Minu Kim

I am a Ph.D. student at OSI Lab, KAIST AI.

As a theory-based practitioner, I approach controllability and interpretability of LLM agents as fundamentally mathematical concepts—properties that should be precisely defined and provable rather than intuitive or heuristic. I am particularly interested in understanding how we can formalize and guarantee the controllability of LLM agents, especially in reinforcement learning settings, where behavior emerges through interaction. Controllability, in this view, becomes a prerequisite for reliable preference alignment and meaningful personalization. Interpretability, likewise, provides the lens through which these mechanisms can be inspected, understood, and trusted. My research aims to develop principled foundations that bring these ideas together in the context of LLM agents.

Publications

Preference Alignment with Flow Matching
Minu Kim, Yongsik Lee, Sehyeok Kang, Jihwan Oh, Song Chong, Se-Young Yun
NeurIPS 2024 (Poster)

DPM: Dual Preference-based Multi-Agent Reinforcement Learning
Sehyeok Kang, Yongsik Lee, Minu Kim, Jihwan Oh, Song Chong, Se-Young Yun
(arXiv preprint)

Guiding Reasoning in Small Language Models with LLM Assistance
Yujin Kim, Euiin Yi, Minu Kim, Se-Young Yun, Taehyeon Kim
(arXiv preprint)

Hierarchical Decomposition Framework for Steiner Tree Packing Problem
Hanbum Ko, Minu Kim, Han-Seul Jeong, Sunghoon Hong, Deunsol Yoon, Youngjoon Park, Woohyung Lim, Honglak Lee, Moontae Lee, Kanghoon Lee, Sungbin Lim, Sungryull Sohn
ICORES 2025 (Best Student Paper Award)

Hierarchical Decomposition Framework for Feasibility-hard Combinatorial Optimization
Hanbum Ko, Minu Kim, Han-Seul Jeong, Sunghoon Hong, Deunsol Yoon, Youngjoon Park, Woohyung Lim, Honglak Lee, Moontae Lee, Kanghoon Lee, Sungbin Lim, Sungryull Sohn
ICML 2023 Workshop on Sampling and Optimization in Discrete Space

ReSPack: A Large-Scale Rectilinear Steiner Tree Packing Data Generator and Benchmark
Kanghoon Lee, Youngjoon Park, Han-Seul Jeong, Sunghoon Hong, Deunsol Yoon, Sungryull Sohn, Minu Kim, Hanbum Ko, Moontae Lee, Honglak Lee, Kyunghoon Kim, Euihyuk Kim, Seonggeon Cho, Jaesang Min, Woohyung Lim
NeurIPS 2022 Workshop on Synthetic Data for Empowering ML Research

Education

Bachelor of Science
(2018.03. ~ 2023.02.)
Korea Advanced Institute of Science and Technology (KAIST)
Dept. of Mathematical Sciences

MS/Ph.D Integrated
(2023.03. ~ Current)
Korea Advanced Institute of Science and Technology (KAIST)
Kim Jaechul Graduate School of AI

Research Experiences

Research Intern at Fundamental Research Lab, LG AI Research
(2022.03. ~ 2022.08.)

Honors & Awards

ICORES 2025 Best Student Paper Award
Hierarchical Decomposition Framework for Steiner Tree Packing Problem
Hanbum Ko, Minu Kim, Han-Seul Jeong, Sunghoon Hong, Deunsol Yoon, Youngjoon Park, Woohyung Lim, Honglak Lee, Moontae Lee, Kanghoon Lee, Sungbin Lim, Sungryull Sohn

NeurIPS 2021 Competition - Machine Learning for Combinatorial Optimization
Overall 4th place (dual task), 2nd in student leaderboard (dual task)

Talent Award of Korea 2016
(대한민국 인재상 2016)

Zayed Future Energy Prize 2016
Global High School Winner ($100,000 USD)

Scholarships & Memberships

Alumni Academic Scholarship, March 2019 - December 2021 ($10,000 USD) Korea Advanced Institute of Science and Technology (Daejeon, Republic of Korea) Student Council of the Department of Mathematical Sciences (2019 - 2021) Korea Advanced Institute of Science and Technology (Daejeon, Republic of Korea) President of Student Council, 2020 Chairman of Emergency Committee of Student Council, Winter 2020

Teaching & Mentoring Experiences

Lecturer at Chungcheongbuk-do Education Research and Information Institute Gifted Education Center, Korea (Government Institute) (2024 ~ Current)
Reinforcement Learning & Generative AI

KAIST Major Course Student Tutor (2019 Fall)
Engineering Statistics II (IE341)

KAIST Basic Elective Course Student Lecturer (2020 Spring, 2021 Spring)
Introduction to Linear Algebra (MAS109)

Page updated

Google Sites

Report abuse