Generative Data Augmentation using LLMs improves Distributional Robustness in Question Answering

TL;DR

This paper investigates how data generated by large language models can be used to improve the robustness of question answering systems against natural distribution shifts, focusing on domain generalization.

Contribution

It introduces a two-step data augmentation method using generative models to enhance QA datasets and demonstrates improved robustness under distribution shifts.

Findings

Generated data improves QA model robustness

Augmentation leads to better domain generalization

Experiments on multiple datasets confirm effectiveness

Abstract

Robustness in Natural Language Processing continues to be a pertinent issue, where state of the art models under-perform under naturally shifted distributions. In the context of Question Answering, work on domain adaptation methods continues to be a growing body of research. However, very little attention has been given to the notion of domain generalization under natural distribution shifts, where the target domain is unknown. With drastic improvements in the quality and access to generative models, we answer the question: How do generated datasets influence the performance of QA models under natural distribution shifts? We perform experiments on 4 different datasets under varying amounts of distribution shift, and analyze how "in-the-wild" generation can help achieve domain generalization. We take a two-step generation approach, generating both contexts and QA pairs to augment existing datasets. Through our experiments, we demonstrate how augmenting reading comprehension datasets with generated data leads to better robustness towards natural distribution shifts.