Xingyu Zhu 朱星宇
Xingyu Zhu 朱星宇

PhD Student

About Me

I am a second year PhD student in Computer Science at Princeton University working on theoretical machine learning and language modeling. I am fortunate to be advised by Professor Sanjeev Arora. I did my undergrad at Duke, where I was fortunate to be advised by Professor Rong Ge.

Download CV
Interests
  • Machine Learning Theory
  • Language Modeling
  • Theoretical Computer Science
Education
  • Ph.D. in Computer Science

    Princeton University

  • B.S. in Math & Computer Science

    Duke University

Research Interests
My interest spans across both theoretical and empirical ML. Recently I am especially interested in understanding the power of large language models through a semi-theoretical lens. I have also worked on optimization dynamics of deep neural nets. Besides, I am also broadly interested in theoretical computer science and algorithmic fairness.
Recent News

New Paper - On the Power of Context-Enhanced Learning in LLMs

Check out our new paper on context-enhanced learning! When training LLMs, putting helpful information purely in-context can exponentially improve sample efficiency while leading to little verbatim memorization of the in-context materials!

Featured Publications
Recent Publications
(2025). On the Power of Context-Enhanced Learning in LLMs. arXiv.
(2023). Understanding Edge-of-Stability Training Dynamics with a Minimalist Example. ICLR 2023.
(2023). Fairness in the Assignment Problem with Uncertain Priorities. AAMAS 2023.
(2022). Dissecting Hessian: Understanding Common Structure of Hessian in Neural Networks. arXiv.
Talks