模型规模对通过 LIME 实现的 LLM 事后可解释性的影响
The Effect of Model Size on LLM Post-hoc Explainability via LIME
AI总结：
本文通过 LIME 分析四种规模 DeBERTaV3 在 NLI 与零样本分类上的事后解释，发现规模提升改善性能却不提升解释合理性，并揭示忠实度指标局限。
摘要：
大型语言模型（LLM）正变得越来越大，以提升性能。然而，人们对这一趋势如何影响可解释性知之甚少。本工作在自然语言推理（NLI）和零样本分类（ZSC）任务上，探究了四种不同规模 DeBERTaV3 模型的 LIME 解释。我们依据解释对模型内部决策过程的忠实度及其合理性（即与人类解释的一致性）来评估这些解释。关键发现是，尽管模型性能有所提升，但模型规模增大与合理性并不相关，这表明随着模型规模增加，LIME 解释与模型内部过程之间存在错位。我们的结果进一步表明，在 NLI 情境下，忠实度指标存在局限性。
英文摘要：
Large language models (LLMs) are becoming bigger to boost performance. However, little is known about how explainability is affected by this trend. This work explores LIME explanations for DeBERTaV3 models of four different sizes on natural language inference (NLI) and zero-shot classification (ZSC) tasks. We evaluate the explanations based on their faithfulness to the models' internal decision processes and their plausibility, i.e. their agreement with human explanations. The key finding is that increased model size does not correlate with plausibility despite improved model performance, suggesting a misalignment between the LIME explanations and the models' internal processes as model size increases. Our results further suggest limitations regarding faithfulness metrics in NLI contexts.
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。
