Wenbo Yu

~Wenbo_Yu1

One Perturbation is Enough: On Generating Universal Adversarial Perturbations against Vision-Language Pre-training Models

MIBench: A Comprehensive Benchmark for Model Inversion Attack and Defense