让大模型根据题目难易自动调节推理长度，省资源又不丢准确率。
DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Models
- 按题目难度动态调整推理步数，难题多想，简单题少想。
- 平均减少30%以上推理令牌消耗，复杂题准确率不变。
- 适合需要高效推理的部署场景，尤其看重计算成本的应用。
近期慢思考推理模型在复杂推理任务中表现优异，但常出现过度思考（对简单问题生成冗余推理步骤），导致计算资源浪费。现有缓解策略统一减少推理令牌，可能损害需长推理的难题表现。本文提出难度自适应慢思考（DAST）框架，使模型能自主根据问题难度调整思维链（CoT）长度。首先提出令牌长度预算（TLB）度量难度，再通过预算感知奖励塑造与预算偏好优化实现DAST。该方法对简单任务惩罚过长回答，对复杂任务激励充分推理。在多种数据集与模型规模上实验表明，DAST有效缓解过度思考（平均减少30%以上令牌使用），同时保持复杂问题的推理准确率。代码与模型已开源。
原文摘要 · Abstract (English)
Recent advancements in slow thinking reasoning models have shown exceptional performance in complex reasoning tasks. However, these models often exhibit overthinking (generating redundant reasoning steps for simple problems), leading to excessive computational resource usage. While current mitigation strategies uniformly reduce reasoning tokens, they risk degrading performance on challenging tasks that require extended reasoning. This paper introduces Difficulty-Adaptive Slow Thinking (DAST), a novel framework that enables models to autonomously adjust the length of Chain-of-Thought (CoT) based on problem difficulty. We first propose a Token Length Budget (TLB) metric to quantify difficulty, then leverage budget-aware reward shaping and budget preference optimization to implement DAST. DAST penalizes overlong responses for simple tasks while incentivizing sufficient reasoning for complex problems. Experiments on diverse datasets and model scales demonstrate that DAST effectively mitigates overthinking (reducing token usage by over 30\% on average) while preserving reasoning accuracy on complex problems. Our codes and models are available at https://github.com/AnonymousUser0520/AnonymousRepo01.
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。