VeriThinker: Learning to Verify Makes Reasoning Model Efficient
This addresses the efficiency problem for users of large reasoning models by reducing computational costs, though it is an incremental improvement on existing methods.
The paper tackles the problem of Large Reasoning Models (LRMs) overthinking and producing unnecessarily long reasoning chains, which increases inference costs, by introducing VeriThinker, an approach that fine-tunes LRMs through an auxiliary verification task to compress reasoning chains. The result is a substantial reduction in reasoning tokens (e.g., from 3790 to 2125 on MATH500) while maintaining or slightly improving accuracy (e.g., from 94.0% to 94.8% on MATH500).
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