1 paper
Xiaomeng Wang, Zhengyu Zhao, Martha Larson
Large Vision-Language Models (LVLMs) are susceptible to typographic attacks, which are misclassifications caused by an attack text that is added to an image. In this paper, we intr…