Ψ-0.6: unified world model reasons in words, depth and motion streams, transfers to real robots
FaeriaManic · reddit · 2026-10-28
Researchers introduced Ψ-0.6, a unified vision-language world model that extends chain-of-thought beyond words into "streams of thought": words, depth, motion, camera pose, and imagined scenes, with RL learning when to reason through which representation.
- Surpasses leading open-weight models on complex spatial reasoning (MindCube, zero-shot)
- Beats other world models on FlowBench for object motion prediction; trained only on simulated rooms, it transfers to real robot tabletop manipulation
- One model acts as perceiver, VLM, world model, reasoner, and policy
Paper, website, and code are all public.
Related event: Stanford's Psi-0.6: unified world model reasons with streams of thought(3 posts)→
More from Embodied
- NVIDIA unveils ThinkWorld: robots learn motion intelligence from human videos via shared dynamics — CMHungSteven · 2026-10-28
- Houston Robotaxi called most 'fake' AV service map: 40 min to cross 2 blocks — reed · 2026-10-28
- Indie roboticist gets first anthropometry-based robot walking with agent-driven design loop — IanPritchard · 2026-10-28
- Reward AI's OM-1 Learns Manipulation Skills From Human Demonstrations for Cross-Body Transfer — CyberRobooo · 2026-10-28
- 4 Stanford freshmen build assistive walker robot in 36 hours, win grand prize at CalHacks 13.0 — Scobleizer · 2026-10-28
- Just Action Model: decoupling what separates VLA from world-action models — Yonghao Yu · 2026-10-28