链式推理真的不可解释性吗？链式推理可以在不提示 verbalization 的情况下保持忠实
Is Chain-of-Thought Really Not Explainability? Chain-of-Thought Can Be Faithful without Hint Verbalization
AI总结：
本文挑战了链式推理的忠实性定义，指出基于提示注入的偏差特征指标过于狭隘，通过新指标显示推理预算影响提示 verbalization，并通过因果中介分析证明非 verbalized 提示可通过链式推理影响预测，呼吁更广泛的可解释性工具。
摘要：
近期工作使用偏差特征指标将链式推理（CoT）标记为不可信，如果其忽略了影响预测的提示注入提示。我们争论该指标采用了一种狭义的忠实性概念，并将不可信与不完整混淆，这种压缩是将分布式Transformer计算转化为线性自然语言叙述所需的。在多跳推理任务中，许多被偏差特征标记为不可信的CoT被其他指标判断为忠实，某些模型超过50%。通过新的faithful@k指标，我们显示更大的推理时间预算显著增加提示verbalization（某些设置中高达90%），表明明显的不可信可能源于紧缩的token限制。通过因果中介分析，我们进一步证明即使非verbalized提示也可以通过CoT因果中介预测变化。因此，我们呼吁不要仅依赖基于提示的评估，并倡导更广泛的可解释性工具，包括因果中介和破坏性度量。我们不声称所有CoT都是忠实的，只说明缺乏提示词本身并不证明不可信。
英文摘要：
Recent work, using the Biasing Features metric, labels a CoT as unfaithful if it omits a prompt-injected hint that affected the prediction. We argue this metric adopts a narrow notion of faithfulness and confuses unfaithfulness with incompleteness, the lossy compression needed to turn distributed transformer computation into a linear natural language narrative. On multi-hop reasoning tasks with instruct-tuned and reasoning models, many CoTs flagged as unfaithful by Biasing Features are judged faithful by other metrics, exceeding 50% in some models. With a new faithful@k metric, we show that larger inference-time budgets greatly increase hint verbalization (up to 90% in some settings), suggesting much apparent unfaithfulness is due to tight token limits. Using Causal Mediation Analysis, we further show that even non-verbalized hints can causally mediate prediction changes through the CoT. We therefore caution against relying solely on hint-based evaluations and advocate a broader interpretability toolkit, including causal mediation and corruption-based metrics. We do not claim all CoTs are faithful, only that the absence of hint words alone does not prove unfaithfulness.
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。
