Tsinghua and Partners Propose Collective Learning for Robot Fleets

Tsinghua, AgiBot and Infinigence AI proposed Collective Experience Learning (CEL), a paradigm where deployed robots share execution trajectories and failure segments to teach each other, significantly improving policy success rates. The project is available on Hugging Face.

2026-10-20 ~ 2026-10-20 · 2 related posts

• Fleet learning for robots: sharing execution experience boosts policy success — Chao Yu · 2026-10-20
• Tsinghua and AgiBot's Collective Experience Learning lets robot fleets teach each other from failures — KyeGomezB · 2026-10-20