New MT3 Paradigm Enables Robots to Learn 1000 Tasks in Under 24 Hours
chris_j_paxton · x · 2026-07-31
A new study from Imperial College London introduces Multi-Task Trajectory Transfer (MT3), a highly efficient imitation learning paradigm that made the cover of Science Robotics.
The core idea involves decomposing manipulation trajectories into sequential alignment and interaction phases, combined with retrieval-based generalisation. In the few-shot regime (<10 demonstrations per task), this decomposition improves data efficiency by an order of magnitude over traditional single-phase behavioral cloning.
Thanks to this efficiency, MT3 can teach a robot 1000 distinct everyday tasks in under 24 hours of human demonstration time (approx. 17 hours), while generalising to unseen object instances. The team validated the method's capabilities and limitations across more than 5,650 real-world rollouts.
Related event: New Robot Learning Paradigm on Science Robotics: 1000 Tasks in a Day(2 posts)→
More from Embodied
- NVIDIA launches RTX Spark for local AI on Windows PCs — ostrisai · 2026-10-16
- Transforming quadruped robot runs on real hardware: per-pose RL policies in Isaac switched on morph — _Stocko_ · 2026-10-16
- Foldable iPhone Duo reportedly sells out launch-day inventory, some deliveries slip to 2027 — Polymarket · 2026-10-16
- KinoGen synthesizes robot loco-manipulation reference motions directly via kinematically grounded video generation — ChongZzZhang · 2026-10-16
- Science Robotics paper introduces F³, predictive event-based vision beyond optical flow — CSProfKGD · 2026-10-16
- iPhone Duo foldable appears on South African retailer iStore at R47,999 in apparent leak — shauntrennery · 2026-10-16