Embodied AI Alliance: Riemann Dynamics & Partners Build 1M-Hour Data Foundation

量子位 · wechat · 2026-08-06

Riemann Dynamics has partnered with Lightwheel AI and Noitom Robotics to build a foundational dataset for embodied AI, setting a target to collect 1 million hours of robot training data by the end of 2026—double the current estimated global inventory.

The partnership addresses the bottleneck in embodied AI data scale and quality with clear division of labor: Noitom handles motion capture for real action data, Lightwheel AI provides simulation and evaluation, and Riemann Dynamics (formerly Kunlun's Matrix team) focuses on core world model training. Riemann previously trained its Riemann-1.0 model on 232k hours of data.

This collaboration aims to close the loop of 'collection-simulation-training-evaluation,' with some data open-sourced to the community to test the limits of Scaling Laws in the physical world.

Original post →

More from Companies & People

Companies & People channel →