Xiaodong Chen

Xiaodong Chen

Master

Welcome to Xiaodong Chen’s Homepage

👋 About Me

Hello! I am Xiaodong Chen, currently a second-year M.S. student at Renmin University of China, School of Information, where I am advised by Prof. Jing Zhang. During my M.S. studies so far, I have authored 6 papers, with 3 as first author, published at top-tier venues including ACL and ICLR.

Prior to this, I completed my B.Eng. degree at Xi’an Jiaotong University, School of Computer Science and Technology (2020–2024), graduating 1st out of 33 in my cohort.

I previously worked as a Research Intern at the AGI Center, Ant Research Institute (Mar. 2025 – Oct. 2025), supervised by Jianguo Li and Haoxing Chen.

Most recently, I interned with the GLM Foundation Model Group at Zhipu AI (Nov. 2025 – Apr. 2026), working under the guidance of Zhengxiao Du.

🔬 Research Interests

My research focuses on:

  • 🏗️ Pre-training Model Architecture Improvement
  • ✂️ Model Compression & Inference Acceleration
  • ⚡ Speculative Decoding

📫 Feel free to reach out via email:
chenxiaodong@ruc.edu.cn

🥇 Honors and Awards

  • China National Scholarship (2025)
  • Renmin University of China Academic Excellence Scholarship — First Prize (2025)
  • Renmin University of China Academic Excellence Scholarship — First Prize (2024)
  • Xi’an Jiaotong University Academic Scholarship — First Prize (2022)

📝 Selected Publications

LLM-Streamline Streamlining Redundant Layers to Compress Large Language Models
Xiaodong Chen, Yuxuan Hu, Jing Zhang, Yanling Wang, Cuiping Li, Hong Chen
International Conference on Learning Representations (ICLR), 2025 (Spotlight)
arXiv GitHub
P2 Law P2 Law: Scaling Law for Post-Training After Model Pruning
Xiaodong Chen, Yuxuan Hu, Xiaokang Zhang, Yanling Wang, Cuiping Li, Hong Chen, Jing Zhang
Annual Meeting of the Association for Computational Linguistics (ACL), 2025
arXiv
MoBE MoBE: Mixture-of-Basis-Experts for Compressing MoE-based LLMs
Xiaodong Chen, Mingming Ha, Zhenzhong Lan, Jing Zhang, Jianguo Li
International Conference on Learning Representations (ICLR), 2026
arXiv GitHub
DND DND: Boosting Large Language Models with Dynamic Nested Depth
Tieyuan Chen, Xiaodong Chen, Haoxing Chen, Zhenzhong Lan, Weiyao Lin, Jianguo Li
International Conference on Learning Representations (ICLR), 2026
arXiv

🚀 Major Project Contribution

GLM-5: From Vibe Coding to Agentic Engineering Zhipu AI · Core Contributor

My Contribution: Parameter-Shared MTP Layer
Proposed a novel parameter-shared Multi-Token Prediction (MTP) layer for speculative decoding, which:

  • ✅ Improves acceptance rate of speculative tokens, accelerating inference speed
  • ✅ Integrated into GLM-5 serving pipeline for efficient deployment

arXiv