Hyo Seo (Claudia) Kim

Ph.D. student

Department of Electrical and Computer Engineering

Illinois Institute of Technology

Hi! I'm a second-year Ph.D. student in Computer Engineering at Illinois Tech, advised by Prof. Ren Wang. My research focuses on reliable model adaptation, with interests in model editing, machine unlearning, model merging, continual learning, and adversarial robustness.

News

08/2026: Invited reviewer for IEEE Transactions on Neural Networks and Learning Systems (TNNLS).
05/2026: One paper accepted to ICML 2026.
04/2026: One paper accepted for Oral Presentation, ICLR 2026 Workshop on Principled Design for Trustworthy AI.
02/2026: Served as a reviewer for ICLR 2026 Workshop.
10/2025: Invited reviewer for Pattern Recognition Letters (Elsevier).

08/2025: Started my Ph.D. journey at Illinois Tech.
05/2025: One paper accepted to ICML 2025.
12/2024: One paper accepted to Pattern Recognition Letters.
12/2024: Served as an invited reviewer for CPAL 2025.
10/2024: One paper accepted to NeurIPS 2024 Workshop on Adaptive Foundation Models.
10/2024: Started visiting research at Michigan State University.
07/2024: Joined NAVER AI Lab as a Research Intern.
04/2024: One paper accepted to Pattern Recognition Letters.
07/2023: Visiting researcher at the Tübingen AI Center.
09/2022: Joined Sogang University as an M.S. student.

Publications

HyperFix: Combinatorial Nonlinear Correction for Task Vector Merging [paper]

arXiv preprint, 2026

HyperFix learns a shared correction function for model merging across the combinatorial space of task subsets.

Authors: Hyo Seo Kim, Ren Wang

MoCo-EA: Exploiting Adversarial Mode Connectivity for Efficient Evolutionary Attacks [paper] [code]

ICML 2026 (Poster) & ICLR 2026 Workshop (Oral)

MoCo-EA reveals that adversarial examples lie on connected manifolds rather than isolated points.

Authors: Hyo Seo Kim, Gang Luo, Can Chen, Binghui Wang, Yue Duan, Ren Wang

NegMerge: Sign-Consensual Weight Merging for Machine Unlearning [paper] [code]

ICML 2025 (Poster) & NeurIPS 2024 Workshop (Poster)

NegMerge introduces model merging into machine unlearning by combining sign-consistent task vectors.

Work done during an internship at NAVER AI Lab.

Authors: Hyo Seo Kim, Dongyoon Han*, Junsuk Choe* (* corresponding author)

Weakly-supervised Incremental learning for Semantic segmentation with Class Hierarchy [paper]

PRL 2024

Authors: Hyo Seo Kim, Junsuk Choe

Curriculum Learning with Class-label Composition for Weakly Supervised Semantic Segmentation [paper]

PRL 2025

Authors: Dongjun Hwang, Hyo Seo Kim, Doyeol Baek, Hyunbin Kim, Inhye Kye, Junsuk Choe

Education

Illinois Institute of Technology, Illinois, USA (Aug. 2025 - now)

Ph.D. in Computer Engineering
Advisor: Ren Wang

Sogang University, Seoul, South Korea (Aug. 2022 - Aug. 2025)

M.S. in Computer Science and Engineering
Advisor: Junsuk Choe

Philipps-University, Marburg, Germany (Mar. 2021 - Aug. 2021)

Exchange program

Sogang University, Seoul, South Korea (Feb. 2018 - Aug. 2022)

B.A. in Economics
B.S. in Computer Science and Engineering
Graduated with Cum laude

Work Experiences

Visiting Scholar, Michigan State University, Michigan, United States (Oct. 2024 - Apr. 2025, Host: Prof. Sijia Liu)
Research Intern, NAVER AI Lab, Seoul, South Korea (Jul. 2024 - Oct. 2024, Mentor: Dr. Dongyoon Han)
Visiting Scholar, Tübingen AI Center, Tübingen, Germany (Jul. 2023 - Aug. 2023, Host: Prof. Seong Joon Oh)

Web Developer Intern, SK C&C, Seoul, South Korea (Jul. 2022 - Aug. 2022)
ERP Developer Intern, LG CNS, Seoul, South Korea (May 2022 - Jun. 2022)
Undergraduate Intern, Sogang BaSE Lab, Seoul, South Korea (Aug. 2021 - Aug. 2022, Advisor: Prof. Sooyong Park)
QA Intern, LIKELION, Seoul, South Korea (Dec. 2020 - Mar. 2021)

Academic Services

Conference Volunteer: IEEE MASS 2025
Conference Reviewer: CPAL 2025, ICLR 2026 Workshop on Principled Design for Trustworthy AI
Journal Reviewer: Pattern Recognition Letters (Elsevier) 2025, IEEE TNNLS 2026

Page updated

Google Sites

Report abuse