1 paper · 1 filter
Bryan E. Tuck, Rakesh M. Verma
Adversarial text attacks remain a persistent threat to transformer models, yet existing defenses are typically attack-specific or require costly model retraining, leaving a gap for…