When Alignment Fails: Multimodal Adversarial Attacks on Vision-Language-Action Models

作者：Yuping Yan, Yuhan Xie, Yixin Zhang, Lingjuan Lyu, Handing Wang, Yaochu Jin · 单位：TGAI Lab, School of Engineering, Westlake University, Pennsylvania State University Sony Research, Sony Xidian University · 会议/期刊：arXiv · 日期：2025-11-20 · 来源：Low-Level Learning-Based Action Modelling / Input Modelling / 2D Vision Language Action Models with Robustness / Attack