word perturbation

adversarial attack (2) model robustness (2) adversarial robustness (2) nlp model (1) word substitution perturbation (1) deep nlp model (1) randomized perturbation defense (1) character perturbation (1) robustness enhancement (1) model confidence (1)