I am an undergraduate student majoring in Biomedical Engineering and double majoring in Mathematics at Korea University. I previously conducted research as an intern at the Multimodal Interactive Intelligence Laboratory (MIIL), under the supervision of Prof. Paul Hongsuck Seo. My research interests lie in multimodal generative models, especially diffusion-based approaches for cross-modal alignment, synthesis, and understanding. I am currently focused on generative modeling in the video domain, exploring how temporal dynamics and multimodal signals can be jointly modeled within modern deep learning frameworks. Publications
DGMO: Training-Free Audio Source Separation through Diffusion-Guided Mask Optimization
Geonyoung Lee*, Geonhee Han*, and Paul Hongsuck Seo (* Equal contribution)
Interspeech 2025 oral
Geonyoung Lee*, Geonhee Han*, and Paul Hongsuck Seo (* Equal contribution)
Interspeech 2025 oral
Research Experiences
MIIL Lab, Korea University — Undergraduate Research Intern
Advisor: Prof. Paul Hongsuck Seo
Jul. 2024 – Jun. 2025Advisor: Prof. Paul Hongsuck Seo
Education
Korea UniversityB.S. in Biomedical Engineering, College of Health Science
Korea UniversityB.S. in Mathematics, College of Science
Academic Activities
KUBIG AI Society — Generative Models & CV Track Member, Administrator
Jul. 2024 – Jul. 2025BEAT Society — AI Track Member
Mar. 2023 – Jul. 2023B.I.O Union Society (Yonsei × Korea Univ.)
Sep. 2022 – Dec. 2022Research Interests
- Diffusion Models for Audio and Vision
- Multimodal Learning (Text-Audio-Visual Alignment)
- Reinforcement Learning for Robotics
- Multimodal Learning (Text-Audio-Visual Alignment)
- Reinforcement Learning for Robotics