I’m a second-year M.S. student at the Multimodal Language Processing Lab, KAIST, advised by Prof. KyungTae Lim. I studied computer science and mathematics at Seoul National University.
I work on the alignment and interpretability of multimodal language models, with an interest in AI safety. I enjoy working on cases where apparent correctness conceals a failure. An answer may not be grounded in what the model actually observed; a harm may survive filtering because it reads as benign. I look for these failures, identify what is going wrong and why it goes unnoticed, then explore ways to address them.
Publications
* denotes equal contribution.
Updates
Other
- Opera: I’ve been taking classical vocal lessons for KAIST grad students for about a year now. I sang in ensembles at the 6th KAIST Classical Vocal Concert, and I’m thrilled to be singing Tosca’s Vissi d’arte at the upcoming 7th.
- A cappella: I sing with Instru, an a cappella group from my undergrad years. I especially love singing jazz. Some of my favorite performances: Mas Que Nada (The Idea of North), Dreamer (Laufey), Words (The Real Group).
- Languages: Korean, English and Mandarin Chinese.