GiHoon Kim

GiHoon Kim

Ph.D. student in diffusion & multimodal generation.

About

I am a Ph.D. student in Data Science at Seoul National University, advised by Prof. Taesup Kim (Learning Adaptation Algorithm Lab). My research broadly studies diffusion models—their learning and inference mechanisms, guidance/control behaviors, and how these translate into practical capabilities for generative modeling. Within this direction, I am particularly interested in personalized diffusion, aiming to maintain identity/attribute consistency while preventing distributional drift during adaptation, as well as multimodal generation (text-to-image, video).

Education

  • Seoul National University, Ph.D. in Data Science, Mar. 2024 - Present
    Learning Adaptation Algorithm Lab, Advisor: Prof. Taesup Kim
  • KAIST, M.S. in Culture Technology, Sep. 2021 - Dec. 2023
    Visual Media Lab, Advisor: Prof. Junyong Noh
  • Handong Global University, B.S. in Computer Science & Electronic Engineering, Mar. 2015 - Aug. 2021

Research Interests

  • Personalized Diffusion
    • distribution drift prevention
    • identity consistency
    • semantic anchoring for robust personalization
  • Multimodal Generative Modeling
    • text-to-image generation
    • video generation
  • Controllable Generation
    • guidance and control for targeted edits/attributes
    • diffusion for inverse problems

If you would like to discuss research or potential collaboration, feel free to reach out: gihoon.kim@snu.ac.kr.

News

  • First-author paper accepted at ICLR 2026: Preserve and Personalize: Personalized Text-to-Image Diffusion Models without Distributional Drift
  • Open-source research implementations (one repository reached 65 GitHub stars)