RoboTTT Extends Robot Memory to 5 Minutes
DrJimFan · x · 2026-07-16
RoboTTT: Bringing Test-Time Training to Robot Learning
- This repost highlights a collaborative research project from Stanford SVL and NVIDIA Robotics focusing on test-time training in robot learning.
- The authors state they natively scaled the robot model to a context of 8,000 timesteps—roughly equivalent to 5 minutes of "muscle memory"—while keeping inference costs unchanged.
- Instead of previous robot policies that could only process a few frames at a time and quickly forgot recent events, this approach compresses history into a small internal network within the model, enabling the robot to learn continuously over longer periods.
- The paper/method is named RoboTTT: with each new sensor input, a gradient update is applied to the internal core, utilizing a fixed-size hidden state to carry longer experiences.
Related event: RoboTTT brings test-time training to robot policies(6 posts)→
More from Embodied
- Amazon and Google sold 600M+ smart speakers, so why no AGI-era successor? — julianlehr · 2026-09-11
- ECCV26 Oral: Flow Matching Enables Single-Stage Multi-View Point Cloud Registration — ducha_aiki · 2026-09-11
- Polish developers build iPhone app that detects nearby Meta smart glasses — Low-Honeydew6483 · 2026-09-11
- Ant's Afu health AI hits 150M users, unveils AI+hardware health alliance at Bund Summit — APPSO · 2026-09-11
- Johns Hopkins Launches Full-Stack Hands-on Robot Learning Class with SO-101 Arm Kits — _krishna_murthy · 2026-09-11
- SyncWorld: In-Context Robot World Model Simulates Unseen Views and Embodiments Zero-Shot — ChongZzZhang · 2026-09-11