Seungyoo Lee

Ph.D. Student @ KAIST AI · Data‑Centric LLMs & Multimodal Models

Seungyoo Lee

Hi! I am a Ph.D. student at KAIST AI, advised by Prof. Juho Lee at the SIML Lab. I study LLMs and multimodal models from a data-centric perspective, with a focus on synthetic data generation and understanding model behavior through data.

Previously, I was a research intern at Kakao, working on encoder-free multimodal models and domain-specific LLMs, and a partner researcher at Markr.AI. Before that, I worked with Prof. Xiaopeng Yang (POSTECH) on medical image analysis.

Contact: punctuate [at] kaist [dot] ac [dot] kr

Download CV

News
Selected Publications

(*) denotes equal contribution, (†) denotes co-corresponding authors. See publications for the full list.

NeurIPS
SAVE: Sparsity-Aware Influence Estimation for Vocabulary-Expanded LLMs
Seungyoo Lee, Giung Nam, Seanie Lee, Hyungi Lee, Juho Lee
In The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS 2026) Accepted · Sep 24, 2026
ICML
OCNR: Stabilizing Self-Play by Mitigating Iteration-Collapse With One-Class Novelty Rewards
Seungyoo Lee, Giung Nam, Hyungi Lee†, Juho Lee†
In Proceedings of the 43rd International Conference on Machine Learning (ICML 2026), PMLR 306 Accepted · Apr 30, 2026
NeurIPS
PANGEA: Projection-Based Augmentation with Non-Relevant General Data for Enhanced Domain Adaptation in LLMs
Seungyoo Lee, Giung Nam, Moonseok Choi, Hyungi Lee†, Juho Lee†
In Advances in Neural Information Processing Systems 38 (NeurIPS 2025) Accepted · Sep 18, 2025
ICLR
Variational Bayesian Pseudo-Coreset
Hyungi Lee, Seungyoo Lee, Juho Lee
In The Thirteenth International Conference on Learning Representations (ICLR 2025) Accepted · Jan 22, 2025