RoboTTT: 8K Context Robot Policies
scott_e_reed · x · 2026-07-15
This research introduces RoboTTT, extending robot policies to an 8K timesteps visuo-motor context. It pushes the usable context length beyond current SOTA levels while maintaining constant inference latency.
Demonstrated capabilities include:
- One-shot imitation from human video demos using only a few minutes of experience
- Continuous improvement during deployment
- Self-recovery from disturbances
- Completion of a 5-minute, 10-step assembly task
Comments emphasize that such TTT methods are highly suitable for robot learning and are likely agnostic to whether the backbone is a VLA or a video model.
Related event: RoboTTT brings test-time training to robot policies(6 posts)→
More from Embodied
- Apple responds after iPhone Duo inner displays crack along the hinge — mobileraj · 2026-10-27
- VLA models encode task concepts as curved multi-dimensional geometry, 19-author study finds — ninamiolane · 2026-10-27
- Munro Live Posts Full Cybercab Engineering Deep-Dive Video — yunta_tsai · 2026-10-27
- Tesla's Robotaxi Fleet Keeps Growing in Austin — yunta_tsai · 2026-10-27
- Building Halo Infinite Forge maps with local LLMs, DCjson and Blender MCP — swagonflyyyy · 2026-10-27
- SoftBank-Backed XAG Ships Drones With Autonomous Docks That Clean, Recharge and Refill — pstAsiatech · 2026-10-27