Guangyu Li

Welcome!

My name is Guangyu Li, a graduate student at University of Science and Technology of China (USTC). Now, I am an incoming PhD student at the Department of Industrial Engineering and Decision Analytics in HKUST.

Open to collaborations! Feel free to reach out if our research interests align.

Research Interests

• Computational Statistics: Numerical Optimization, Monte Carlo Methods.
• Probabilistic ML & Generative AI: Scalable & Efficient Probabilistic Models, Uncertainty Estimation, Bayesian Methods.

Furthermore, I am interested in various directions and hope to explore additional research fields in the future, combining theoretical statistical foundations with AI implementations.

Publications/Preprints

[1] Accelerating Discrete Langevin Samplers via Continuous Intermediates [pdf]
Guangyu Li and Ruqi Zhang , ProML 2026 Workshop.
[2] Understanding Generalization and Forgetting in In-Context Continual Learning [arxiv]
Guangyu Li, Meng Ding and Lijie Hu, ICML 2026.
[3] The Generalization Error Bound of LoRA [Under Review][[pdf]]
Haojun Wu, Guangyu Li and Canhong Wen

News

• [May. 2026] One paper accepted to ProML 2026 Workshop.
• [Apr. 2026] One paper accepted to ICML 2026.
• [Nov. 2024] The website is built.

Powered by Jekyll and Minimal Light theme.