Few Tokens Matter: Entropy Guided Attacks on Vision-Language Models

Published in arXiv preprint, 2025