Prof. Soobin Um - Kookmin University AI·SW Graduate School

Research Overview 연구 개요

Develops trustworthy and responsible AI systems with emphasis on generative models applied to scientific discovery and innovation. Recent work focuses on minority-focused generation using diffusion models and contrastive noise optimization.

생성 모델을 과학적 발견과 혁신에 적용하여 신뢰할 수 있고 책임감 있는 AI 시스템을 개발합니다. 최근 연구는 확산 모델을 활용한 소수 집단 중심 생성과 대조적 노이즈 최적화에 초점을 맞추고 있습니다.

Research Areas 연구 분야

Generative Modeling
Trustworthy AI
Diffusion Models
Scientific Discovery

생성 모델링
신뢰할 수 있는 AI
확산 모델
과학적 발견

Major Achievements 주요 연구 성과

Minority-Focused Text-to-Image Generation (CVPR 2025, Oral, Top 0.74%)
Diverse Text-to-Image Generation via Contrastive Noise Optimization (ICLR 2026)

Honors & Awards 수상

CVPR 2025 Oral (Top 0.74%)
AAAI 2023 Oral

Recent Publications 주요 논문

D. Seo, S. Um, S. Lee, J. C. Ye, and H. Chung, "Physics-Guided and Fabrication-Aware Inverse Design of Photonic Devices Using Diffusion Models," ACS Photonics, 2026
B. Kim, S. Um, and J. C. Ye, "Diverse Text-to-Image Generation via Contrastive Noise Optimization," ICLR, 2026
S. Um, B. Kim, and J. C. Ye, "Boost-and-Skip: A Simple Guidance-Free Diffusion for Minority Generation," ICML, 2025
S. Um and J. C. Ye, "Minority-Focused Text-to-Image Generation via Prompt Optimization," CVPR (Oral), 2025
S. Um and J. C. Ye, "Self-Guided Generation of Minority Samples Using Diffusion Models," ECCV, 2024
S. Um, S. Lee, and J. C. Ye, "Don't Play Favorites: Minority Guidance for Diffusion Models," ICLR, 2024
S. Um and C. Suh, "A Fair Generative Model Using LeCam Divergence," AAAI (Oral), 2023