Pengfei Wang

Decoupling memory conditioning from video generation allows for more data-efficient training and better spatial consistency in long-horizon video generation, even when exploring novel scenes.

Achieve robust, real-time 3D multi-object tracking in panoramic views by representing object states on a sphere, sidestepping the limitations of image-plane trackers and redundant Euclidean formulations.

Offline RL can be made more robust to distribution shift by directly optimizing against worst-case transition dynamics within an uncertainty set, leading to policies that avoid unreliable out-of-distribution actions.

Achieve geometrically consistent and explorable 3D scene generation from a single image by reformulating the problem as multi-view stereo matching on anchor views projected from a panorama.