|
Sangyoon Lee I’m a third-year M.S.-Ph.D. student in GSAI (Graduate School of Artificial Intelligence) at POSTECH Efficient Learning Lab (EffL), advised by Prof. Jaeho Lee. My research interests lie in accelerating neural network training and understanding optimization dynamics through the lens of spectral and simplicity biases. I have mainly focused on neural fields, exploring methods to accelerate and improve training efficiency. |
Publications |
|
Fast Training of Sinusoidal Neural Fields via Scaling Initialization
Taesun Yeom*, Sangyoon Lee*, Jaeho Lee, ICLR, 2025 arXiv / code We propose a simple yet effective approach for accelerating neural field training. |
|
In Search of a Data Transformation That Accelerates Neural Field Training
Junwon Seo*, Sangyoon Lee*, Kwang In Kim, Jaeho Lee, CVPR, 2024 (Oral Presentation) arXiv / code / demo Random pixel permutation accelerates neural field training by removing easy-to-fit patterns, guiding the network away from fixating on lowfrequency components. |
|
Beware of the Batch Size: Hyperparameter Bias in Evaluating LoRA
Sangyoon Lee, Jaeho Lee, arXiv Vanilla LoRA remains a strong baseline once batch size is properly tuned, and provide a practical guideline for small-scale proxies to tune batch size in LoRA fine-tuning. |
|
HyperCLOVA X 8B Omni
NAVER Cloud HyperCLOVA X Team Tech Report, 2026 arXiv |
|
Multi-frame Restoration for High-rate Lissajous Confocal Laser Endomicroscopy
Minhee Lee, Sangyoon Lee, Jaeho Lee, MICCAI, 2026 arXiv |
Education |
|
|
M.S.-Ph.D in Artificial Intelligence Pohang University of Science and Technology (POSTECH), South Korea |
2024.02 – Present |
|
B.S. in Computer Science and Engineering Pohang University of Science and Technology (POSTECH), South Korea |
2018.03 – 2023.08 |
Experience |
|
|
Research Intern at NAVER HyperClovaX Vision Team Worked on Multimodal Pre-Training. |
2025.10 – 2026.03 |