让大模型自动判断何时该多想，节省计算又不丢准确率。
When to Continue Thinking: Adaptive Thinking Mode Switching for Efficient Reasoning
- 根据题目难易动态调节思考长度，避免无意义长推理。
- 在多个模型上减少32.5%计算量，准确率损失仅1.2%。
- 适合追求高效推理、安全可控的部署场景。
大型推理模型（LRMs）通过长推理链实现优异表现，但常因冗余推理导致过高计算开销，尤其在简单任务上。本文系统量化了LRMs在长思考与无思考模式下的性能上限，发现模型在生成答案时存在“内部自恢复机制”，可隐式补全推理过程。基于此，提出自适应自恢复推理（ASRR）框架，抑制不必要的推理并实现隐式修复。通过引入精度感知的长度奖励调节，ASRR根据问题难度自适应分配推理资源，在多个基准和模型上实验显示，相比GRPO，ASRR在1.5B和7B模型上分别减少32.5%和25.7%的推理预算，准确率损失仅为1.2%和0.6% pass@1，且在安全基准上显著提升无害率（最高+21.7%）。结果表明ASRR能有效实现高效、自适应、更安全的推理。
原文摘要 · Abstract (English)
Large reasoning models (LRMs) achieve remarkable performance via long reasoning chains, but often incur excessive computational overhead due to redundant reasoning, especially on simple tasks. In this work, we systematically quantify the upper bounds of LRMs under both Long-Thinking and No-Thinking modes, and uncover the phenomenon of "Internal Self-Recovery Mechanism" where models implicitly supplement reasoning during answer generation. Building on this insight, we propose Adaptive Self-Recovery Reasoning (ASRR), a framework that suppresses unnecessary reasoning and enables implicit recovery. By introducing accuracy-aware length reward regulation, ASRR adaptively allocates reasoning effort according to problem difficulty, achieving high efficiency with negligible performance sacrifice. Experiments across multiple benchmarks and models show that, compared with GRPO, ASRR reduces reasoning budget by up to 32.5% (1.5B) and 25.7% (7B) with minimal accuracy loss (1.2% and 0.6% pass@1), and significantly boosts harmless rates on safety benchmarks (up to +21.7%). Our results highlight the potential of ASRR for enabling efficient, adaptive, and safer reasoning in LRMs.
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。