On September 1, Liangyuan Xinchuang released and open-sourced the LightNav-0 universal navigation brain. The navigation brain relies on Real2Sim2Real technology. Based on 2000+ real indoor and outdoor scenes, the team synthesized 4,000+ hours of post-training data on a large scale to achieve zero-sample real machine navigation across entities, tasks, and scenes. LightNav-0 is an achievement of large-scale alignment in the three-stage paradigm of Liangyuan's new “large-scale pre-training, large-scale alignment, and large-scale deployment”, providing a reusable technical path for large-scale production of post-robot training data.

On September 1, Liangyuan Xinchuang released and open-sourced the LightNav-0 universal navigation brain. The navigation brain relies on Real2Sim2Real technology. Based on 2000+ real indoor and outdoor scenes, the team synthesized 4,000+ hours of post-training data on a large scale to achieve zero-sample real machine navigation across entities, tasks, and scenes. LightNav-0 is an achievement of large-scale alignment in the three-stage paradigm of Liangyuan's new “large-scale pre-training, large-scale alignment, and large-scale deployment”, providing a reusable technical path for large-scale production of post-robot training data.