FlowHijack: A Dynamics-Aware Backdoor Attack on Flow-Matching Vision-Language-Action Models

Xinyuan An, Tao Luo, Gengyun Peng +3

Vision-Language-Action (VLA) models are emerging as a cornerstone for robotics, with flow-matching policies like $\ddot{I}{€}_0$ showing great promise in generating smooth, continuous actio…