Dongwon Kim

POSTECH Computer vision lab. kdwon@postech.ac.kr

prof_pic.jpeg

I am a postdoctoral researcher at KAIST, working with Prof. Jeany Son. I previously completed my BS and PhD at POSTECH Computer Vision Lab where I worked with Prof. Suha Kwak. My research centers on whether machines can learn representations with the right abstraction and hierarchy, spanning work in compositional representation (SelfMod), multi-modalities (DivE, SaG, MaskGen), and world model (ongoing, CompACT).

News

Feb 2026 A paper on world model and decision-time planning is accepted at CVPR 2026 (early version appeared in CoRL-LSRW workshop).
Feb 2026 I won POSTECH CSE Best Research Award 2025, recognizing the best research among PhD graduates.
Jul 2025 A paper on 1-dimensinal tokenization and text-to-image generation is accepted at ICCV 2025.
Jul 2025 I have completed my defense - now I’m officially a Ph.D. (thesis title: Learning Compositional Visual Representations for Vision-Language Understanding and Generation)
Sep 2024 A paper about object-centric learning is accepted at NeurIPS 2024.

Education

Mar, 2015 - Aug, 2019 B.S. in Computer Science & Engineering
POSTECH, Pohang, South Korea
Sep, 2019 - Aug, 2025 Integrated M.S & Ph.D in Computer Science & Engineering
POSTECH, Pohang, South Korea
Advisor: Prof. Suha Kwak


Experience

Nov, 2025 - Present Postdoctoral researcher
KAIST, Daejeon, KR
Jun, 2024 - Nov, 2024 Research Intern
Fundamental Research Team, ByteDance SEED, San Jose, US
  • Developed efficient text-to-image generative model using compact text-aware 1D tokens
  • First-authored paper “Democratizing Text-to-Image Masked Generative Models…” (accepted to ICCV 2025)


Publications

  1. 16world.png
    Planning in 8 Tokens: A Compact Discrete Tokenizer for Latent World Model
    Dongwon Kim, Gawon Seo, Jinsung Lee, Minsu Cho, and Suha Kwak
    In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun 2026
    (An early version appeared in LSRW workshop in CoRL 2025)
  2. maskgen.png
    Democratizing Text-to-Image Masked Generative Models with Compact Text-Aware One-Dimensional Tokens
    Dongwon Kim*, Ju He*, Qihang Yu*, Chenglin Yang, Xiaohui Shen, Suha Kwak, and Liang-Chieh Chen
    In IEEE/CVF International Conference on Computer Vision (ICCV), Jan 2025
    arXiv PDF Code
  3. 158flux.png
    1.58-bit FLUX
    Chenglin Yang, Celong Liu, Xueqing Deng, Dongwon Kim, Xing Mei, Xiaohui Shen, and Liang-Chieh Chen
    In arXiv preprint, Dec 2024
    arXiv
  4. selfmod.png
    Bootstrapping Top-down Information for Self-modulating Slot Attention
    Dongwon Kim, Seoyeon Kim, and Suha Kwak
    In Neural Information Processing Systems (NeurIPS), Dec 2024
    arXiv
  5. plot.png
    PLOT: Text-based Person Search with Part Slot Attention for Corresponding Part Discovery
    Jicheol Park, Dongwon Kim, Boseung Jeong, and Suha Kwak
    In European Conference on Computer Vision (ECCV), Oct 2024
  6. risclip.jpg
    Extending CLIP’s Image-Text Alignment to Referring Image Segmentation
    Seoyeon Kim, Minguk Kang, Dongwon Kim, Jaesik Park, and Suha Kwak
    In Annual Conference on the North American Chapter of the Association for Computational Linguistics (NAACL), Jun 2024
    Code
  7. sag.png
    Shatter and Gather: Learning Referring Image Segmentation with Text Supervision
    Dongwon Kim*, Namyup Kim*, Cuiling Lan, and Suha Kwak
    In IEEE/CVF International Conference on Computer Vision (ICCV), Oct 2023
    arXiv Code Website
  8. dive.png
    Improving Cross-Modal Retrieval With Set of Diverse Embeddings
    Dongwon Kim, Namyup Kim, and Suha Kwak
    In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun 2023
    (Highlight, 235/9155 = 2.5%)
    arXiv Code Website
  9. restr.png
    ReSTR: Convolution-Free Referring Image Segmentation Using Transformers
    Namyup Kim, Dongwon Kim, Cuiling Lan, Wenjun Zeng, and Suha Kwak
    In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun 2022
    arXiv Website
  10. stml.png
    Self-Taught Metric Learning Without Labels
    Sungyeon Kim, Dongwon Kim, Minsu Cho, and Suha Kwak
    In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun 2022
    arXiv Code Website
  11. et.png
    Embedding Transfer With Label Relaxation for Improved Metric Learning
    Sungyeon Kim, Dongwon Kim, Minsu Cho, and Suha Kwak
    In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun 2021
    arXiv Website
  12. pa.gif
    Proxy Anchor Loss for Deep Metric Learning
    Sungyeon Kim, Dongwon Kim, Minsu Cho, and Suha Kwak
    In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun 2020
    arXiv Code Website

Professional Services

Reviewer
  • IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)
  • IEEE/CVF International Conference on Computer Vision (ICCV), 2023
  • IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022-2023
  • European Conference on Computer Vision (ECCV), 2022
  • Winter Conference on Applications of Computer Vision (WACV), 2023
  • Asian Conference on Computer Vision (ACCV), 2022

Honors and Awards

POSTECH CSE Best Research Award, POSTECH, 2025 POSTECHIAN Fellowship, POSTECH, 2023
  • $5,000 grant
BK21 Best Paper Award, POSTECH GSAI, 2023
  • Self-Taught Metric Learning without Labels (CVPR 2022)
Qualcomm Innovation Fellowship Winner, Qualcomm Korea Corp., 2022
  • Self-Taught Metric Learning without Labels (CVPR 2022)
  • ReSTR: Convolution?free Referring Image Segmentation Using Transformers (CVPR 2022)
NAVER x POSTECH AI DAY The 2nd and 3rd Prize, 2022
  • ReSTR: Convolution-free Referring Image Segmentation Using Transformers (CVPR 2022)
Qualcomm Innovation Fellowship Winner, Qualcomm Korea Corp., 2021
  • Embedding Transfer with Label Relaxation for Improved Metric Learning (CVPR 2021)
IPIU Best Paper Award, 2021
  • Embedding Transfer with Label Relaxation for Improved Metric Learning (CVPR 2021)