verify the correctness of CoT solutions, the LRMs inherently become more discerning about the necessity of subsequent self-reflection steps, thereby effectively suppressing overthinking. Extensive experiments validate that VeriThinker substantially reduces reasoning chain lengths while maintaining or even slightly improving accuracy. When applied to DeepSeek-R1-Distill-Qwen-7B, our approach reduces reasoning tokens on MATH500 from 3790 to 2125 while improving accuracy by 0.8% (94.0% to 94.8%), and on AIME25, tokens decrease from 14321 to 10287 with a 2.1% accuracy gain (38.7% to 40.8%). Additionally, our experiments demonstrate that VeriThinker can also be zero-shot generalized to speculative reasoning. Code is available at https://github.com/czg1225/VeriThinker