AI RESEARCH
One Perturbation, Two Failure Modes: Probing VLM Safety via Embedding-Guided Typographic Perturbations
arXiv CS.CV
•
ArXi:2604.25102v1 Announce Type: new Typographic prompt injection exploits vision language models' (VLMs) ability to read text rendered in images, posing a growing threat as VLMs power autonomous agents. Prior work typically focus on maximizing attack success rate (ASR) but does not explain \emph{why} certain renderings bypass safety alignment. We make two contributions.