论文标题
选择偏见引起大语言模型中的虚假相关性
Selection Bias Induced Spurious Correlations in Large Language Models
论文作者
论文摘要
在这项工作中，我们显示了大型语言模型（LLMS）如何在由于数据集选择偏差而导致的其他无条件自变量之间学习统计依赖性。为了证明效果，我们开发了一项掩盖的性别任务，可以应用于伯特家庭模型，以揭示预测的性别代词与各种看似性别中性变量（如日期和位置）之间的虚假相关性，例如预先训练（未经修改的）Bert和Roberta大型模型。最后，我们提供了一个在线演示，邀请读者进一步实验。
In this work we show how large language models (LLMs) can learn statistical dependencies between otherwise unconditionally independent variables due to dataset selection bias. To demonstrate the effect, we developed a masked gender task that can be applied to BERT-family models to reveal spurious correlations between predicted gender pronouns and a variety of seemingly gender-neutral variables like date and location, on pre-trained (unmodified) BERT and RoBERTa large models. Finally, we provide an online demo, inviting readers to experiment further.
