RoboTok: Internet-Scale Data Engine Retrieves Human Videos for Dexterous Robot Learning
Rice-RobotPI-Lab · hf · 2026-09-05
Rice University's RoboPI Lab released RoboTok, an internet-scale data engine for dexterous manipulation learning. It builds a latent motion space from 3D hand trajectories to retrieve relevant human manipulation videos from the web, then uses them to improve robot policy training, addressing the scarcity of manipulation demonstration data.
More from Embodied
- GPT-6 Astra Scores 95% on Robot Control, 6.2x Fewer Tokens Than Fable 5.1's 40% — scaling01 · 2026-09-05
- Humanoids take over the sewing shop; author details a six-agent trading desk built on Grok Bot — mustafamhus · 2026-09-05
- Beijing's Hokmind shows lifelike Primal M1/F1 humanoids with real-time emotional AI conversation — CyberRobooo · 2026-09-05
- AR glasses turn hand and voice gestures into robot navigation, winning open source prize — GregCook2011 · 2026-09-05
- Scoble drives his 8-year-old robot to follow the newest robot to an SF office party — Scobleizer · 2026-09-05
- HUG: dexterous grasping trained on zero robot data, already running on robots worldwide — LerrelPinto · 2026-09-05