Bing Liu (刘冰)

Hello! I’m Bing.

I am currently a fourth-year PhD student at the College of Control Science and Engineering, Zhejiang University.

Bio. I am advised by Chengcheng Zhao and co-advised by Li Chai and Peng Cheng. Previously, I was a visiting PhD student at the School of Mathematics, Peking University, working with Kun Yuan. I received my bachelor’s degree in Automation from Tongji University in 2023.

Research. My research currently focuses on the design and understanding of machine learning optimizers, building upon my prior work in distributed optimization. In particular, I’m a big fan of the Shampoo optimizer and have recently started exploring schedule-free optimization.

Industry. Since December 2025, I have been a research intern with the LLM pretraining team at Ant Group, where I deeply involved in developing the Ling 3.0 model series (latest: Ling 3.1 Flash), focusing on optimizers and late-stage model averaging to reduce training noise during pretraining.

bing_liu@zju.edu.cn Google Scholar Photo

Publications

Representative works

  1. How Bregman Divergences Shape Shampoo

    Bing Liu, Wenjie Zhou, Chengcheng Zhao, Hongtao Zhang, Boao Kong, Felix Dangel, Wu Lin

    Preprint

    * In collaboration with Felix Dangel and Wu Lin, coauthors of KL-Shampoo.

  2. Row-Stochastic Matrices Can Provably Outperform Doubly Stochastic Matrices in Decentralized Learning

    Bing Liu, Boao Kong, Limin Lu, Kun Yuan, Chengcheng Zhao

    ICML 2026

    * An extended journal version coming soon.

  3. Privacy-Preserving Resilient Vector Consensus

    Bing Liu, Chengcheng Zhao, Li Chai, Peng Cheng, Jiming Chen

    IEEE Transactions on Automatic Control (TAC) Full Paper

Other First-author works

  1. Rethinking Bregman Divergences in Kronecker-Factored Optimizers

    An early version of "How Bregman Divergences Shape Shampoo," presented at HiLD, ICML 2026.

  2. ImprovDML: Improved Trade-off in Private Byzantine-Resilient Distributed Machine Learning

    Preprint · Under major revision at Automatica

  3. Trade-off between Privacy and Accuracy in Resilient Vector Consensus

    American Control Conference (ACC), 2024

Selected Honors

  • National Scholarship for Doctoral Students (Top 2%), 2026
  • First Prize, National Finals of the 17th National University Students Smart Car Competition, 2022
  • First-Class Award, China High School Physics Olympiad, Hebei, 2018

Academic Service

  • Reviewer: IEEE Transactions on Automatic Control (TAC), IEEE Transactions on Signal Processing (TSP), International Conference on Learning Representations (ICLR), American Control Conference (ACC), and IEEE Conference on Decision and Control (CDC).