PeopleofSuperintelligence

Researcher

Mingyue Xu

Mingyue Xu is a computer scientist whose work connects learning theory and statistics with questions about how machine learning generalizes. She earned degrees in mathematics at Shanghai Jiao Tong University and data science at Columbia before joining Purdue's computer science graduate program in 2023. Her early papers explored distributed statistical inference and dimension reduction; with Steve Hanneke, she classified universal learning rates for empirical risk minimization at NeurIPS 2024 and COLT 2025. In 2026, she, Gal Vardi and Itay Safran proved that grokking can occur in ridge regression, and ICML gave their paper an Outstanding Paper Honorable Mention. She is a fourth-year PhD student at Purdue, advised by Steve Hanneke.

Timeline

  1. 2021–2022

    Mathematics and data science degrees

    She earns a B.S. in mathematics at Shanghai Jiao Tong University in 2021 and an M.S. in data science at Columbia University in 2022.

  2. 2023

    A graduate student at Purdue

    She joins Purdue's Department of Computer Science as a graduate student in fall 2023.

  3. 2023

    Distributed sparse inference

    With Jiyuan Tu, Weidong Liu, and Xiaojun Mao, she coauthors a 2023 study of distributed semi-supervised sparse statistical inference. Its method uses a multi-round distributed debiased estimator that combines labeled and unlabelled data.

  4. 2023–2025

    Smoothed gradients for dimension reduction

    She and Gan Yuan, Samory Kpotufe, and Daniel Hsu submit a 2023 paper on sufficient dimension reduction for multi-index models. It shows that estimating the expected smoothed gradient outer product can achieve a fast parametric convergence rate; the work appears in the SIAM Journal on Mathematics of Data Science in 2025.

  5. 2024

    Four learning rates for ERM

    At NeurIPS 2024, she and Steve Hanneke show that empirical risk minimization in the realizable case has four possible universal learning rates. The paper classifies which concept classes fall into each rate.

  6. 2025

    Agnostic learning rates

    At COLT 2025, she and Steve Hanneke show that agnostic empirical risk minimization has three possible universal learning rates: exponential, super-root, or arbitrarily slow.

  7. 2026

    Grokking in ridge regression

    In 2026, she, Gal Vardi, and Itay Safran prove that gradient descent with weight decay can produce grokking in over-parameterized ridge regression. Their work shows that hyperparameter tuning can control the delay, and ICML gives the paper an Outstanding Paper Honorable Mention.

  8. As of October 2026

    A Purdue computer science PhD

    As of October 2026, she is a fourth-year PhD student in computer science at Purdue University, advised by Steve Hanneke. Her work focuses on foundations of machine learning.

Career

Education

  1. 2022Columbia UniversityM.S. in Data Science
  2. 2021Shanghai Jiao Tong UniversityB.S. in Mathematics
  3. Purdue UniversityPhD student

From the biography.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Columbia University. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  2. 84% of people shaping AI with recorded undergraduate and doctoral schools changed institutions for their doctorateIncludes Shanghai Jiao Tong University. 84% of people shaping AI with recorded undergraduate and doctoral schools changed institutions for their doctorate. Across 2,116 people with both an undergraduate degree and a doctorate recorded, we trace which undergraduate schools feed which doctoral programs, how often students stay at the same school, and how paths differ by birth-country region.
  3. Nonprofit roles are 2.3 times as common in AI safetyIncludes Columbia University. We estimate where people focused on safety, alignment and evaluation work. Nonprofit roles appear for 25.1% of safety people with known affiliations, compared with 11.0% in other areas.
  4. Universities had a net outflow of 2,068 moves in AI careersIncludes Columbia University. Our estimates of career flows and median h-index across universities and companies in AI, including the role of joint appointments and earlier academic jobs.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Mingyue Xu yet. Send us a free-licensed photo.

  1. Mingyue Xu - Department of Computer Science - Purdue University cs.purdue.edu. Checked 21 October 2026.
  2. Distributed Semi-Supervised Sparse Statistical Inference arxiv.org. Checked 21 October 2026.
  3. Efficient Estimation of the Central Mean Subspace via Smoothed Gradient Outer Products arxiv.org. Checked 21 October 2026.
  4. Mingyue Xu | Computer Science PhD Student raymondmyxu.github.io. Checked 21 October 2026.
  5. Universal Rates of Empirical Risk Minimization papers.nips.cc. Checked 21 October 2026.
  6. Universal Rates of ERM for Agnostic Learning proceedings.mlr.press. Checked 21 October 2026.
  7. To Grok Grokking: Provable Grokking in Ridge Regression arxiv.org. Checked 21 October 2026.
  8. Announcing the ICML 2026 Awards – ICML Blog blog.icml.cc. Checked 21 October 2026.