Robot Models Learn Depth and Shape to Grasp Better

Original: XGenAct: Geometry-Enhanced World Action Models through Cross-Task Generation

Source: arxiv.org ↗