Minseo Koo
M.S. Student @ GIST DS Lab
Generalizing LLM reasoning beyond ARC
I am an M.S. student at the DS Lab , Gwangju Institute of Science and Technology (GIST), advised by Prof. Sundong Kim.
I’m currently working on using LLMs to solve abstract reasoning tasks such as the ARC-AGI benchmark, with a focus on reasoning strategies that generalize rather than overfit to a single benchmark. Previously, I worked on platform-invariant topic modeling and small language models for dialogue and text classification (see Publications ).
Research Interests
Generalizable Reasoning with LLMs
Using large language models to solve abstract reasoning tasks like the ARC (Abstraction and Reasoning Corpus) benchmark - and, more broadly, understanding what makes a reasoning strategy generalize rather than overfit to one benchmark.
Text Mining & Dialogue Systems
Making text mining and NLU systems robust across data sources - from platform-invariant topic modeling to intent/utterance classification.
Publications
- Platform-Invariant Topic Modeling via Contrastive Learning to Mitigate Platform-Induced Bias Findings of the Association for Computational Linguistics: EMNLP 2024 · 2024
News
CV
Education
Awards
- KSC 2023 Encouragement Award (Undergraduate) — Research to Mitigate Platform-Induced Topic Modeling Bias (2024)
- SW Festival 2023 Excellent Project Award — AI Mental Care Chatbot (2023)