About
I am currently an M.S. student at the Kim Jaechul Graduate School of AI at KAIST, advised by Prof. Sung Ju Hwang. I received my B.E. in Computer Science and Engineering from Korea University.
My research focuses on developing more capable, efficient, and adaptive large language models and LLM-based systems. I am particularly interested in how language models and agents can reason, act, and accumulate knowledge efficiently over long horizons, while minimizing unnecessary computation, token usage, and external retrieval. My recent interests include token-efficient LLM agents, memory and knowledge integration, and parametric mechanisms for efficient adaptation.
Education
News
- Jun. 2026 — Released MemRefine, a framework for storage-budgeted long-term agent memory compression.
- Dec. 2025 — Joined the MLAI Lab at KAIST as a research intern.
Publications
-
MemRefine: LLM-Guided Compression for Long-Term Agent Memory
Minjae Kim, Jinheon Baek, Soyeong Jeong, and Sung Ju Hwang
arXiv preprint, 2026 -
Enhancing Multi-Hop Complex Query Retrieval Efficiency through the Integration of RAG and Graph RAG
Seong-min Lee, He-eun Kim, Dong-young Lee, Min-seon Kang, Minjae Kim, Soo-yeol Yang, and Young-sook Hwang
Annual Conference on Human and Cognitive Language Technology (HCLT), 2025
Research Experience
Honors & Awards
-
Top Excellence Award (2026)
2026 Spring Individual Research and Capstone Design Competition, College of Informatics, Korea University -
Top Excellence Award / President of IITP Award (2025)
2025 SW-Centered University Outstanding Works Competition
Awarded by the Institute for Information & Communications Technology Planning & Evaluation (IITP) -
Top Excellence Award (2024)
2024 Re:ALThon, Korea University SW-Centered University Hackathon