Four visual jailbreak techniques on VLMs achieve substantially higher success rates than equivalent text attacks, exposing a cross-modality gap in safety alignment.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Jailbreaking Vision-Language Models Through the Visual Modality
Four visual jailbreak techniques on VLMs achieve substantially higher success rates than equivalent text attacks, exposing a cross-modality gap in safety alignment.