Ropedia introduces W0, a world model for human and robot interactions
A post says W0 was trained on 10,000 hours of human video and 3,000 hours of robot video.
TLDR
A post sharing Ropedia’s announcement says W0 predicts what humans and robots will see as they move and interact with the world. It claims W0 outperformed baselines across six human and robot benchmarks and supports real-time, minute-long VR interaction. On a real humanoid, the post says model-based planning improved task progress to 55%, compared with 45% for the best single policy.
Combined views
2K
1 Source, first seen ago