I'm a first-year CS PhD student at CIS, University of Pennsylvania working under the supervision of Professor Jiatao Gu.
Previously, I completed my undergraduate studies at Columbia College, Columbia University in New York with a degree in Mathematics and Computer Science. I was fortunate enough to be advised by Professor Changxi Zheng at the Columbia Computer Graphics Group, where I also learned a lot from Dr. Rundi Wu.
My research interests include 3D/4D world modeling, video generation, and vision for embodied tasks.
arXiv, 2026
UniLAP learns a continuous latent action policy from action-free videos using flow matching, enabling adaptation across embodiments without requiring future frames at inference.
arXiv, 2026
OmniDream transforms a silent monocular video into an immersive panoramic experience with spatially aligned audio, using object-centric sound generation and physics-based acoustic simulation.
arXiv, 2026
PointAction bridges the gap between generative video models and robust robot control by jointly predicting RGB frames and dynamic 3D pointmaps, which serve as a universal, geometry-aware action interface.
* Equal contribution.
SIGGRAPH, 2025
A method for recovering dynamic 3D indoor lighting from a video by learning a light field with help from a 2D diffusion model.