- I am the founding director of Celia Large Model Application Lab @HKRC, leading a team of 50+ talented researchers and engineers. Our mission bridges cutting-edge AI research with real-world smart-device applications, spanning Large Language Models (LLMs), multimodal intelligence (vision, audio, omni, etc.), AI agents, and embodied AI. We are always looking for passionate researchers, engineers, and interns to join our team.
- Before that, I was a senior researcher at SenseTime Group where I investigated on-device multi-modal models including vision language models (VLMs) and diffusion models (DMs).
- I hold a PhD from MMLab, CUHK, supervised by Prof. Xiaogang Wang and Prof. Hongsheng Li.
News
- [Aug., 2026] Five papers accepted by EMNLP 2026 (3 main conference and 2 findings).
- [July, 2026] Released Boogu, the state-ossf-the-art open-sourced image generation model.
- [June, 2026] Two papers accepted by ECCV 2026.
- [May, 2026] Excited to see our Aura featured and highly regarded in the latest Interaction Models released by Thinking Machines Lab!
- [May, 2026] Three papers accepted by ICML 2026.
- [Apr., 2026] Released Aura, a SOTA streaming video understanding framework.
- [Apr., 2026] One paper accepted by ACL 2026.
- [Mar., 2026] Released CoVe, a novel agentic post-training framework.
- [Feb., 2026] Released Capybara, the first unified video generation & editing model.
- [Feb., 2026] Three papers accepted by CVPR 2026.
- [Jan., 2026] Two papers accepted by ICLR 2026.
- [Oct., 2025] Check out our new work MathCanvas, the first unified multi-model that enables thinking while drawing auxiliary lines!
- [Aug., 2025] One paper accepted by EMNLP 2025.
- [July, 2025] Released a novel reinforcement learning algorithm GHPO for LLM post-training.
- [Apr., 2025] Released a SOTA level Image-to-Video model Pusa.
- [Sep., 2024] Joined Huawei Hong Kong Research Center as TopMinds.
- [July, 2022] Joined Sensetime as senior researcher.
- [June, 2022] Graduated from MMLab, CUHK.
Visitors
