Yumin Choi

About

I am an M.S. student in Artificial Intelligence at KAIST, where I work with Prof. Sung Ju Hwang and his research group, MLAI Lab. Before joining KAIST, I received my B.S. in Mechanical Engineering and Artificial Intelligence from Korea University in 2025. During my undergraduate studies, I worked on research projects involving MLLMs at MLV Lab with Prof. Hyunwoo J. Kim, and studied machine learning through Data Science & AI Society (KUBIG) under the guidance of Prof. Sungbin Lim.

My research is broadly about Self-Evolving AI: systems that continuously improve by accumulating, organizing, and reusing knowledge from external sources and their own experience. One way I approach this is through LLM-driven optimization, where LLMs generate high-quality solutions and prioritize promising candidates before committing to expensive evaluations. My recent work studies experience-driven improvement for long-horizon LLM agents, including pre-task memory construction and hindsight self-distillation from validated trajectories.

Publications

EvolveTrade: Experience-Driven Policy Refinement for Self-Evolving LLM Trading Agents

Sehee Kim*, Yumin Choi*, Minki Kang, Sung Ju Hwang

arXiv, 2026
It Takes Two: Complementary Self-Distillation for Contextual Integrity in LLMs

Sangwoo Park*, Woongyeong Yeo*, Seanie Lee, Yumin Choi, Hyomin Lee, Kangsan Kim, Jinheon Baek^†, Seong Joon Oh^†, Sung Ju Hwang^†

Conference on Neural Information Processing Systems (NeurIPS), 2026
PREPING: Building Agent Memory without Tasks

Yumin Choi, Sangwoo Park, Minki Kang, Jinheon Baek^†, Sung Ju Hwang^†

Conference on Neural Information Processing Systems (NeurIPS), 2026 · Project page
THINKSAFE: Self-Generated Safety Alignment for Reasoning Models

Seanie Lee*, Sangwoo Park*, Yumin Choi, Gyeongman Kim, Minki Kang, Jihun Yun, Dongmin Park, Jongho Park, Sung Ju Hwang

Conference on Neural Information Processing Systems (NeurIPS), 2026
T-MAP: Red-Teaming LLM Agents with Trajectory-aware Evolutionary Search

Hyomin Lee, Sangwoo Park, Yumin Choi, Sohyun An, Seanie Lee, Sung Ju Hwang

Conference on Empirical Methods in Natural Language Processing (EMNLP Main, Oral), 2026
HINT-SD: Targeted Hindsight Self-Distillation for Long-Horizon Agents

Woongyeng Yeo*, Yumin Choi*, Taekyung Ki, Sung Ju Hwang

Findings of Empirical Methods in Natural Language Processing (EMNLP Findings, Short), 2026
Super Library Agent: Joint Generation and Maintenance of Multiple Applications Beyond the Single Codebase

Daegyu Sung*, Yukyeong Lee*, Geon Park, Yumin Choi, Sung Ju Hwang

Findings of Empirical Methods in Natural Language Processing (EMNLP Findings), 2026
Multimodal Prompt Optimization: Why Not Leverage Multiple Modalities for MLLMs

Yumin Choi*, Dongki Kim*, Jinheon Baek^†, Sung Ju Hwang^†

International Conference on Learning Representations (ICLR), 2026
System Prompt Optimization with Meta-Learning

Yumin Choi*, Jinheon Baek*, Sung Ju Hwang

Conference on Neural Information Processing Systems (NeurIPS), 2025

*: equal contribution; †: equal advising.

Experience

Jun. 2024 - Present

KAIST MLAI Lab

Research Assistant · Advisor: Prof. Sung Ju Hwang

Nov. 2023 - May 2024

Korea University MLV Lab

Undergraduate Research Assistant · Advisor: Prof. Hyunwoo J. Kim

Jul. 2023 - Jun. 2024

Korea University Data Science & AI Society KUBIG

Project Team Leader · Advised by Prof. Sungbin Lim

Aug. 2022 - Jan. 2024

Korea University Anam Hospital

Research Assistant · Clinical ML with national insurance data, focusing on transfusion outcomes and ML model interpretation.

Education

Feb. 2027 expected

KAIST

M.S. in Artificial Intelligence

Feb. 2025

Korea University

B.S. in Mechanical Engineering & Artificial Intelligence