Yiqiao Zhong

Do LLMs reason as we do? A synthetic study of transformers’ learning dynamics for compositions

• Invited talk at CS Machine Learning Lunch and Meetings, Feb 2026

How do LLMs generalize on out-of-distribution tasks? insights from model's internal representations

• Invited talk at Hong Kong University, IDS workshop, May 2025

• Invited talk at Peking University Mathematics Department, June 2025

• Invited talk at Univeristy of Pennsylvania ASSET seminar, Sep 2025

• Invited talk at University of Michigan CSP seminar, Nov 2025

Why are the logits of trained models distorted? A theory of overfitting for imbalanced classification

• Invited talk at Georgia Tech, Stochastics Seminar at Mathematics, Mar, 2025

Do You Interpret Your t-SNE Embeddings Correctly? A Perspective from Map-Continuity and Leave-One-Out

• Invited talk at University of Wisconsin–Madison, BMI seminar, Apr 2025

Can large language models solve compositional tasks? A study of out-of-distribution generalization

• Invited talk at Princeton University Wilks departmental seminar, Dec 2024

• Invited talk at CFE-CMStatistics, King's College London, Dec 2024

• Invited talk at Brin workshop University of Maryland, Nov 2024

• Invited talk at Washington State University, departmental seminar, Nov 2024

A Statistical Perspective of LLMs: Geometry and Embeddings in Transformers

• Invited talk at SIAM-MDS, Atlanta, Georgia, Oct 2024

• Invited talk at JCSDS, Kunming, China, July 2024

A Geometric Journey into the World of Large Language Models

• Invited talk at Machine Learning Seminar, Department of Computer Science and Engineering, University of Minnesota, Nov 2023

• Invited talk at Machine Learning Lunch Meetings, UW Madison, Nov 2023

Interpolation Phase Transition in Neural Networks: Memorization and Generalization under NT model

• Invited talk at Annual Allerton Conference 2023, “Learning and Estimation in High Dimensions” session, Sep 2023

• Invited talk at EcoSta conference 2023, Aug 2023

• Invited talk at Research School of Finance, Actuarial Studies and Statistics, Australian National University, Sep 2022.

• Invited talk at IST colloquium, Stanford University, Sep 2021; Wilks Statistics Seminar, Princeton University, Oct 2021; Neyman Statistics Seminar, UC Berkeley, Oct 2021.

Spectral Methods and Nonconvex Optimization: A Modern Statistical Perspective

• Invited talk at Statistics Department, Harvard University, MA, March 2019.

• Same talk at DPMMS, University of Cambridge, UK, Jan 2019.

Near-optimal bounds for phase synchronization

• Invited talk at SIAM Annual Meeting, Portland, OR, July 2018.

• Poster presentation at Bridging Mathematical Optimization, Information Theory, and Data Science, Princeton, May 2018.

• Invited talk at IDeAS seminar, The Program in Applied and Computational Mathematics, Princeton University, May 2017.

Spectral algorithm without trimming or cleaning works for exact recovery in SBM

• Invited talk at Joint Mathematics Meetings, San Diego, Jan 2018.

• Poster presentation at UCLA Workshop on Deep Learning Techniques, Los Angeles, Feb 2018.

eigenvector perturbation and robust covariance estimation

• Poster Presentation at Workshop on Networks, Random Graphs and Statistics, Columbia University, May 2016