Scholar
Jing Xu
Google Scholar ID: jlrroGQAAAAJ
Tsinghua University
Citations & Impact
Citations
282
H-index
7
i10-index
6
Publications
13
Co-authors
0
Contact
Publications
Browse publications on Google Scholar (top-right) ↗
Resume (English only)
Academic Achievements
- Publications: Understanding Nonlinear Implicit Bias via Region Counts in Input Space, ICML, 2025; Scalable Model Merging with Progressive Layer-wise Distillation, ICML, 2025; Functionally Constrained Algorithm Solves Convex Simple Bilevel Problems, Neurips, 2024; Random Masking Finds Winning Tickets for Parameter Efficient Fine-tuning, ICML, 2024; On Bilevel Optimization without Lower-level Strong Convexity, COLT, 2024; Towards Data-Algorithm Dependent Generalization Analysis: a Case Study on Overparameterized Linear Regression, Neurips, 2023; Quantifying the Variability Collapse of Neural Networks, ICML, 2023; Faster Gradient-Free Algorithms for Nonsmooth Nonconvex Stochastic Optimization, ICML, 2023.
Research Experience
- Quantitative Research Intern at Citadel Securities (Jun. 2025 – Sept. 2025), built LLM pipelines to extract signals and build alphas from text-based alternative dataset; Machine Learning Intern at Moonshot AI (Feb. 2025 – Jun. 2025), developed efficient and stable optimization algorithms (e.g., Muon and its variants) for LLM pre-training; Quantitative Research Intern at Jump Trading (Jun. 2024 – Aug. 2024), conducted alpha analysis for China’s stock market.
Education
- Received a B.S. degree in artificial intelligence from Peking University in 2021, advised by Professor Liwei Wang; currently pursuing a Ph.D. in computer science at the Institute for Interdisciplinary Information Sciences, Tsinghua University, advised by Professor Andrew Chi-Chih Yao, the A.M. Turing laureate of 2000.
Background
- A fifth-year Ph.D. student in computer science at the Institute for Interdisciplinary Information Sciences in Tsinghua University. Research interests lie at the intersection of theoretical and applied machine learning. On the theoretical side, focuses on establishing provable guarantees for the generalization and optimization of machine learning algorithms. Empirically, has hands-on experience with large-scale LLM pre-training and is committed to designing efficient optimization algorithms that improve the scalability and performance of pre-training.
Co-authors