Xingyuan Intelligence Launches Cross-Entity Data Engine Chorus, Exploring Robot 'Lexemes' and System Scaling
“行元智能”杨世琦：不做本体，押注机器人“词元” | 首发
Xingyuan Intelligence founder Yang Shiqi proposes a strategy of 'not building bodies, but betting on robot lexemes' in the field of embodied intelligence, and launches the cross-entity data engine Chorus. This engine processes multi-source human operation videos through cleaning and 3D reconstruction to enable the reuse of human experience across different robot bodies, aiming to address the issue of insufficient training data for robots and promote system scaling development.
The article highlights the development of 'Chorus', a cross-robotic-body data engine designed to enable scalable training data for robot learning by processing human operation videos. It emphasizes the importance of data scaling over hardware or model innovation, and outlines a system scaling approach that integrates data, models, and real-world deployment.
Source: Robotics — Chinese web discovery · Read original article ↗
English translation unavailable. Choose the original article to read it.
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
human demonstrations
40
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
把人类数据 “ 扩容40倍 ” 机器人训练数据短缺是行业老生常谈的痛点。
Open source S18
Source：Robotics — Chinese web discovery · finance.sina.com.cn