Homepage · Google Scholar · Email
I study representation learning through distributional and generative perspectives, especially self-supervised learning and its connection to generative learning. I also work on the theory of language models.
I am pursuing my PhD in Applied Mathematics at The Hong Kong Polytechnic University, advised by Prof. Defeng Sun and Prof. Houduo Qi. Prof. Yuling Jiao, my master's advisor at Wuhan University, continues to advise me during my PhD, and we work closely together.
My research brings generative learning tools into self-supervised representation learning by treating the latter as a distribution-matching problem. DM develops this formulation; FBDM extends it through flow matching.
Recasting self-supervised learning as distribution matching opens the door to generative learning tools for representation learning.
What if a generative flow could learn representations rather than generate samples?
Learning a Flow to Self-Supervised Representations
We introduce a minimax approach to debias existing self-supervised learning methods. This adversarial formulation improves downstream performance while helping establish theoretical guarantees for the learned representations.
Adv-SSL: Adversarial Self-Supervised Representation Learning with Theoretical Guarantees, NeurIPS 2025. Code
We develop a theoretical framework to model and understand zero-shot prediction, in-context learning, and chain-of-thought reasoning. We seek to explain how demonstrations and intermediate reasoning steps can improve performance along the progression from zero-shot prediction to in-context learning and chain-of-thought.
Beyond the Prompt in Large Language Models: Comprehension, In-Context Learning, and Chain-of-Thought
Implementations of machine learning algorithms, for studying and adapting the underlying methods.
Authors are listed alphabetically by surname in all publications.
For publications, talks, and academic service, visit my personal homepage.