AI RESEARCH
Investigating Adversarial Robustness of Multi-modal Large Language Models
arXiv CS.CV
•
ArXi:2606.03713v1 Announce Type: new Multi-modal Large Language Models (MLLMs) achieve strong performance on vision-language tasks, but incorporating visual inputs through a vision encoder (e.g., CLIP) substantially expands the attack surface, making these models vulnerable to visual adversarial perturbations. Prior defenses typically preserve compatibility with pretrained MLLMs by enforcing strict alignment to CLIP's original embedding space during adversarial fine-tuning; while practical, this constraint fundamentally limits achievable robustness.