UniAction: unified action conditioning lets one world model handle mixed action types
ZeYanjie · x · 2026-10-27
Researchers introduced UniAction, a unified action-conditioning framework for world models. Existing action-conditioned world models are highly heterogeneous—object forces, camera movements, hand motions, robot arms—each typically built with its own conditioning design. UniAction drives a single world model with all these action types, even mixing them within one generation, and jointly trained it outperforms the action-specific experts. Led by Bowen Xue; a detailed thread to follow.
More from Embodied
- Mundo turns a few room photos into an explorable, simulatable 3D world — at 1.5 FPS — mark_k · 2026-10-27
- Zurich Physical AI Hackathon puts attendees in a room with $1.5M of humanoids — ___Mufasaa · 2026-10-27
- Apple responds after iPhone Duo inner displays crack along the hinge — mobileraj · 2026-10-27
- VLA models encode task concepts as curved multi-dimensional geometry, 19-author study finds — ninamiolane · 2026-10-27
- Munro Live Posts Full Cybercab Engineering Deep-Dive Video — yunta_tsai · 2026-10-27
- Tesla's Robotaxi Fleet Keeps Growing in Austin — yunta_tsai · 2026-10-27