Why Should Adversarial Perturbations be Imperceptible? Rethink the Research Paradigm in Adversarial NLP

Open in new window