#artificial intelligence
Mar 2026
Understanding and Defending VLM Jailbreaks via Jailbreak-Related Representation Shift
It is observed that VLMs can clearly distinguish benign inputs from harmful ones in their representation space, and even among harmful inputs, jailbreak samples form a distinct internal state that is separable from refusal samples.
Zhi-Hua Wei, Qiang Li, J. Ruan et al.
· arXiv.org · 6 citations