SJTU's UVTA teaches robot dexterous manipulation from 1,000 human tactile demos per task
siyuanhuang95 · x · 2026-10-10
Researchers at Shanghai Jiao Tong University introduce UVTA (Unified Visual-Tactile-Action Modeling), arguing human touch — not just human motion — is the scalable path to robot dexterity.
- System & data: a passive 22-DoF exoskeleton syncs hand motion, fingertip pressure, wrist pose and wrist-mounted RGB; 1,000 human demos plus 150 robot demos per task (wrist RGB, 20-D touch, 31-D action, 30 Hz). Dataset on Hugging Face.
- Model: a 384-D visual token and 20-D tactile token fuse into a 404-D interaction representation, trained via a diffusion action head producing 16-step chunks of wrist and finger-joint actions plus future-tactile prediction.
- Paper, code and project page are public.
More from Embodied
- Scoble at Tesla Gigafactory: Workers Wear VR-Like Rigs to Train Optimus, 20,000 Already There — Scobleizer · 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
- Dyna launches Taku semi-humanoid robot to remove the human babysitting bottleneck — JasonMa2020 · 2026-10-10
- First-ever humanoid deathmatch was absolutely wild — cixliv · 2026-10-10