New! Try Everlyn.ai — The fastest video model we built.
I am a Professor at HKUST AMC and a Visiting Professor at Yale University. My research focuses on advancing AGI, particularly in image and video generation. Previously, I worked at Meta as a Research Scientist. I am also the co-founder of Everlyn, with Yann LeCun as our advisor. I earned PhD in computer science from USC and completed my undergraduate studies at USTC (SCGY). I am honored to be a recipient of the USC Distinguished Alumni Award.
News
Students & Interns
Grant
Teaching
September 2025
- “Hierarchical Fine-Grained Preference Optimization for Physically Plausible Video Generation” by Haodong Chen (my intern) accepted to NeurIPS 2025.
- “When Semantics Mislead Vision: Mitigating Large Multimodal Models Hallucinations in Scene Text Spotting and Understanding” by Yexin Liu (my PhD student) accepted to NeurIPS 2025.
- “VideoGen-of-Thought: Step-by-Step Generation of Multi-Shot Videos with Minimal Manual Intervention” by Mingzhe Zheng (my PhD student) accepted as an oral presentation at the NeurIPS 2025 NextVid Workshop.
- Serving as Area Chair for ICLR 2026.
- Giving a talk at HKUST-GZ CMA Seminar.
- Yann LeCun joins Everlyn as our advisor.
July 2025
- “Meta-Learning for Speeding Up Large Model Inference in Decentralized Environments”, first-authored by my PhD student Zihao Wang, has been accepted to COLM 2025.
- Internship placements: Congratulations to my first-year PhD students for securing research internships at Kuaishou (Kling), Tencent, and Bytedance.
June 2025
- “An Integrated Framework for Autoregressive Video Generation and Understanding”, for which I serve as Principal Investigator, has been awarded funding under the GRF/ECS 2025-26 scheme.
- Two papers accepted to ICCV 2025:
- “DreamDance: Animating Human Images by Enriching 3D Geometry Cues from 2D Poses” (with Yatian as my intern)
- “Model Reveals What to Cache: Profiling-Based Feature Reuse for Video Diffusion Models” (with my PhD students Xuran, Yexin, and Yaofu)
- Serving as Area Chair for AAAI-26.
- Serving as Associate Editor for APSCNNIPA Transactions on Signal and Information Processing.
- Visiting Meta AI, New York.
- Interviewed by RTHK on GenAI and video generation, discussing applications in Hong Kong local community. (Cantonese)
- Interviewed by CNN on Embodied AI, discussing the current status, challenges, and future directions.
February 2025
January 2025
Paper Highlights
December 2024
- Our video project was approved for the HKSTP Incubation Program (3 years).
- Meta-Learning for Speeding Up Large Model Inference (AAAI DAI Workshop)
- Project funded by TSSSU-O.
- Towards Secure and Private AI: A Framework for Decentralized Inference (NeurIPS Workshop on RBFM)
November 2024
- Delivered a talk at DevCon Bangkok.
September 2024
- Visited Abu Dhabi and met with the Royal Family to discuss collaboration.
- Delivered a talk in Singapore at Token2049.
August 2024
- Hosted a roundtable at Foresight 2024 in Hong Kong.
- Invited as a panel speaker at ABS 2024 in Taipei.
- Visited CUHKSZ.
April 2024
- Foresight interview.
- Invited as a panel speaker at the Harvard Blockchain Conference 2024.
March 2024
- Our project was selected by Binance Labs for incubation.
Current Students
Zihao Wang
Yexin Liu
Xuran Ma
Mingzhe Zheng
Jiajun Zha
Yaofu Liu
Haoze Zheng
Wenjie Shu
Weixiao Wang
Zhefan Rao (with Qifeng Chen)
Dengyang Jiang
Group attending ICLR 2025 in Singapore
Past Students and Interns
Xianfeng Wu
Yajing Bai
Harold Chen
Yatian Pang (with Li Yuan)
Feilong Tang
Bin Ma
Sen Liang
Yuxiang Lin
Xin Yang (with Qifeng Chen)
Yiyu Zhuang
Dongli Xu
Grants
- GRF/ECS 2025: The General Research Fund / Early Career Scheme.
- TSSSU: The Technology Start‑up Support Scheme for Universities helps university teams start technology businesses and commercialise R&D results.
- HKSTP Incubation Programme: A 3‑year incubation programme from Hong Kong Science and Technology Parks.
- HKUST‑KAIST Seed Grant: A collaborative seed grant that supports joint research projects between HKUST and KAIST.