AI RESEARCH
Exposing Vulnerabilities in Visible-Infrared VLMs: A Unified Geometric Adversarial Framework with Cross-Task Transferability
arXiv CS.CV
•
ArXi:2605.22273v1 Announce Type: new Vision-language models (VLMs) have achieved strong performance across diverse multimodal tasks, but their adversarial robustness in visible-infrared (VIS-IR) scenarios remains underexplored. This gap is critical because VIS-IR sensing is widely used in real-world perception systems to reliable understanding under challenging imaging conditions. To address this cross-modal threat setting, we propose CFGPatch, a curved-edge fractal geometric adversarial patch framework for attacking VIS-IR VLMs.