Preprint
Sep 2026
Proxy2World: Learning to Generate Worlds From Lightweight Proxies without Seeing Them
Proxy2World is presented, a controllable world model that learns depth-conditioned RGB generation and joint RGBD generation through cross-modal flow matching from ordinary posed RGBD videos, without training on authored proxy-video pairs.
Hong-Li Xu, Wei-Long Yan, An-Bang Wang et al.
· 0 citations