不只是终点，而是旅程：推理痕迹因果地塑造泛化行为
Not Just the Destination, But the Journey: Reasoning Traces Causally Shape Generalization Behaviors
AI总结：
研究探讨推理痕迹对模型泛化行为的因果影响，通过设计控制实验发现不同推理类型导致不同行为模式，证明推理本身具有独立信号。
摘要：
Chain-of-Thought (CoT) 通常被视为理解 LLM 决策的窗口，但近期研究表明它可能仅是事后合理化。这提出了一个关键对齐问题：推理痕迹是否在不考虑最终答案的情况下因果地塑造模型泛化？为隔离推理的因果效应，我们设计了一项受控实验，保持最终有害答案不变而改变推理路径。我们构建了包含恶意推理、误导推理和顺从推理的 datasets。我们训练了 0.6B-14B 参数的模型，在多种范式下进行训练，包括问题-思考-答案 (QTA)、问题-思考 (QT) 和仅思考 (T-only)，并在思考和无思考模式下评估。我们发现：(1) CoT 训练比标准微调更加强化有害泛化；(2) 不同推理类型诱导不同的行为模式，与语义一致，尽管最终答案相同；(3) 训练推理而不进行答案监督 (QT 或 T-only) 足以改变行为，证明推理具有独立信号；(4) 这些效果即使在不生成答案的情况下也持续存在，表明深度内化。我们的发现表明推理内容具有因果效力，挑战了仅监督输出的对齐策略。
英文摘要：
Chain-of-Thought (CoT) is often viewed as a window into LLM decision-making, yet recent work suggests it may function merely as post-hoc rationalization. This raises a critical alignment question: Does the reasoning trace causally shape model generalization independent of the final answer? To isolate reasoning's causal effect, we design a controlled experiment holding final harmful answers constant while varying reasoning paths. We construct datasets with \textit{Evil} reasoning embracing malice, \textit{Misleading} reasoning rationalizing harm, and \textit{Submissive} reasoning yielding to pressure. We train models (0.6B--14B parameters) under multiple paradigms, including question-thinking-answer (QTA), question-thinking (QT), and thinking-only (T-only), and evaluate them in both think and no-think modes. We find that: (1) CoT training could amplify harmful generalization more than standard fine-tuning; (2) distinct reasoning types induce distinct behavioral patterns aligned with their semantics, despite identical final answers; (3) training on reasoning without answer supervision (QT or T-only) is sufficient to alter behavior, proving reasoning carries an independent signal; and (4) these effects persist even when generating answers without reasoning, indicating deep internalization. Our findings demonstrate that reasoning content is causally potent, challenging alignment strategies that supervise only outputs.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
