Few-ShotAdversarialPromptLearningon Vision-LanguageModels

In this paper, to address these issues, we propose a few-shot adversarial prompt framework where adapting input sequences with limited data makes significant adversarial robustness improvement.