<strong>Dunyao</strong> Xue

Institute of Statistics and Big Data

Renmin University of China

Beijing, China

Welcome to my page! I am Dunyao Xue, a PhD student in Statistics at the Institute of Statistics and Big Data, Renmin University of China, advised by Dr. Cheng Meng and Dr. Wenlin Dai.

Prior to this, I received a BSc degree in Mathematics from the Cuiying Honors School, Lanzhou University.

My research interests include data compression, optimal transport, ensemble learning, and large language model architectures. I am especially interested in applying statistical and machine learning methods to large language models (LLMs), with the goal of improving the performance or efficiency of model training and inference.

data compression optimal transport Mixture-of-Experts LLM decoding

First/co-first author: ICML / ICLR / JCGS

Recent work includes MP-MoE at ICML, SEINT at ICLR, and core-elements subsampling for ALS at JCGS.

currently

Statistics PhD, Data Science and Artificial Intelligence track

Working with Cheng Meng and Wenlin Dai at the Institute of Statistics and Big Data.

also interested in

dataset distillation, generative models, and model compression

Committed to bringing statistical learning methods into frontier large language models.

Statistical compression core elements, dataset distillation, compact statistical learning

Transport geometry invariant metrics, cross-space comparison, structured distributions

Efficient LLMs MoE routing, ensemble pruning, geometry-aware decoding

If you are interested in my research, feel free to drop me an email.

news

more updates

Jul 06, 2026 update

From July 6 to 11, I attended ICML 2026 in Seoul and presented our MoE work, MP-MoE.

Presenting the MP-MoE poster at ICML 2026 in Seoul

Jul 01, 2026 update

I attended Huawei’s “Zijin Summit” Young Scholars Paper Sharing Seminar and shared our optimal transport work, SEINT.

Group photo at Huawei's Zijin Summit Young Scholars Paper Sharing Seminar

Jun 05, 2026 update

Project “Large Language Model Inference Optimization Based on Ensemble Pruning” received institute-level support as a 2026 Graduate Student Scientific Research Fund Project of the Institute of Statistics and Big Data, one of eight funded projects institute-wide.

May 01, 2026 update

Our paper Breaking the Echo Chamber: A Dynamic Ensemble Pruning Perspective on MoE was accepted by ICML.

Apr 23, 2026 update

I was invited to attend the International Seminar on Foundational Artificial Intelligence (FAIC) and presented a poster. View Poster

selected publications

full list

ICLR

SEINT: AN EFFICIENT SE (p)-INVARIANT TRANSPORT METRIC DRIVEN BY POLAR TRANSPORT DISCREPANCY-BASED REPRESENTATION

Junyi Lin^*, Dunyao Xue^*, Jun Yu, Hongteng Xu, and Cheng Meng

International Conference on Learning Representations, 2026

Co-first author

PDF Code Poster
JCGS

Core-elements Subsampling for Alternating Least Squares

Dunyao Xue, Mengyu Li, Cheng Meng, and Jingyi Zhang

Journal of Computational and Graphical Statistics, 2026

First author

PDF Code Poster
ICML

Breaking the Echo Chamber: A Dynamic Ensemble Pruning Perspective on MoE

Xinlai Kang^*, Dunyao Xue^*, Zhengbo Wang, Chengshuo Du, Cheng Meng, Hanting Chen, Hang Zhou, and Xinghao Chen

Forty-third International Conference on Machine Learning, 2026

Co-first author

PDF Code Poster