Hanyang Li

About

Hanyang Li

I am a Ph.D. candidate in the Department of Industrial Engineering and Operations Research (IEOR) at the University of California, Berkeley, advised by Professor Ying Cui.

My research focuses on the design and analysis of optimization algorithms, especially for nonsmooth and ill-conditioned problems. I study how algorithms can make reliable progress on these landscapes by building and refining models of their local geometry. My recent interests also include optimization under precision and memory constraints. I examine how quantization affects optimization dynamics and use these insights to design algorithms for memory-efficient model training and inference.

Email: hanyang_li AT berkeley DOT edu

Education

Publications and Preprints

Theory and Algorithms for Nonsmooth Optimization

  1. Subgradient Regularization: A Descent-Oriented Subgradient Method for Nonsmooth Optimization

    Hanyang Li, Ying Cui

    Submitted, 2025

    [code]

  2. Variational Theory and Algorithms for a Class of Asymptotically Approachable Nonconvex Problems

    Hanyang Li, Ying Cui

    Mathematics of Operations Research, 51(1), 1–34

    [slide]

    • Katta G. Murty Prize for Best Paper in Optimization, UC Berkeley (2025).
  3. A Decomposition Algorithm for Two-Stage Stochastic Programs with Nonconvex Recourse Functions

    Hanyang Li, Ying Cui

    SIAM Journal on Optimization, 34(1), 306–335

    [code] [slide]

    • Runner-up, Dupacova–Prekopa Best Student Paper Prize (2023).
    • Third Place, INFORMS JFIG Paper Competition (2022).

Optimization for Memory-Efficient Training and Inference

  1. Beyond Shadow Weights: Quantization-Aware Training as Quantized-Endpoint Descent

    Sheng-An Xu, Hanyang Li, Jianhao Ma, Ying Cui

    Submitted, 2026

  2. Understanding Quantization-Aware Training: Gradients at Quantized Weights Bias to the Low-Loss Basin

    Hanyang Li, Jianhao Ma, Ying Cui

    Submitted, 2026

  3. Rounding in Preconditioner Space: Redesigning 4-bit AdamW Optimizer-State Quantization

    Hanyang Li, Shao Tang, Daniel Braithwaite, Gregory Dexter, Leonardo Neves, Abhishek Shivanna, Aman Gupta, Hiroto Udagawa, Daniel Silva, Rohan Ramanath

    Submitted, 2026

    • Preliminary version appears in the NeurIPS 2026 Workshop on Optimization for Machine Learning (OPT 2026).

2026

  1. Beyond Shadow Weights: Quantization-Aware Training as Quantized-Endpoint Descent

    Sheng-An Xu, Hanyang Li, Jianhao Ma, Ying Cui

    Submitted, 2026

  2. Understanding Quantization-Aware Training: Gradients at Quantized Weights Bias to the Low-Loss Basin

    Hanyang Li, Jianhao Ma, Ying Cui

    Submitted, 2026

  3. Rounding in Preconditioner Space: Redesigning 4-bit AdamW Optimizer-State Quantization

    Hanyang Li, Shao Tang, Daniel Braithwaite, Gregory Dexter, Leonardo Neves, Abhishek Shivanna, Aman Gupta, Hiroto Udagawa, Daniel Silva, Rohan Ramanath

    Submitted, 2026

    • Preliminary version appears in the NeurIPS 2026 Workshop on Optimization for Machine Learning (OPT 2026).

2025

  1. Subgradient Regularization: A Descent-Oriented Subgradient Method for Nonsmooth Optimization

    Hanyang Li, Ying Cui

    Submitted, 2025

    [code]

  2. Variational Theory and Algorithms for a Class of Asymptotically Approachable Nonconvex Problems

    Hanyang Li, Ying Cui

    Mathematics of Operations Research, 51(1), 1–34

    [slide]

    • Katta G. Murty Prize for Best Paper in Optimization, UC Berkeley (2025).

2024

  1. A Decomposition Algorithm for Two-Stage Stochastic Programs with Nonconvex Recourse Functions

    Hanyang Li, Ying Cui

    SIAM Journal on Optimization, 34(1), 306–335

    [code] [slide]

    • Runner-up, Dupacova–Prekopa Best Student Paper Prize (2023).
    • Third Place, INFORMS JFIG Paper Competition (2022).

Talks and Presentations

  • INFORMS Annual Meeting · San Francisco, CASession: Optimization Methods for Quantizing and Pruning LLMs · Upcoming
    11/2026
  • Cornell ORIE Young Researchers Workshop · Ithaca, NYPoster presentation · Upcoming
    10/2026
  • INFORMS Optimization Society Conference · Atlanta, GASession: Recent Advances in Stochastic and Large-Scale Optimization
    03/2026
  • Hong Kong University of Science and TechnologyJoint IEDA/ISOM Seminar
    01/2026
  • Hong Kong Polytechnic UniversityAMA Seminar Series on Young Scholars in Optimization and Data Science
    01/2026
  • International Conference on Continuous Optimization · Los Angeles, CASession: First-order Methods for Nonsmooth Constrained Optimization
    07/2025
  • INFORMS Annual Meeting · Phoenix, AZSession: Recent Advances in Nonsmooth Optimization
    10/2023
  • XVI International Conference on Stochastic Programming · Davis, CAStudent Paper Prize Competition
    07/2023

Teaching

University of California, Berkeley · Graduate Student Instructor

  • INDENG 242A: Machine Learning and Data AnalyticsFall 2025

University of Minnesota, Twin Cities · Teaching Assistant

  • IE 3012: Optimization IISpring 2022
  • IE 5441: Financial Decision MakingSpring 2022

University of Science and Technology of China · Teaching Assistant

  • Mathematical Analysis B1Fall 2020