GreenWAM: A World Foundation Model for Robotics

What happened: Researchers present GreenWAM, a world foundation model for robotics that serves as both an action-conditioned world model and a policy across 20 robot families, from bimanual arms to dexterous humanoids. Key numbers: 20 robot families; top-tier trajectory, depth and 3D accuracy on Why it matters: The authors report it outperforms every compared world-action model and video generator, and on a real humanoid it succeeds more often than policies adapted from π0.5 and Cosmos 3 Edge. Controlled experiments show robot specialization improves world prediction and control together.

Researchers present GreenWAM, a world foundation model for robotics that serves as both an action-conditioned world model and a policy across 20 robot families, from bimanual arms to dexterous humanoids.

The authors report it outperforms every compared world-action model and video generator, and on a real humanoid it succeeds more often than policies adapted from π0.5 and Cosmos 3 Edge. Controlled experiments show robot specialization improves world prediction and control together.