Xiaomi-Robotics-1: Scaling Vision-Language-Action Models with over 100K Hours of Real-World Trajectories
- Published
- Source
- arXiv
- Paper number
- 660
- Field
- Robotics
- arXiv ID
- 2607.15330
Key points
- An automatic VLM-based labeling pipeline is applied to more than 100,000 hours of real UMI manipulation trajectories.
- The pretraining stage shows a clear scaling law, with performance improving consistently as both data and model size increase.
- On RoboCasa365, it reaches 57.6 percent, which beats the previous best of 46.6 percent.
- On a real robot, it learns new tasks from less than 10 hours of data and reaches an average success rate of 75 percent.
Paper links
External research summaries. These are not HDATF publications or measured product results.