Xiaomi's Robotics-1 Tests VLA Scaling Laws with 100k Hours of Real Data
stepjamUK · x · 2026-08-11
The author discusses the practical relationship between data scale and model capability in robot learning, noting that previous research often disconnected pre-training scaling curves from real-world robot performance.
The review highlights Xiaomi's Robotics-1, a foundation VLA model pre-trained on over 100,000 hours of real-world manipulation data collected via handheld UMI devices. The paper is praised for testing whether scale behaves as scaling laws predict and for being honest about where those predictions break down upon contact with a real physical environment.
More from Embodied
- Starkey Omega AI Hearing Aid Wins Healthcare Eagles Award — BrandonSawalich · 2026-08-11
- AgenticROS Integrates RealSense for Web-Based Robot Teleoperation — chrismatthieu · 2026-08-11
- XPENG's AI Chief on VLA 2.0, Autonomous Driving Gaps, and Europe Timeline — kimmonismus · 2026-08-11
- Enact Launches Post-Training Infrastructure to Fix Robot Model Failures in Real World — ycombinator · 2026-08-11
- Osseus (YC S26) Launches Agentic Platform to 10x Robot Development — ycombinator · 2026-08-11
- China's Magnetic Wall-Climbing Robot Uses VR Teleoperation for High-Risk Jobs — lukas_m_ziegler · 2026-08-11