通过概率差值自动停止推理，减少无效思考并提升效率。
ThinkBrake: Efficient Reasoning via Log-Probability Margin Guided Decoding
- 在句末检测最高延续词与结束符的概率差，差值过小时停止推理。
- 在数学和科学问答任务中，减少30%的思考token，准确率不降反升。
- 无需训练，可直接部署，适合追求高效推理的实用场景。
大型推理模型（LRMs）在链式思维（CoT）推理中消耗大量推理计算，虽提升数学、科学问答和工具使用性能，但导致过度推理：模型常提前得出正确中间答案，继续推理后反而覆盖为错误结果。我们首次证明，若在每个句子边界注入</think>并事后选择最佳停止点，可使平均准确率提升8%，同时将思考token减少72%，揭示了严重的过度推理问题。受此启发，我们提出ThinkBrake，它在句末监控最高延续候选词与</think> token之间的对数概率差，当差值缩小即停止推理。ThinkBrake无需训练，在数学、科学问答和工具使用基准上均实现良好的精度-效率权衡，思考token使用量最多减少30%。此外，我们提供理论分析，表明ThinkBrake等价于测试时重对齐，并对</think> token给予奖励加分。
原文摘要 · Abstract (English)
Large Reasoning Models (LRMs) allocate substantial inference-time compute to Chain-of-Thought (CoT) reasoning, improving performance on mathematics, scientific QA, and tool usage. However, this introduces overthinking: LRMs often reach a correct intermediate solution, continue reasoning, and overwrite it with an incorrect answer. We first demonstrate that oracle stopping--where we inject </think> at every sentence boundary and select the best stopping point in hindsight--improves average accuracy by 8% while reducing thinking tokens by 72%, exposing substantial overthinking. Motivated by this finding, we propose ThinkBrake, which monitors the log-probability margin between the top continuation token and </think> at sentence boundaries, stopping reasoning when this margin narrows. ThinkBrake requires no training and achieves favorable accuracy-efficiency trade-offs across math, scientific QA, and tool usage benchmarks, reducing thinking token usage by up to 30%. Furthermore, we provide theoretical analysis showing that ThinkBrake is equivalent to test-time realignment with a reward bonus for the </think> token.
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。