让大模型提前停止推理，提速又省力。
FlashThink: An Early Exit Method For Efficient Reasoning
- 用验证模型判断何时可提前终止推理
- 在4个基准上缩短77%以上推理内容，准确率不变
- 适合需要高效推理的部署场景
大型语言模型在推理任务中表现优异，但常生成过长的推理过程，造成显著计算开销。我们观察到，即使面对简单问题，模型仍会生成冗长推理，违背直觉。初步实验表明，在生成过程中某一时刻，模型已能正确输出答案，无需完成全部推理。为此，我们提出一种验证机制，精准识别可提前退出的时机。在四个不同基准上的综合实验表明，所提方法FlashThink能有效缩短推理内容，同时保持模型准确率。对Deepseek-R1和QwQ-32B模型，推理长度分别减少77.04%和77.47%，准确率未下降。
原文摘要 · Abstract (English)
Large Language Models (LLMs) have shown impressive performance in reasoning tasks. However, LLMs tend to generate excessively long reasoning content, leading to significant computational overhead. Our observations indicate that even on simple problems, LLMs tend to produce unnecessarily lengthy reasoning content, which is against intuitive expectations. Preliminary experiments show that at a certain point during the generation process, the model is already capable of producing the correct solution without completing the full reasoning content. Therefore, we consider that the reasoning process of the model can be exited early to achieve the purpose of efficient reasoning. We introduce a verification model that identifies the exact moment when the model can stop reasoning and still provide the correct answer. Comprehensive experiments on four different benchmarks demonstrate that our proposed method, FlashThink, effectively shortens the reasoning content while preserving the model accuracy. For the Deepseek-R1 and QwQ-32B models, we reduced the length of reasoning content by 77.04% and 77.47%, respectively, without reducing the accuracy.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。