Hwiyeong Lee

About Me

Hi, I’m Hwiyeong Lee, a Ph.D. student at the Graduate School of Data Science, Seoul National University, advised by Prof. Taesup Kim as a member of LAAL.

I’m motivated to explore how LLMs acquire and process knowledge, and importantly, how we can manipulate it post hoc. Specifically, I approach LLMs as objects of scientific study, aided by the emerging line of work in the field of Mechanistic Interpretability to peel back the curtain on their inner workings. Ultimately, I aim to solve practical challenges like Knowledge Editing and Unlearning through a clear understanding of how LLMs operate.

News

[Aug. 2026] 🎓 Completed My M.S. & Starting My Ph.D. at SNU!
After two amazing years at Hanyang University, I’m excited to have completed my M.S. and to begin a new chapter as a Ph.D. student at Seoul National University. I’m deeply grateful to everyone who supported me throughout my master’s journey, and I look forward to the challenges and opportunities ahead! ✨
[May. 2026] 🎉 Paper Accepted to ICML 2026!
My first-author paper “Query Lens: Interpreting Sparse Key-Value Features with Indirect Effects” has been accepted to ICML 2026. It feels especially meaningful to present this work in Seoul, the city where I was born and raised! 🇰🇷
[Aug. 2025] 🎉 Paper Accepted to EMNLP 2025!
Excited to share that my first-author paper “Does Localization Inform Unlearning? A Rigorous Examination of Local Parameter Attribution for Knowledge Unlearning in Language Models” has been accepted to EMNLP 2025 Main! Looking forward to presenting our work in Suzhou! 🚀

Publications

arXiv

How Learning Governs Unlearning across the Memorization–Generalization Spectrum

Hwiyeong Lee^*, Hyelim Lim^*, Ingyu Bang, Hoki Kim^†, Taeuk Kim^†

arXiv preprint, 2026.

PDF BibTex

ICML

Query Lens: Interpreting Sparse Key-Value Features with Indirect Effects

Hwiyeong Lee, Ingyu Bang, Uiji Hwang, Hyelim Lim, Taeuk Kim^†

International Conference on Machine Learning (ICML), 2026.

PDF Code BibTex

EMNLP

Does Localization Inform Unlearning? A Rigorous Examination of Local Parameter Attribution for Knowledge Unlearning in Language Models

Hwiyeong Lee, Uiji Hwang, Hyelim Lim, Taeuk Kim^†

Conference on Empirical Methods in Natural Language Processing (EMNLP), 2025.

PDF Code BibTex

KCC

Uncovering Hidden Vulnerabilities in Machine Unlearning: Adversarial Attack as a Probe and Pruning as a Solution

Hwiyeong Lee, Jeonghyeon Kim, Sangheum Hwang^†

Korea Computer Congress (KCC), 2024.

PDF

Educations

[Sep. 2026 - Present] Ph.D. in Data Science, Seoul National University
[Sep. 2024 - Aug. 2026] M.S. in Artificial Intelligence Semiconductor Engineering, Hanyang University
[Mar. 2018 - Aug. 2024] B.S. in Industrial and Information Systems Engineering, Seoul National University of Science and Technology