AgiBot World 2026 Theme 3: Reinforcement Learning

From the catalogue. Listed by one researcher from one source page.

• id: agibot-world-2026-theme-3-reinforcement-learning
• kind: real manipulation scripted or autonomous
• robot or device: AgiBot real robots (class: mobile manipulator)
• how collected: real robot RL rollouts, failures kept
• size as stated: 14 tasks, 11,430 trajectories
• hours: not stated | hours only claimed: none | episodes: 11,430
• year: 2026
• organisation: AgiBot
• licence: not stated (class: not stated)
• access: gated (class: gated)
• link (dataset): https://x.com/AGIBOTofficial/status/2095402214324072731
• note: Announced on X 3 Sept 2026; numeric gated repos on agibot-world HF org.

Confirm the licence at the link before relying on it.