1. Personal Information
Full Name: Xiu Li
Email: li.xiu@sz.tsinghua.edu.cn
Title: Professor & Doctoral Advisor
Affiliation: Shenzhen International Graduate School, Tsinghua University, Shenzhen, Guangdong, China
2. Academic Summary
Xiu Li is a Professor at the Shenzhen International Graduate School of Tsinghua University with expertise in artificial intelligence, computer vision, and reinforcement learning. She has published over 200 academic papers in top AI and computer vision venues such as CVPR, NeurIPS, ICML, and ICLR. Her work has been cited 10,000+ times and she has been included in the Stanford & Elsevier top 2% global AI scientists list for multiple consecutive years (2022–2024).
3. Education and Training
Ph.D., Tsinghua University
Visiting Scholar, University of California, Irvine (2016–2017)
Visiting Scholar, Georgia Institute of Technology (2007–2008)
Visiting Scholar, The Hong Kong Polytechnic University (2006–2007)
Visiting Scholar, The University of Hong Kong (2005)
(Exact Ph.D. graduation year not explicitly listed on public site)
4. Professional Experience
Professor, Shenzhen International Graduate School, Tsinghua University (2016–present)
Associate Researcher and roles at multiple leading universities during early career (2000s)
5. Research Interests
Professor Li’s research spans a broad range of topics in artificial intelligence and machine learning, including:
Artificial Intelligence & Machine Learning
Computer Vision
Reinforcement Learning
Multimodal perception and generation
Deep learning for visual understanding
6. Scholarly Output
Professor Li has published 200+ academic papers with high impact in top journals and major AI/vision conferences (e.g., CVPR, NeurIPS, ICML, ICLR). Her research includes work on diffusion models for low-level vision, reinforcement learning algorithms, multimodal animation and pose estimation benchmarks, and advanced deep learning techniques.
Notable publications include:
Diffusion Models in Low-Level Vision: A Survey (IEEE TPAMI, 2025)
UniHead: Unifying Multi-Perception for Detection Heads (IEEE TNNLS, 2024)
Off-Policy RL Algorithms Can Be Sample-Efficient … (Information Sciences, 2024)