Donghyun Son

prof_pic.png

I’m currently looking for Ph.D. opportunities that are aligned with my interests. Feel free to reach me!

I am an undergraduate researcher majoring in Computer Science and Engineering at Seoul National University, working on efficient machine learning.

I aim to build an intelligent system that delivers strong problem-solving capabilities at low-cost. I believe efficiency is key to achieving the goal, and focus on two directions: lowering the inference cost of large models, and improving training efficiency. My current interests include:

  • efficient long-context inference
  • data- and compute-efficient training of large models
  • system-level optimization of ML workloads

I have worked on making large language models efficient, including KV cache quantization (NeurIPS 2025, with Prof. Sungjoo Yoo) and systems for agentic workflows (under review, with Dr. Saurabh Agarwal and Prof. Aditya Akella).

Aside from research, I enjoy algorithmic problem solving and have competed in competitive programming contests. You can find me on codeforces and BOJ.

Links: Github | CV | Google Scholar | X

 

Selected Publications

  1. NeurIPS 2025
    NSNQuant.png
    NSNQuant: A Double Normalization Approach for Calibration-Free Low-Bit Vector Quantization of KV Cache
    Donghyun Son, Euntae Choi, and Sungjoo Yoo
    In Advances in Neural Information Processing Systems (NeurIPS), 2025
    arXiv
  2. WSDM 2023 (Oral)
    wsdm23.png
    Reliable decision from multiple subtasks through threshold optimization: Content moderation in the wild
    Donghyun Son*, Byounggyu Lew*, Kwanghee Choi, and 5 more authors
    In Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining (WSDM), 2023
    arXiv
  3. OOD-CV@ICCV23
    ood-cv23.png
    Gradient estimation for unseen domain risk minimization with pre-trained models
    Byounggyu Lew*, Donghyun Son*, and Buru Chang
    In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023
    arXiv