Ph.D Student
The University of Hong Kong
About Me
I am a researcher at Anuttacon, studying the intelligence underlying human behavior, performance, and interaction: how machines/avatars can learn to move, express, and respond the way humans do. Generative models are my primary instrument.
My research interests lie on three threads:
- Video as a behavior prior: pretraining large video generation models (LPM 1.0) as a unified foundation for behavior and performance
- Real-time diffusion-based motion generation: compact generative frameworks that respond to people at interactive rates, spanning full-duplex conversational characters, human-human and human-object interaction, and multimodal control
- Structure before scaling: as action generation scales up in character control and robotics, I work on the core problems beneath it (representation design, data laws, model scaling)
Formerly, I was a Ph.D. student in the Department of Computer Science at the University of Hong Kong, supervised by Prof. Taku Komura. My Ph.D. work argues that human motion needs the right representation as an action space before it can benefit from scale. That conviction still runs through everything I build, from monocular video-based motion capture to real-time character controllers over increasingly complex action spaces.
Before it, I received my B.Eng. and M.Sc. degrees from Shandong University, supervised by Prof. Baoquan Chen. I spent three wonderful years (2017-2020) at the Reality Capture Group in Beijing Film Academy and the CFCS in Peking University, collaborating with the best research scientist Kfir Aberman.
I used to intern at Electronic Arts (2020), miHoYo (2022), FAIR Meta (2022-2023), and Tencent (2023-2024). Thanks for the wonderful experiences and the great friends I met there!