Photo of Qingming Liu

Qingming Liu (刘青明)

Ph.D. Student @ HKUST

I am a Ph.D. student at The Hong Kong University of Science and Technology, where I will be advised by Prof. Wenhan Luo and Prof. Ping Tan. My research focuses on 3D vision and physical intelligence, with particular interests in embodied AI, world models, and sim-to-real learning.

I received my M.Phil. in Computer Science from The Chinese University of Hong Kong, Shenzhen, supervised by Prof. Kui Jia and Prof. Zhen Liu, to whom I am deeply grateful for their guidance and support. Prior to my M.Phil., I worked closely with Prof. Yuan Liu, and I am equally grateful for his mentorship along the way.

More about me

I received my B.Eng. in Artificial Intelligence from Beijing Jiaotong University, where I was fortunate to have Prof. Runmin Cong introduce me to research. Before my M.Phil., I spent one year at the City University of Hong Kong with Prof. Junhui Hou, where I worked on dynamic scene reconstruction and started collaborating with Prof. Yuan Liu. During my M.Phil., I was also a research intern at DexForce, a company founded by my supervisor Prof. Kui Jia, working on Physics-aware 3D generation for embodied AI.

My recent goal is to build 3D-native world models that support action-conditioned, interaction-aware simulation — going beyond static asset generation toward models that understand how the physical world moves and reacts.

Outside research, I enjoy badminton and swimming.

Email / CV / Google Scholar / GitHub / LinkedIn / Address

Education

News

Selected Publications

(* denotes equal contribution)

PAct: Part-Decomposed Single-View Articulated Object Generation
Qingming Liu, Xinyue Yao, Shuyuan Zhang, Yueci Deng, Guiliang Liu, Zhen Liu, Kui Jia
ACM SIGGRAPH Asia, 2026
project page / arXiv / code
Given a single-view input, PAct generates complete part-decomposed geometry and infers the kinematic structure, producing physically plausible articulated 3D objects.
Nabla-R2D3: Effective and Efficient 3D Diffusion Alignment with 2D Rewards
Qingming Liu*, Zhen Liu*, Dinghuai Zhang, Kui Jia
NeurIPS, 2025
project page / arXiv / OpenReview / code
The first effective and sample-efficient RL framework that aligns 3D-native diffusion models with human preferences using only 2D reward feedback.
MoDGS: Dynamic Gaussian Splatting from Casually-captured Monocular Videos with Depth Priors
Qingming Liu*, Yuan Liu*, Jiepeng Wang, Xianqiang Lyv, Peng Wang, Wenping Wang, Junhui Hou
ICLR, 2025
project page / arXiv / OpenReview / code / video
With 3D-aware initialization and an ordinal depth loss, MoDGS learns dynamic Gaussian splatting for novel-view synthesis from casually captured (even static) monocular videos.

Experience

Academic Service