Loading...
標記為「vision-model」的 1 篇文章
Attack vision-language models through adversarial images with embedded text, perturbations, and visual trojans.