I'm a research intern at KAIST AI under the supervision of Prof. Seong Joon Oh, where I will continue my PhD starting in September 2026. Previously, I received my M.S. in Artificial Intelligence in 2026 and my B.S. in Computer Science and Engineering with Magna Cum Laude honors in 2024, both from Sogang University.
My research lies in the field of AI Safety, with a particular interest in building models that are both fair, responsible, and transparent. Specifically, I am interested in machine unlearning, which seeks to selectively remove the influence of specific data to protect user privacy, copyright, and data ownership. Recently, I have been exploring unlearning techniques in both large language models (LLMs) and large vision-language models (LVLMs). Also, I have developed a growing interest in model explainability, as understanding how a model forgets certain knowledge is crucial for ensuring trustworthy and verifiable unlearning. By combining unlearning and explainability, my goal is to contribute toward safer and more accountable AI systems.
News
01/2026: 1 paper accepted @ ICLR 2026.
01/2026: Joined the STAI group at KAIST AI.
11/2025: 🏆 Granted AI SeoulTech Graduate Scholarship by Seoul Scholarship Foundation ($3.5K).
09/2025: 1 paper accepted @ NeurIPS 2025.
07/2025: 1 paper accepted @ COLM 2025.
01/2025: Served as a visiting scholar at the Tübingen AI Center (Host: Prof. Dr. Seong Joon Oh).
03/2024: 🏆 Granted Digital Human & Entertainment Scholarship by Smilegate ($27K).
03/2024: 🏆 Granted Albatross Fellowship by Sogang University ($21K, top 10% rank honor).
Publications
*Co-first Authors, †Co-corresponding Authors
❒ Knowledge Vector Weakening: Efficient Training-free Unlearning for Large Vision-Language Models [code]
Yejin Kim, Dongjun Hwang, Sungmin Cha†, Junsuk Choe†.
Preprint.
[3] Enhancing Multi-Image Understanding through Delimiter Token Scaling [code]
Minyoung Lee, Yeji Park, Dongjun Hwang, Yejin Kim, Seong Joon Oh, Junsuk Choe.
ICLR 2026.
[2] OVS Meets Continual Learning: Towards Sustainable Open-Vocabulary Segmentation [code]
Dongjun Hwang, Yejin Kim, Minyoung Lee, Seong Joon Oh, Junsuk Choe.
NeurIPS 2025.
[1] Improving Fisher Information Estimation and Efficiency for LoRA-based LLM Unlearning [code]
Yejin Kim*, Eunwon Kim*, Buru Chang†, Junsuk Choe†.
COLM 2025.