Abstract
Large Language Models (LLMs) excel in text classification, but their complexity hinders interpretability, making it difficult to understand the reasoning behind their predictions. Explainable AI (XAI) methods like LIME and SHAP offer local explanations by identifying influential words, but they rely on computationally expensive perturbations. These methods typically generate thousands of perturbed sentences and perform inferences on each, incurring a substantial computational burden, especially with LLMs. To address this, we propose \underline{P}erturbation-free \underline{L}ocal \underline{Ex}planation (PLEX), a novel method that leverages the contextual embeddings extracted from the LLM and a ``Siamese network" style neural network trained to align with feature importance scores. This one-off training eliminates the need for subsequent perturbations, enabling efficient explanations for any new sentence. We demonstrate PLEX's effectiveness on four different classification tasks (sentiment, fake news, fake COVID-19 news and depression), showing more than 92\% agreement with LIME and SHAP. Our evaluation using a ``stress test" reveals that PLEX accurately identifies influential words, leading to a similar decline in classification accuracy as observed with LIME and SHAP when these words are removed. Notably, in some cases, PLEX demonstrates superior performance in capturing the impact of key features. PLEX dramatically accelerates explanation, reducing time and computational overhead by two and four orders of magnitude, respectively. This work offers a promising solution for explainable LLM-based text classification.
Abstract (translated)
大型语言模型（LLMs）在文本分类任务中表现出色，但它们的复杂性阻碍了可解释性，使得理解其预测背后的逻辑变得困难。像LIME和SHAP这样的可解释人工智能(XAI)方法通过识别有影响力的词汇来提供局部解释，但是这些方法依赖于计算成本高昂的数据扰动。通常情况下，为了生成成千上万个扰动的句子并对其进行推理，这种方法会产生巨大的计算负担，尤其是对于LLMs而言。 为了解决这个问题，我们提出了一种名为“无扰动局部解释”（PLEX）的新方法。PLEX利用从大型语言模型中提取的上下文嵌入和一种类似“Siamese网络”的神经网络来与特征重要性得分对齐。这种方法通过一次性的训练消除了后续需要进行扰动的需求，从而能够为任何新的句子提供高效的解释。 我们在四个不同的分类任务（情感分析、假新闻识别、新冠假新闻检测以及抑郁症检测）上展示了PLEX的有效性，并且PLEX在92%以上的测试中与LIME和SHAP达成一致。通过使用“压力测试”进行评估，PLEX准确地识别出有影响力的词汇，在这些单词被移除后导致分类准确性下降的情况类似于LIME和SHAP的表现。 值得注意的是，在某些情况下，PLEX展示了捕捉关键特征影响的优越性能。更重要的是，PLEX极大地加快了解释过程，将时间和计算开销分别减少了两个和四个数量级。这项工作为基于LLM的文本分类提供了可解释性的有希望的解决方案。
URL
https://arxiv.org/abs/2507.10596
