思维惯性：大语言模型在被要求停止思考时仍会继续思考

Thinking Inertia: LLMs Keep Thinking When Told Not To

October 8, 2026

作者: Dianqiao Lei, Kevin Qinghong Lin, Pan Lu, Philip Torr, James Zou

cs.AI

摘要

大型语言模型（LLM）越来越多地自带明确的“思考模式”，但与之相对的“不思考”却受到的关注少得多。我们从两个维度研究 LLM 的不思考行为。a. 如何度量不思考？先前工作通常通过代理方式来定义不思考，例如禁用思考模式或不存在长轨迹。这些代理方式并不可靠：被禁用的思考模式仍可能输出推理，而长轨迹可能包含填充内容而非真正的推理。我们改为将每个回答归一化为答案前轨迹和最终答案，并从三个层面进行评估：（i）空思考率（Empty-Thinking Rate），用于衡量严格仅答案合规性；（ii）指令感知的问题—答案前轨迹相关性（instruction-aware Question-Pre-answer Relevance），用于衡量问题与答案前轨迹之间的相似度；（iii）LLM 作为评判者的显式推理率（LLM-as-judge Explicit Inference Rate），用于衡量可见的显式推理。这些指标共同区分仅答案输出、相关但非推理的文本以及显式推理。b. 不思考在不同任务和模型间如何变化？我们针对布尔题、多项选择题和开放性问题，在六个 LLM 上评估了六种提示干预。我们发现，显式的不思考控制无法可靠地消除可见推理。模型反而表现出“思考惯性”：即使在严格控制下，显式推理仍持续存在，并随着答案空间开放而更加普遍。在布尔题和多项选择题任务上，准确率保持稳定；而开放性任务则揭示了仅答案合规性与任务准确率之间的权衡。将相同问题跨答案空间改写后表明，提供候选答案会使仅答案式回答更容易生成。这些发现确立了不思考是一种非平凡能力：不能仅凭模型设置或指令就假定能够停止显式推理，它值得与推理能力一道接受系统评估。

English

Large Language Models (LLMs) increasingly ship with explicit "thinking modes", yet their counterpart, "no-thinking", has received far less attention. We study LLMs' no-thinking behavior along two axes. a. How to measure no-thinking? Prior work typically defines no-thinking through proxies such as a disabled thinking mode or the absence of long traces. These proxies are unreliable: disabled thinking modes may still emit reasoning, while long traces may contain filler rather than genuine inference. We instead normalize each response into a pre-answer trace and final answer, and evaluate it at three levels: (i) Empty-Thinking Rate for strict answer-only compliance; (ii) instruction-aware Question-Pre-answer Relevance for similarity between the question and pre-answer trace; and (iii) LLM-as-judge Explicit Inference Rate for visible explicit inference. Together, these metrics distinguish answer-only output, relevant but non-inferential text, and explicit inference. b. How does no-thinking vary across tasks and models? We evaluate six prompting interventions on six LLMs across Boolean, multiple-choice, and open-ended questions. We find that explicit no-think controls cannot reliably eliminate visible inference. Models instead exhibit "Thinking Inertia": explicit inference persists even under strict controls and becomes more prevalent as the answer space opens. Accuracy remains stable on Boolean and multiple-choice tasks, whereas open-ended tasks reveal a trade-off between answer-only compliance and task accuracy. Rewriting the same questions across answer spaces shows that supplying candidate answers makes answer-only responses easier to produce. These findings establish no-thinking as a non-trivial capability: stopping explicit reasoning cannot be assumed from model settings or instructions alone and deserves systematic evaluation alongside reasoning ability.