Xiaomi-Robotics-1 shows robot motion improves more from data than bigger models
The Decoder · rss · 2026-07-21
Xiaomi-Robotics-1 suggests that, for robot motion training, more data matters more than a bigger model.
The key result: Xiaomi trained the system on more than 100,000 hours of motion data collected from people using camera-equipped handheld grippers, rather than from robots.
The article says:
- Increasing training data improved performance far more than scaling model size.
- The gains had not yet plateaued.
- Even so, absolute success rates were still low.
It is a useful reminder that robot learning may be bottlenecked by data quality and scale, not just parameter count.
More from Embodied
- NVIDIA open-sources FLORA, a robotics data engine ~125x cheaper than GPT-6 Astra — nvidia · 2026-10-27
- Embodied AI startups need a different operational discipline, investor says — ___Mufasaa · 2026-10-27
- π0.5 study: raising demos from 10 to 438 lifts Place Ball in Bowl success from 7% to 83% — stingning · 2026-10-27
- GPT-6 Astra controls dexterous hands zero-shot at just 5.2% success in 500 trials — stingning · 2026-10-27
- HandsOn benchmark: 66 verified tasks and 18,000 human demos for dexterous manipulation — stingning · 2026-10-27
- DexPort: a handheld interface with a real robot hand for dexterous-hand data collection — pabbeel · 2026-10-27