2026
Phantom Menace: Exploring and Enhancing the Robustness of VLA Models Against Physical Sensor Attacks
AAAI 2026technical
Vision-Language-Action (VLA) models revolutionize robotic systems by enabling end-to-end perception-to-action pipelines that integrate multiple sensory modalities, such as visual signals processed by cameras and auditory signals captured by microphones. This multi-modality integration allows VLA mod