Pro-0 trains dexterous robot hands from human demos with zero on-robot data
chris_j_paxton · x · 2026-11-12
A Stanford-affiliated team unveiled Pro-0, co-designing tactile data-collection gloves with ProHand and a learning pipeline to train dexterous robot policies purely from human demonstrations — with zero on-robot training data.
Early results cover three short tasks; a video shows the full human data-collection and robot trajectory-replay pipeline. The team presented a poster at the UMI Arena workshop at CoRL in Austin on 11/12.
More from Embodied
- Liquid AI lays out roadmap to put AI on half the world's devices via model-hardware co-design — JosephJacks_ · 2026-11-12
- Wayve unveils Learned Safety Envelope: end-to-end trained live risk prediction for AI driving — alexgkendall · 2026-11-12
- Chris Paxton endorses "the age of deploy cos" to bridge robotics' demo-to-deployment gap — chris_j_paxton · 2026-11-12
- CORL happy hour draws 1300+ RSVPs, more than the entire conference did two years ago — JasonMa2020 · 2026-11-12
- Physical Intelligence clarifies: zero-shot LLM already beats fine-tuned open VLA — paigeinsf · 2026-11-12
- Goal-conditioned locomotion in sim: recovery policy survives stairs and slopes — burny_tech · 2026-11-12