In and Out-of-Domain Text Adversarial Robustness via Label Smoothing
This addresses adversarial robustness for NLP practitioners, but it is incremental as it applies an existing regularization method to a new context.
The paper tackles the vulnerability of NLP models to adversarial attacks by studying label smoothing strategies, finding that they significantly improve robustness in models like BERT against various attacks, with analysis showing reduced over-confident errors.
Recently it has been shown that state-of-the-art NLP models are vulnerable to adversarial attacks, where the predictions of a model can be drastically altered by slight modifications to the input (such as synonym substitutions). While several defense techniques have been proposed, and adapted, to the discrete nature of text adversarial attacks, the benefits of general-purpose regularization methods such as label smoothing for language models, have not been studied. In this paper, we study the adversarial robustness provided by various label smoothing strategies in foundational models for diverse NLP tasks in both in-domain and out-of-domain settings. Our experiments show that label smoothing significantly improves adversarial robustness in pre-trained models like BERT, against various popular attacks. We also analyze the relationship between prediction confidence and robustness, showing that label smoothing reduces over-confident errors on adversarial examples.