2022
Distinguishing Non-natural from Natural Adversarial Samples for More Robust Pre-trained Language Model
ACL 2022findings
Recently, the problem of robustness of pre-trained language models (PrLMs) has received increasing research interest. Latest studies on adversarial attacks achieve high attack success rates against PrLMs, claiming that PrLMs are not robust. However, we find that the adversarial samples that PrLMs fa…