Kwanhee Lee
Hi, I’m a second-year Master’s student at Graduate School of Artificial Intelligence, POSTECH, advised by Prof. Namhoon Lee. Currently, I’m an Academic Visitor in DASLab at ISTA, hosted by Prof. Dan Alistarh.
I design optimization algorithms for scalable and robust compression of large-scale foundational models, aiming to make them efficient and usable for everyone. I also have general interest on optimization for deep learning.
For more details, here’s my CV
If you have any questions or want to collaborate with me, feel free to reach me via kwanhee.lee@postech.ac.kr
news
| May 23, 2026 | We released Kimi-K2.5-P48-NVFP4-W4A4-Preview — Kimi-K2.5 with paired 4:8 sparsity and NVFP4 W4A4 quantization, targeting Blackwell sparse tensor cores. More coming soon! |
|---|---|
| Apr 19, 2026 | Will be serving as a reviewer for NeurIPS 2026 and attending ICLR 2026 in Rio de Janeiro, 🇧🇷. See you in Rio! |
| Jan 26, 2026 | Our paper on extreme LLM sparsity has been accepted to ICLR 2026! |
| Jan 05, 2026 | Our paper SAFE won the Best Paper Award at POSTECH-GSAI BK21 2025! |
| Nov 17, 2025 | Our new paper on extreme LLM sparsity won Best Paper Award at JKAIA 2025! |
selected publications
-
The Unseen Frontier: Pushing the Limits of LLM Sparsity with Surrogate-Free ADMMIn ICLR, 2026JKAIA 2025 Best Paper Award -
SAFE: Finding Sparse and Flat Minima to Improve PruningIn ICML, 2025Spotlight(top 2.6%) poster | 2025 POSTECH-GSAI BK21 Best Paper Award -
