QWM: one world model zero-shot transfers locomotion across quadruped robots at CoRL
breadli428 · x · 2026-10-10
ETH Zurich's leggedrobotics group presents the Quadrupedal World Model (QWM) at CoRL 2026: a single morphology-conditioned generative dynamics model in which policies are trained entirely in imagination. By conditioning on scale-invariant physical features, QWM learns across heterogeneous quadrupeds and transfers zero-shot to unseen morphologies — no fine-tuning, adaptation, or warm-up — while a model-free policy given the same morphology info degrades out of cohort. Authors claim it's the first world model showing zero-shot cross-embodiment transfer within quadrupeds.
More from Embodied
- Robots now get up with ease: big upgrades in balance and control — cixliv · 2026-10-10
- Scoble at Tesla Gigafactory: Workers Wear VR-Like Rigs to Train Optimus, 20,000 Already There — Scobleizer · 2026-10-10
- SJTU's UVTA teaches robot dexterous manipulation from 1,000 human tactile demos per task — siyuanhuang95 · 2026-10-10
- SpaceX building industrial base to make up to 1,000 Starships a year, with Florida GigaBay 11x bigger than current Megabay — XFreeze · 2026-10-10
- 20-minute explainer breaks down how Tesla trains FSD: 8 cameras, 36fps, 2B signals to 2 outputs — PTrubey · 2026-10-10
- Real Steel IRL: Humanoid Robots Noisy Boy and Raider-72 Box Each Other — zealcaiden · 2026-10-10