Nan Xue

I'm leading a research team at RobbyAnt, dedicated to self-supervised learning of structured, metric, and actionable representations for robotics. Previously, I spent 10 years working on structured visual geometry — line segments, wireframes, and planes — in computer vision. My profile photo is computed by my HAWPv3.
Google Scholar  /  Github  /  Twitter  /  LinkedIn  /  Email

profile photo

Recent News

2026. 07: Meet LingBot 2.0! I lead LingBot-Vision for self-supervised vision pretraining, and take part in LingBot-Video for video pretraining and LingBot-VA 2.0 for video-action modeling.
2026. 03: Our paper LiP-Map: Interacted Planes Reveal 3D Line Mapping has been accepted to IEEE TPAMI.
2026. 01: Our new work LingBot-VA on causal world modeling for robot control is now available on arXiv with code.
2026. 01: Our new work LingBot-Depth for masked depth modeling is now available on arXiv with code.
2025. 10: Our NeurIPS paper Plana3r is released with code. See you in San Diego!

Selected Publications