一作ACL 2026Poster· Natural Language Processing
Seeing No Evil: Blinding Large Vision-Language Models to Safety Instructions via Adversarial Attention Hijacking
Jingru Li · Wei Ren · Tianqing Zhu
摘要
Large Vision-Language Models (LVLMs) rely on attention-based retrieval of safety instructions to maintain alignment during generation. Existing attacks typically optimize image perturbations to maximize harmful output likelihood, but suffer from slow convergence due to gradient conflict between a...