Universal adversarial attacks cause output perturbation 90 times more often than precise target injection in VLMs, with only 2 verbatim successes out of 6615 tests.
AnyAttack: Towards large-scale self-supervised adversarial attacks on vision-language models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
VisInject: Disruption != Injection -- A Dual-Dimension Evaluation of Universal Adversarial Attacks on Vision-Language Models
Universal adversarial attacks cause output perturbation 90 times more often than precise target injection in VLMs, with only 2 verbatim successes out of 6615 tests.