About me
I am an undergraduate student in the Department of Probability and Statistics, School of Mathematical Sciences, Peking University, pursuing a B.S. in Statistics (expected 2027).
My research interests lie in large language models, mechanistic interpretability, training data attribution, and alignment. Broadly, I am interested in understanding how training data shapes models’ internal mechanisms and behaviors, and how such understanding can enable targeted interventions and more reliable AI systems.
At Peking University, I work with Prof. Liangming Pan in the PKU Interpretable Large Language model And Reasoning (PILLAR) Group on mechanistic data attribution and training-data interventions for LLMs. I have also worked with Prof. Ziyu Yao as a research intern at George Mason University, studying the internal representations and use of world models in planning agents.
My recent work includes Mechanistic Data Attribution (MDA) (ICML 2026 Oral (Top 0.7%)), From Reweighting to Rewriting (arXiv preprint), and Represented but Not Used, a research project on implicit world modeling in planning agents.
I am applying to PhD programs for Fall 2027. Please feel free to reach out at luoyuzzhang [at] stu [dot] pku [dot] edu [dot] cn.