Minseok Kim

Ph.D. Candidate @ VILAB, KAIST

minseok_profile_2.jpg

📧 alstjrx1x1@kaist.ac.kr

🐙 github.com/Minseokkim-0124/

📍 291 Daehak-ro, Yuseong-gu, Daejeon 34141

Hello! 👋 I am a 2nd-year Ph.D. student at VILAB, advised by Prof. Kuk-Jin Yoon in the Department of Mechanical Engineering at KAIST.

🔬 Research Interests. My research spans robust 3D perception (LiDAR / Radar) and data-efficient learning (Category Discovery & Dataset Distillation). Recently, my focus has shifted toward understanding human intention in egocentric scenarios — exploring how AI systems can interpret first-person videos to advance next-generation AR / VR and human-robot interaction technologies.

💡 Current Directions.

  • Egocentric video understanding (gaze, hand, intention)
  • Vision-Language Models (VLM) for scene understanding
  • Generalized / Open-world 3D perception

📬 I am always happy to discuss ideas and collaborate on egocentric vision, VLMs, or 3D perception. Feel free to reach out via email!

🎸 Outside of research, I enjoy playing electric guitar, photography, and working out.

News

May 13, 2026 Personal webpage is now live!

Education

2025.03 – now · Ph.D. candidate, Mechanical Engineering, KAIST

2023.03 – 2025.02 · M.S., Mechanical Engineering, KAIST

2017.03 – 2023.02 · B.S., Mechanical Engineering, Hanyang University

Publications

  1. Preprint
    Score Locally, Solve Globally: Training-free Set Decoding for Vision-Language Models
    Minseok Kim, Jaewoo Jeong, and Kuk-Jin Yoon
    2027
    Preprint
  2. Preprint
    Solving LiDAR Inverse Problems using LiDAR Diffusion Models
    Changgyoon Oh*, Minseok Kim*, Hoyong Kwon*, Seonkyu Song, and Kuk-Jin Yoon
    2026
    Preprint
    Video
  3. Preprint
    PointDPS: Diffusion Posterior Sampling for Point Inverse Problems
    Changgyoon Oh*, Seonkyu Song*, Yuhwan Jeong*, Minseok Kim, and Kuk-Jin Yoon
    2026
    Preprint
    Video
  4. Preprint
    Zero-Shot Coordination under Partial Observability by Predicting Intention of Partner
    Yuhwan Jeong*, Hoyong Kwon*, Jinnyeong Yang*, Minseok Kim, Jihun Kim, and Kuk-Jin Yoon
    2026
    Preprint
    Video
  5. Preprint
    Gaze-hand Trajectory Merging for Efficient Egocentric Video Understanding
    Jaewoo Jeong*, Minseok Kim*, Yujin Bae*, Sung-Hoon Yoon, and Kuk-Jin Yoon
    2026
    Preprint
  6. CVPR
    MMDD.png
    Multimodal Distribution Matching for Vision-Language Dataset Distillation
    Jongoh Jeong*, Hoyong Kwon*, Minseok Kim*, and Kuk-Jin Yoon
    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2026
    Video
  7. WACV
    GCDLSS.png
    Generalized Category Discovery for LiDAR Semantic Segmentation
    Minseok Kim, Jiyong Boo, and Kuk-Jin Yoon
    In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2026
  8. ECCV
    OCDSS.png
    On-the-fly Category Discovery for LiDAR Semantic Segmentation
    Hyeonseong Kim*, Sung-Hoon Yoon*, Minseok Kim, and Kuk-Jin Yoon
    In European Conference on Computer Vision, 2024
  9. ECCV
    RL3DOD.png
    LiDAR-based All-weather 3D Object Detection via Prompting and Distilling 4D Radar
    Yujeong Chae, Hyeonseong Kim, Changgyoon Oh, Minseok Kim, and Kuk-Jin Yoon
    In European Conference on Computer Vision, 2024

Projects

  • Development of Mobility and Connectivity Platform for Unmanned Autonomous Delivery
  • Development of quadrupedal robot system technology for surveillance, reconnaissance, and search missions