publications

2026

  1. Hardware-Native Joint Sparse-Quantization for Trillion-Scale Mixture-of-Experts
    Kwanhee Lee, Namhoon Lee, and Dan Alistarh
    arXiv preprint, 2026
    arXiv Code HuggingFace
  2. hidden_plain_sight.png
    Hidden in Plain Sight: The Overlooked Significance of Canonical Elements for Extreme LLM Sparsity
    Hyeondo Jang, Kwanhee Lee, Dongyeop Lee, and 1 more author
    In EMNLP, 2026
    Main conference | CKAIA 2026 Best Paper Award
    arXiv Code
  3. unseen_frontier.png
    The Unseen Frontier: Pushing the Limits of LLM Sparsity with Surrogate-Free ADMM
    Kwanhee Lee, Hyeondo Jang, Dongyeop Lee, and 2 more authors
    In ICLR, 2026
    JKAIA 2025 Best Paper Award
    arXiv Code OpenReview Thread
  4. dynamic.jpg
    Dynamic Network Compression via Probabilistic Channel Pruning
    Kwanhee Lee, and Hyang-Won Lee
    Neural Networks, 2026
    Paper

2025

  1. safe.png
    SAFE: Finding Sparse and Flat Minima to Improve Pruning
    Dongyeop Lee, Kwanhee Lee, Jinseok Chung, and 1 more author
    In ICML, 2025
    Spotlight(top 2.6%) poster | 2025 POSTECH-GSAI BK21 Best Paper Award
    arXiv Code OpenReview