Instruction-tuned large language models (LLMs) excel at many tasks, and will even provide explanations for their behavior. Since these models are directly accessible to the public, there is a risk that convincing and wrong explanations can lead to unsupported confidence in LLMs. Therefore, interpretability-faithfulness of self-explanations is an important consideration for AI Safety. Assessing the interpretability-faithfulness of these explanations, termed self-explanations, is challenging as the models are too complex for humans to annotate what is a correct explanation. To address this, we propose employing self-consistency checks as a measure of faithfulness. For example, if an LLM says a set of words is important for making a prediction, then it should not be able to make the same prediction without these words. While self-consistency checks are a common approach to faithfulness, they have not previously been applied to LLM's self-explanations. We apply self-consistency checks to three types of self-explanations: counterfactuals, importance measures, and redactions. Our work demonstrate that faithfulness is both task and model dependent, e.g., for sentiment classification, counterfactual explanations are more faithful for Llama2, importance measures for Mistral, and redaction for Falcon 40B. Finally, our findings are robust to prompt-variations.

翻译：指令微调的大型语言模型（LLMs）在许多任务上表现出色，甚至能为其行为提供解释。由于这些模型可直接被公众访问，存在一种风险：令人信服但错误的解释可能导致对LLMs产生无依据的信任。因此，自我解释的可解释性-忠实度是人工智能安全的重要考量。评估这些被称为自我解释的忠实度颇具挑战，因为模型过于复杂，人类无法标注何为正确的解释。为解决此问题，我们提出采用自洽性检验作为忠实度的度量标准。例如，若一个LLM声称某些词对预测至关重要，那么缺少这些词时，它应无法做出相同预测。尽管自洽性检验是评估忠实度的常见方法，但此前尚未应用于LLM的自我解释。我们将自洽性检验应用于三类自我解释：反事实解释、重要性度量与文本删节。我们的工作表明，忠实度既依赖于任务也依赖于模型，例如在情感分类任务中，对Llama2而言反事实解释更忠实，对Mistral则是重要性度量，而对Falcon 40B则是文本删节。最后，我们的发现对提示变化具有稳健性。