让大模型学会判断题目难易，自动控制推理深度，减少冗余计算。
Think How to Think: Mitigating Overthinking with Autonomous Difficulty Cognition in Large Reasoning Models
- 分两阶段训练：先引导模型识别任务难度，再剔除冗余推理步骤。
- 在简单任务上推理成本降低70%以上，复杂任务降40%，性能稳定。
- 适合需要高效推理的场景，如实时问答、资源受限部署。
近期大型推理模型在复杂任务中表现优异，但常出现过度推理，生成过长且重复的推理过程。我们通过实证分析发现，这些模型在解题前难以像人类一样识别任务难度，导致采用统一推理策略。为此，本文提出一种名为TH2T的新方法，通过两阶段微调，逐步增强模型对任务难度和推理冗余的认知能力。首先，在输出前缀中注入难度提示，引导模型根据任务难易调整推理深度，训练数据融合短路径与长路径；其次，引入冗余提示，监督中间推理步骤以识别并消除无意义模式。在7B/14B/32B模型上的实验表明，该方法在简单任务上可将推理成本降低超70%，复杂任务降低40%，同时保持性能稳定。生成结果显示出明显的难度感知能力与显著减少的冗余（如自我反思和循环推理）。
原文摘要 · Abstract (English)
Recent Large Reasoning Models (LRMs) excel at complex reasoning tasks but often suffer from overthinking, generating overly long and redundant reasoning trajectories. To explore its essence, our empirical analysis reveals that LRMs are primarily limited to recognizing task properties (i.e., difficulty levels) like humans before solving the problem, leading to a one-size-fits-all reasoning process. Inspired by this, a pressing and natural question emerges: Can we explicitly bootstrap such ability to alleviate overthinking in LRMs? In this paper, we propose Think-How-to-Think (TH2T), a novel two-stage fine-tuning strategy that progressively inspires LRMs' difficulty cognition and redundancy cognition of LRMs. Specifically, we first inject difficulty hypnosis into output prefixes to guide the model toward adaptive reasoning depth, trained on a hybrid dataset mixing short and long reasoning paths. Then, we incorporate redundancy hypnosis, which supervises the intermediate reasoning steps to identify and eliminate unnecessary reasoning patterns. Experiments on 7B/14B/32B models demonstrate that TH2T significantly reduces inference costs by over 70% on easy tasks and 40% on hard tasks while maintaining performance stability. The resulting outputs exhibit clear signs of difficulty-aware capabilities and reduced redundancy (e.g., reflection and looping).
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。