The Impacts of Unanswerable Questions on the Robustness of Machine Reading Comprehension Models

TL;DR

This paper investigates whether training on unanswerable questions in SQuAD 2.0 enhances the robustness of machine reading comprehension models against adversarial attacks, revealing hidden robustness and out-of-domain generalization.

Contribution

It demonstrates that training on SQuAD 2.0 introduces hidden robustness in models and presents a new adversarial attack to identify learned artifacts.

Findings

Models trained on SQuAD 2.0 show hidden robustness.

Robustness extends to out-of-domain datasets.

A new adversarial attack reveals artifacts learned from SQuAD 2.0.

Abstract

Pretrained language models have achieved super-human performances on many Machine Reading Comprehension (MRC) benchmarks. Nevertheless, their relative inability to defend against adversarial attacks has spurred skepticism about their natural language understanding. In this paper, we ask whether training with unanswerable questions in SQuAD 2.0 can help improve the robustness of MRC models against adversarial attacks. To explore that question, we fine-tune three state-of-the-art language models on either SQuAD 1.1 or SQuAD 2.0 and then evaluate their robustness under adversarial attacks. Our experiments reveal that current models fine-tuned on SQuAD 2.0 do not initially appear to be any more robust than ones fine-tuned on SQuAD 1.1, yet they reveal a measure of hidden robustness that can be leveraged to realize actual performance gains. Furthermore, we find that the robustness of models fine-tuned on SQuAD 2.0 extends to additional out-of-domain datasets. Finally, we introduce a new adversarial attack to reveal artifacts of SQuAD 2.0 that current MRC models are learning.