On July 8, Ant Group’s LingBot Technology officially released its next-generation embodied foundation model, LingBot-VLA 2.0, as open-source software. This marks a significant enhancement over the LingBot-VLA 1.0 model, which was open-sourced earlier in January of the same year. During its pre-training phase, LingBot-VLA 2.0 was trained on an extensive dataset comprising 60,000 hours of high-quality real-world physical interaction data. This dataset encompasses 20 different configurations sourced from 17 leading robot brands. Additionally, the new version provides improved support for multiple degrees of freedom across key robotic components, including the head, waist, end-effector, and mobile chassis. These enhancements result in greater adaptability to various robot configurations, expanded support for complex movements, and streamlined deployment processes.
