Jueun Kim김주은

I am an M.S. student in Artificial Intelligence at KAIST AI, advised by Prof. Chulhee Yun. My recent research focuses on deep learning optimization, neural network learning dynamics (Edge of Stability), and loss landscapes.

I received my B.S. from Yonsei University with a double major in Computer Science and Mathematics. I’m always open to collaborations on the edge of stability and deep learning optimization. Feel free to reach out!

Email | GitHub | CV | LinkedIn 		Jueun Kim portrait

Yonsei University Purdue University KAIST AI
B.S. in CS, Math
2021-2026		 Visiting Scholar
2023-2024		 M.S. in AI
2026-

News
September 2026 AMUSE was accepted to NeurIPS 2026 as a Spotlight paper. I will present it in Sydney 🇦🇺 this December.
June 2026 AMUSE was accepted to HiLD at ICML 2026.
May 2026 AMUSE: Anytime Muon with Stable Gradient Evaluation was published on arXiv.

Publications
AMUSE FineWeb Llama 720M validation perplexity plot
AMUSE: Anytime Muon with Stable Gradient Evaluation
Jueun Kim*, Baekrok Shin*, Jihun Yun, Beomhan Baek, Minhak Song, Chulhee Yun
Conference on Neural Information Processing Systems (NeurIPS), 2026
Spotlight Presentation (292/30,709 = 0.95%)
We propose AMUSE, a schedule-free Muon optimizer that stabilizes training by gradually shifting gradient evaluation from the fast Muon sequence to a stable averaged sequence, while retaining Muon's rapid progress along flat loss-landscape directions.
arXiv | code | bibtex 
@inproceedings{kim2026amuse,
  title={AMUSE: Anytime Muon with Stable Gradient Evaluation},
  author={Kim, Jueun and Shin, Baekrok and Yun, Jihun and Baek, Beomhan and Song, Minhak and Yun, Chulhee},
  booktitle={Advances in Neural Information Processing Systems},
  note={Spotlight; also accepted to the High-dimensional Learning Dynamics Workshop at ICML 2026},
  year={2026}
}

Projects
RGB-SLAM with Depth Estimation
RGB-SLAM with Depth Estimation
Jueun Kim
Research Project, Purdue University, 2023-2024
Worked on dense visual SLAM and modified the RGB-D model architecture into RGB visual SLAM, removing the requirement for depth data.
Meetable
Meetable
Jueun Kim
Frontend Project, 2023-2024
Developed a website for making appointments among many people and managing a personal calendar.
Project Page

Template based on thiswebsite.