2026
DDGA: Dirichlet Distributional Gradient Aggregation for Transferable Vision-Language Adversarial Attacks
ICML 2026poster
Vision-Language Models (VLMs) achieve remarkable performance on multimodal tasks but remain highly vulnerable to adversarial examples, making transferable attacks essential for realistic robustness evaluation. Recent Adversarial Evolution Triangle (AET) methods improve transferability by interpolati…