BAE: BERT-based Adversarial Examples for Text Classification

Modern text classification models are susceptible to adversarial examples, perturb