I am a Ph.D. student in the SIML lab at KAIST AI, advised by Prof. Juho Lee.
My research has two directions.
- Building efficient AI systems, with a focus on language-model agents. Agents today rely on ever-larger models. I aim to build small agents that adaptively select which capabilities of large models to retain. My research spans pruning, knowledge distillation, and agent harness. (arXiv’26, NeurIPS’26, ICML’24, ICLR’24, ICMLW’23, NeurIPSW’22)
- Building reliable AI systems. Capability alone does not make a model trustworthy. I also aim to build models that are safe and well-calibrated. (ICML’26, ICMLW’25, ECCV’24, arXiv’24, ICMLW’23, arXiv’22)
I am on the job market for positions starting in Spring 2027. I have one year of alternative military service (Technical Research Personnel) remaining.
Education
KAIST
Ph.D. in Artificial Intelligence, advised by Prof. Juho Lee
Mar. 2021 – present
Seoul National University
M.S. in Industrial Engineering, advised by Prof. Deokjoo Lee
Mar. 2019 – Feb. 2021
Thesis: Measuring the Economic Value of Data in Machine Learning: A Cooperative Game Approach
Seoul National University
B.S. in Industrial Engineering, cum laude
Mar. 2014 – Feb. 2019
Publications
Harness-Aware Distillation for Small Language Model Agents
Moonseok Choi, Taehong Moon, Giung Nam†, Juho Lee†
Paper (soon)Project page
arXiv 2026
SKIM: Pruning Large Language Model Agents via Selective Knowledge Informed Masking
Moonseok Choi, Jongwon Jeong, Minki Kang, Giung Nam†, Juho Lee†
Paper (soon)Project page (soon)
NeurIPS 2026
Confidence is not universal: task-dependent calibration and emergent behavior in LLMs
Chaeyun Jang, Moonseok Choi, Yegon Kim, Seungyoo Lee, Juho Lee†, Hyungi Lee†
ICML 2026
Test Time Scaling for Neural Processes
Hyungi Lee, Moonseok Choi, Hyunsu Kim, Kyunghyun Cho, Rajesh Ranganath, Juho Lee
NeurIPS 2025
Projection-Based Augmentation with Non-Relevant General Data for Enhanced Domain Adaptation in LLMs
Seungyoo Lee, Giung Nam, Moonseok Choi, Hyungi Lee†, Juho Lee†
NeurIPS 2025
Verbalized Confidence Triggers Self-Verification: Emergent Behavior Without Explicit Reasoning Supervision
Chaeyun Jang, Moonseok Choi, Yegon Kim, Hyungi Lee, Juho Lee
Workshop on Reliable and Responsible Foundation Models
ICML 2025 Workshop
Safety Alignment Backfires: Preventing Re-emergence of Suppressed Concepts in Fine-tuned Text-to-Image Diffusion Models
Sanghyun Kim, Moonseok Choi, Jinwoo Shin, Juho Lee
arXiv 2024
Safeguard Text-to-Image Diffusion Models with Human Feedback Inversion
Sanghyun Kim, Seohyeon Jung, Balhae Kim, Moonseok Choi, Jinwoo Shin, Juho Lee
ECCV 2024
A Simple Early Exiting Framework for Accelerated Sampling in Diffusion Models
Taehong Moon, Moonseok Choi, EungGu Yun, Jongmin Yoon, Gayoung Lee, Jaewoong Cho, Juho Lee
ICML 2024
Sparse Weight Averaging with Multiple Particles for Iterative Magnitude Pruning
Moonseok Choi*, Hyungi Lee*, Giung Nam*, Juho Lee
ICLR 2024
Towards Safe Self-Distillation of Internet-Scale Text-to-Image Diffusion Models
Sanghyun Kim, Seohyeon Jung, Balhae Kim, Moonseok Choi, Jinwoo Shin, Juho Lee
Workshop on Deployment Challenges for Generative AI
ICML 2023 Workshop
Early Exiting for Accelerated Inference in Diffusion Models
Taehong Moon, Moonseok Choi, Eunggu Yun, Jongmin Yoon, Gayoung Lee, Juho Lee
Workshop on Structured Probabilistic Inference & Generative Modeling
ICML 2023 Workshop
Fine-tuning Diffusion Models with Limited Data
Taehong Moon, Moonseok Choi, Gayoung Lee, Jung-Woo Ha, Juho Lee
Workshop on Score-Based Methods
NeurIPS 2022 Workshop
arXiv 2022
Diffusion Forecasting of Electric Vehicles Based on Consumers’ Preferences: A Korean case
Moonseok Choi, Deokjoo Lee, Jinsu Shin, Piao Ri
Proceedings of the 4th IAEE Eurasian Conference
IAEE Eurasian 2019
Selecting Optimal Locations of Public Charging Stations for Electric Vehicles Using the Big Data of Driving Behaviors: A Case Study of Seoul, Korea
Jihyeok Jung*, Moonseok Choi*, Deokjoo Lee
Proceedings of the 42nd IAEE International Conference
IAEE 2019
* equal contribution, † co-corresponding
Experience
KRAFTON AI
Research intern, core research team
Dec. 2024 – Feb. 2025
Academic services
Conference reviewer
NeurIPS (2022–26), ICLR (2024–26), ICML (2025–26), AAAI (2025–27)