Author name: Ravikumar Balakrishnan, Sanket Mendapara, Ankit Garg

Reading Between the Pixels: Linking Text-Image Embedding Alignment to Typographic Attack Success on Vision-Language Models

Ravikumar Balakrishnan, Sanket Mendapara, Ankit Garg / April 16, 2026

arXiv:2604.12371v2 Announce Type: replace
Abstract: We study typographic prompt injection attacks on vision-language models (VLMs), where adversarial text is rendered as images to bypass safety mechanisms, posing a growing threat as VLMs serve as the …

cs.CV

Reading Between the Pixels: Linking Text-Image Embedding Alignment to Typographic Attack Success on Vision-Language Models

Ravikumar Balakrishnan, Sanket Mendapara, Ankit Garg / April 15, 2026

arXiv:2604.12371v1 Announce Type: new
Abstract: We study typographic prompt injection attacks on vision-language models (VLMs), where adversarial text is rendered as images to bypass safety mechanisms, posing a growing threat as VLMs serve as the perc…