Visual Adversarial Examples Jailbreak Aligned Large Language Models

Xiangyu Qi, Kaixuan Huang, Ashwinee Panda, Peter Henderson, Mengdi Wang, Prateek Mittal

Abstract

Shows adversarial images can jailbreak multimodal LLMs that are robust to text-only attacks, bypassing alignment through the visual channel.

Categories

Tags

Framework Mappings

Cite This Resource

@article{llmsec202500018,
  title = {Visual Adversarial Examples Jailbreak Aligned Large Language Models},
  author = {Xiangyu Qi and Kaixuan Huang and Ashwinee Panda and Peter Henderson and Mengdi Wang and Prateek Mittal},
  year = {2024},
  journal = {AAAI 2024},
  url = {https://arxiv.org/abs/2306.13213},
}