Zhiwei Xu
Scholar

Zhiwei Xu

Google Scholar ID: XzQRU3MAAAAJ
PhD student, University of Michigan
Machine LearningDeep LearningArtificial Intelligence
Citations & Impact
All-time
Citations
69
 
H-index
3
 
i10-index
2
 
Publications
6
 
Co-authors
7
list available
Resume (English only)
Academic Achievements
  • - Publications:
  • * Terminal-Bench: A Benchmark for Agents on Hard, Realistic Tasks in Command Line Interfaces (Submitted)
  • * Recommendations Beyond Catalogs: Diffusion Models for Personalized Generation (Submitted)
  • * Benign Overfitting in Single-Head Attention (NeurIPS 2025)
  • * Let Me Grok for You: Accelerating Grokking via Embedding Transfer from a Weaker Model (ICLR 2025)
  • * Benign Overfitting and Grokking in ReLU Networks for XOR Cluster Data (ICLR 2024)
  • * Inference of Dependency Knowledge Graph for Electronic Health Records (Journal of the Royal Statistical Society, Series B, 2025)
  • - Academic Service: Conference Reviewer @ ICML (2024), ICLR (2025), L4DC (2025); Workshop Reviewer @ W3L (2024)
Research Experience
  • - Teaching Assistant at the University of Michigan, Ann Arbor, including courses such as STATS 500: Statistical Learning, STATS 413: Applied Regression Analysis, STATS/DATASCI 451: Bayesian Data Analysis, and STATS 250: Introduction to Statistics and Data Analysis, responsible for grading homework and conducting office hours
  • - Teaching Assistant at USTC, Real Analysis, Spring 2021, responsible for weekly lab classes and grading homework
Education
  • - University of Michigan, Ann Arbor, Ph.D. in Statistics, 2022 - Present, Advisors: Wei Hu & Yixin Wang
  • - University of Science and Technology of China, B.S. in Statistics, 2018 - 2022, Advisor: Weiping Zhang
Background
  • - Research Interests: Theoretical and scientific understanding of foundation models, particularly focusing on the length generalization and reasoning abilities of Large Language Models
  • - Field: Statistics
  • - Brief Introduction: Fourth-year Ph.D. candidate in Statistics at the University of Michigan, Ann Arbor, advised by Prof. Wei Hu and Prof. Yixin Wang.