自我批评与精炼用于忠实的自然语言解释
Self-Critique and Refinement for Faithful Natural Language Explanations
AI总结：
针对自然语言解释不忠实问题，提出SR-NLE框架，通过自我批评与精炼（含特征归因反馈）迭代改进解释，在三个数据集和四个LLM上将平均不忠实率从54.81%降至36.02%。
摘要：
随着大型语言模型（LLMs）的快速发展，自然语言解释（NLEs）在理解模型预测方面变得越来越重要。然而，这些解释往往无法忠实地代表模型的实际推理过程。尽管现有工作已证明LLMs能够对各种任务进行自我批评并精炼其初始输出，但这一能力在提高解释忠实度方面仍未得到探索。为弥补这一空白，我们引入了自然语言解释的自我批评与精炼（SR-NLE）框架，该框架使模型能够通过迭代的批评与精炼过程，在无需外部监督的情况下，提高其自身解释（特别是事后NLEs）的忠实度。我们的框架利用不同的反馈机制来指导精炼过程，包括自然语言自我反馈，以及一种基于特征归因的新颖反馈方法，该方法突出重要的输入词。我们在三个数据集和四个最先进的LLMs上进行的实验表明，SR-NLE显著降低了不忠实率，我们最佳方法的平均不忠实率为36.02%，而基线为54.81%，绝对降低了18.79%。这些发现表明，所研究的LLMs确实能够精炼其解释，以更好地反映其实际推理过程，仅需通过反馈提供适当指导，无需额外训练或微调。
英文摘要：
With the rapid development of Large Language Models (LLMs), Natural Language Explanations (NLEs) have become increasingly important for understanding model predictions. However, these explanations often fail to faithfully represent the model's actual reasoning process. While existing work has demonstrated that LLMs can self-critique and refine their initial outputs for various tasks, this capability remains unexplored for improving explanation faithfulness. To address this gap, we introduce Self-critique and Refinement for Natural Language Explanations (SR-NLE), a framework that enables models to improve the faithfulness of their own explanations -- specifically, post-hoc NLEs -- through an iterative critique and refinement process without external supervision. Our framework leverages different feedback mechanisms to guide the refinement process, including natural language self-feedback and, notably, a novel feedback approach based on feature attribution that highlights important input words. Our experiments across three datasets and four state-of-the-art LLMs demonstrate that SR-NLE significantly reduces unfaithfulness rates, with our best method achieving an average unfaithfulness rate of 36.02%, compared to 54.81% for baseline -- an absolute reduction of 18.79%. These findings reveal that the investigated LLMs can indeed refine their explanations to better reflect their actual reasoning process, requiring only appropriate guidance through feedback without additional training or fine-tuning.
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。
