Minsik Choi 최민식
Integrated B.S./M.S. student in Computer Science and Engineering at Korea University, advised by Prof. Young Geun Kim. Currently an AI research engineering intern on the Algorithm team at FuriosaAI. Previously at NAVER Cloud, where I worked on the HyperCLOVA X models.
I study how post-training and adaptation change large language models, and use that understanding to make them cheaper to train and to run — merging experts post-trained in parallel into one model, and pruning and hardware-aware optimization for inference on AI accelerators.
Research Interests
- Model merging for parallel post-training of LLMs — combining experts trained separately via SFT or RLVR
- Understanding how post-training and adaptation reshape models, from spectral structure in weight space to attention behavior
- Efficient transformer inference: pruning and hardware-aware optimization for AI accelerators
Publications
* Equal contribution † Corresponding author
-
[C1]
Merging RLVR-Trained Experts via Policy-Shift-Guided Spectral Alignment
-
[C2]
Text Capability Loss in Vision-Language Adaptation: An Attention-Sink Diagnosis
-
[C3]
Extended version of the NeurIPS 2025 Workshop paper below.
-
[C4]
Decentralized Instruction Tuning: Conflict-Aware Splitting and Weight Merging
Workshops
Technical Reports
Writing
Experience
-
AI Research Engineering Intern
Sep 2026 – PresentAlgorithm Team, FuriosaAI, Seoul
-
Research & Engineering Intern
Sep 2025 – Mar 2026NAVER Cloud, Seongnam
Worked on Korea's government-funded sovereign foundation model initiative (독자 파운데이션 모델 프로젝트), building LLM training and evaluation pipelines for instruction tuning and model merging.
-
Undergraduate Research Intern
Feb 2025 – Feb 2026Intelligent Computer Architecture & Systems Lab, Korea University
Hardware-aware transformer optimization and structured pruning for on-device NLP inference.
Service
-
Reviewer
2026OPT 2026: Optimization for Machine Learning, NeurIPS 2026 Workshop
Education
-
Integrated B.S./M.S., Computer Science and Engineering
Sep 2025 – Aug 2027Korea University · Advisor: Prof. Young Geun Kim
-
Exchange Student, Computer Science
Sep 2023 – Feb 2024University of Tübingen, Germany
-
B.S., Computer Science and Engineering
Mar 2022 – Feb 2026Korea University · Admitted via the School of Interdisciplinary Studies · GPA 4.2 / 4.5
Scholarships
-
Graduate Merit-Based Scholarship
Feb 2026 – Aug 2027Korea University · 50% tuition, all semesters
-
Academic Scholarship
2025Woonhae Scholarship Foundation · full tuition for one academic year