Avi: Action from Volumetric Inference

Harris Song, Long Le

We propose Avi, a novel 3D Vision-Language-Action (VLA) architecture that reframes robotic action generation as a problem of 3D perception and spatial reasoning, rather than low-le…