Armfull demonstrates a perception-to-action pipeline in which a Vision-Language-Action (VLA) policy understands a natural-language task instruction, reasons over simulated camera observations, and coordinates two SO-101 robotic arms in MuJoCo to set a dinner table — including a real hand-off where Arm A pours water into a mug that Arm B is holding.
.png&w=256&q=75)
.png&w=256&q=75)