Hao Lin

Few-shot Text Adversarial Attack for Black-box Multi-task Learning

No Access, No Safety: Free Lunch Adversarial Attacks on Black-box NLP Models