Representative work: π4: A Foundation Model for 4D Scene Geometry and Motion
Technical Report, 2026
I am a researcher at Shanghai Artificial Intelligence Laboratory. My research interests lie in Embodied AI, Spatial Intelligence, and Multimodal Generation.
I received my Ph.D. degree from the College of Computer Science and Technology at Zhejiang University in 2023, advised by Prof. Xiaowei Zhou.
I am looking for highly motivated interns at Shanghai AI Laboratory. Drop me an email if you are interested.
* Equal contribution, † corresponding author, ‡ project lead
We pursue physical intelligence by developing AI systems that can perceive the physical world, predict how it evolves, and act within it.
We develop world action models and agentic robotics systems for long-horizon tasks. We connect predictions of physical dynamics with task planning and robot control, and explore how robots can improve through interaction and execution feedback.
We develop methods for 3D/4D reconstruction, motion tracking, and motion capture, alongside large-scale spatial datasets. Agents can call these methods as tools to reason about space, and the same methods help build and annotate training data for manipulation and locomotion.
We develop multimodal generation as a core capability of embodied foundation models. We study how language, images, and actions can guide the generation of video, 3D assets, and motion for simulation, training, and world modeling.
Technical Report, 2026
Technical Report, 2026
arXiv, 2026
Technical Report, 2026
ICCV, 2021
CVPR, 2021 (Oral Presentation)
ECCV, 2020 (Oral Presentation)
iMoCap: Motion Capture from Internet Videos · IJCV, 2022
CVPR, 2019
ICCV, 2025 (Highlight)
ICCV, 2025
CVPR, 2024