基于解释的微调使模型对虚假线索更具鲁棒性
Explanation-based Finetuning Makes Models More Robust to Spurious Cues
AI总结：
本文提出让模型在预测答案外生成自由文本解释的微调方法，显著提升 GPT-3 等 LLM 对虚假线索的分布外鲁棒性，并验证其跨模型规模及自生成解释场景的有效性。
摘要：
大型语言模型（LLM）非常强大，以至于它们有时会学习到标签与任务无关特征之间的相关性，从而导致在分布外数据上的泛化能力较差。我们提出基于解释的微调，作为缓解 LLM 对虚假相关性依赖的一种通用方法。与标准微调中模型仅根据输入预测答案不同，我们对模型进行微调，使其额外生成支持其答案的自由文本解释。为评估我们的方法，我们在包含不同类型虚假线索的人工构建训练集上微调模型，并在没有这些线索的测试集上进行测试。与标准微调相比，在四项分类任务上，就准确率下降而言，我们的方法使 GPT-3（davinci）对虚假线索的鲁棒性显著增强：ComVE（+1.2）、CREAK（+9.1）、e-SNLI（+15.4）和 SBIC（+6.5）。这种效果可泛化到多个模型系列和规模，并且较大模型获得的增益更大。最后，我们的方法在使用模型生成的解释时也表现良好，这表明它适用于更多没有人工撰写解释的数据集。
英文摘要：
Large Language Models (LLMs) are so powerful that they sometimes learn correlations between labels and features that are irrelevant to the task, leading to poor generalization on out-of-distribution data. We propose explanation-based finetuning as a general approach to mitigate LLMs' reliance on spurious correlations. Unlike standard finetuning where the model only predicts the answer given the input, we finetune the model to additionally generate a free-text explanation supporting its answer. To evaluate our method, we finetune the model on artificially constructed training sets containing different types of spurious cues, and test it on a test set without these cues. Compared to standard finetuning, our method makes GPT-3 (davinci) remarkably more robust against spurious cues in terms of accuracy drop across four classification tasks: ComVE (+1.2), CREAK (+9.1), e-SNLI (+15.4), and SBIC (+6.5). The efficacy generalizes across multiple model families and scales, with greater gains for larger models. Finally, our method also works well with explanations generated by the model, implying its applicability to more datasets without human-written explanations.
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。
