Research

LLM Continual Unlearning

I study how large language models can remove requested information one request after another while keeping the knowledge they should retain.

In preparation
Sensitivity Stacking: Intuitive Backpropagation Without Index Hell
Y. Choi, K. Lee, J. M. Hyun
In preparation for ICLR 2027 Blog Post Track
My role: Co-second author (tentative); writing and idea development
Oct 2026
How You Mark It Matters: Particle-Level Framing Effects in a Korean Search Agent
Y. Choi, N. Kwak, H. J. Lee, K. Lee, J. M. Hyun
HCLT 2026, Oral
My role: Co-built experimental data and reviewed the manuscript
Code and draft Conference website
Sep 2026
Self-Claimed Authority Cues and Application-Layer Fact Verification in a Korean-Language Commercial Generative Search API
K. Lee†, Y. Choi†, J. M. Hyun, H. J. Lee, N. Kwak
Korea AI 2026
My role: Co-built experimental data and reviewed the manuscript
Conference website

† Equal first authors