Takuya Koriyama 郡山拓也

I am a third-year Ph.D. student in statistics at the University of Chicago, Booth School of Business.

I’m broadly interested in the mathematical theory and algorithms underlying statistics and machine learning, as well as applied research in modern AI.

Email CV (PDF) Google Scholar GitHub LinkedIn

Research interests

My previous research has focused on high-dimensional statistics, high-dimensional probability and geometry, and combinatorial stochastic processes. Recently I’ve been working on large language models, with a focus on alignment, routing, and model evaluation.

Papers

Google Scholar

Preprints & working papers 3

Power-law clustering with Gibbs partitions: universal estimation and predictive limits

Takuya Koriyama

Working paper
A Markov Chain Approach to Preference Alignment

Takuya Koriyama and Tengyuan Liang

arXiv:2606.22652, 2026

Slides
Existence of solutions to the nonlinear equations characterizing the precise error of M-estimators

Pierre C. Bellec and Takuya Koriyama (alphabetical order)

arXiv:2312.13254, 2023, submitted

Slides

Publications 9

Asymptotic Inference for Exchangeable Gibbs Partitions

Takuya Koriyama

Stochastic Processes and their Applications, 202, 105074, 2026

Slides
Asymptotics of resampling without replacement in robust and logistic regression

Pierre C. Bellec and Takuya Koriyama (alphabetical order)

Bernoulli, 32(4) 3063-3087, 2026
Denoising Diffusions with Optimal Transport: Localization, Curvature, and Multi-Scale Complexity

Tengyuan Liang, Kulunu Dharmakeerthi, and Takuya Koriyama

Transactions on Machine Learning Research, 2835-8856, 2026
Precise asymptotics of bagging regularized M-estimators

Takuya Koriyama, Pratik Patil, Jin-Hong Du, Kai Tan, and Pierre C. Bellec

The Annals of Statistics, 54(2) 796-833, 2026

Code
Asymptotic mixed normality of maximum likelihood estimator for Ewens--Pitman partition

Takuya Koriyama, Takeru Matsuda and Fumiyasu Komaki

Advances in Applied Probability, 58(1) 214-234, 2026

Slides
Phase transitions for the existence of unregularized M-estimators in single index models

Takuya Koriyama and Pierre C. Bellec

Proceedings of the 42nd International Conference on Machine Learning (ICML2025)
Error estimation and adaptive tuning for unregularized robust M-estimator

Pierre C. Bellec and Takuya Koriyama (alphabetical order)

Journal of Machine Learning Research 26 (16), 1-40, 2025
Corrected generalized cross-validation for finite ensembles of penalized estimators

Pierre C. Bellec, Jin-Hong Du, Takuya Koriyama, Pratik Patil and Kai Tan (alphabetical order)

Journal of the Royal Statistical Society Series B: Statistical Methodology, 87(2), 289-318, 2025

Journal Slides Code
Fully Data-driven Normalized and Exponentiated Kernel Density Estimator with Hyvärinen Score

Shunsuke Imai, Takuya Koriyama, Shouto Yonekura, Shonosuke Sugasawa and Yoshihiko Nishiyama

Journal of Business & Economic Statistics 43 (1), 110-121, 2025

Journal

Blog

All posts

No posts yet. Research notes will appear here.

Education & experience

Full CV (PDF)

2024–present

University of Chicago

PhD student in econometrics and statistics, Booth School of Business

Advisor: Tengyuan Liang

2022–2024

Rutgers University

PhD student in Statistics

Advisor: Pierre C. Bellec

2018–2022

University of Tokyo

BE in Mathematical Engineering · Department of Mathematical Engineering and Information Physics

2021–2022

RIKEN Center for Brain Science (CBS)

Student Researcher

Theory of combinatorial stochastic processes and its application to network data analysis