Joowon Kim
I am a Ph.D. student at the Machine Learning and Intelligence Lab (MLILAB) at KAIST, advised by Prof. Eunho Yang. I am also a research intern on the Omni Foundation Model team at NAVER Cloud.
My research interests lie in multimodal learning, with a particular focus on generative models and vision–language models (VLMs). Throughout my work, I favor training-free methods that improve both the performance and the inference-time efficiency of existing models, for instance by routing computation only to where it is needed or by ensembling complementary models. My earlier work developed test-time scaling methods for diffusion and visual auto-regressive models. I then moved to video reasoning, using video generation models as world simulators that let VLMs reason about how a scene unfolds.
More broadly, I am interested in how models represent and simulate the visual world. Recently, I have focused on vision tokenizers, the representation every multimodal model builds on, and on resolving their limitations at the pretraining stage. Looking ahead, I aim to move from representing the world to predicting and acting in it, through world models and world action models for physical AI.
News
- Sep 2026CollabVR is accepted to NeurIPS 2026!
- Jul 2026Joined the Omni Foundation Model team at NAVER Cloud as a research intern.
- Feb 2026Progress by Pieces is accepted to CVPR 2026.
- Sep 2025Started my Ph.D. in Artificial Intelligence at KAIST.
- Jun 2025ELECT is accepted to ICCV 2025.
- Sep 2024Joined the Image Vision team at NAVER Cloud as an intern.
Older news
- Jan 2024Language-Interfaced Tabular Oversampling is accepted to ICLR 2024.
- Sep 2023Started my M.S. in Artificial Intelligence at KAIST.
- Aug 2023Graduated from KAIST with a B.S. in School of Computing.
Publications
* equal contribution
Conference & Journal
Language-Interfaced Tabular Oversampling via Progressive Imputation and Self-Authentication
June Yong Yang*, Geondo Park*, Joowon Kim, Hyeongwon Jang, Eunho Yang
ReviewScore: Misinformed Peer Review Detection with Large Language Models
Hyun Ryu, Doohyuk Jang, Hyemin S. Lee, Joonhyun Jeong, Gyeongman Kim, Donghyeon Cho, Gyouk Chu, Minyeong Hwang, Hyeongwon Jang, Changhun Kim, Haechan Kim, Jina Kim, Joowon Kim, Yoonjeon Kim, Kwanhyung Lee, Chanjae Park, Heecheol Yun, Gregor Betz, Eunho Yang
Prediction Model for the Early Detection of Cachexia in Patients with Pancreatic Adenocarcinoma
H. Jeong, Y. Lee*, Joowon Kim*, Y. Ko, J. Jo, H. Lee, K. Kim, J. Park, S. Bang, S. Park, S. Song, M. Chung
Gastroenterology 162(7): S-705, Digestive Disease Week (DDW) abstract, 2022
Preprints & Under Review
Look Only When Needed: Certificate-Guided Visual Routing for Efficient Diffusion VLMs
Donghyeon Cho*, Joowon Kim*, Seungho Shin, Eunho Yang
Who Should Lead Decoding Now? Tracking Reliable Trajectories for Ensembling Masked Diffusion Language Models
Heecheol Yun, Joonhyung Park, Joowon Kim, Eunho Yang
Experience
NAVER Cloud
- Vision tokenizer pretraining for omni foundation models
- Real-time (full-duplex) video understanding models
NAVER Cloud
- Supervised fine-tuning of a text-to-image based background inpainting model for advertising
- Performance optimization, evaluation metric development, and scaling to SD 3.0 and DiT-based models
Machine Learning & Intelligence Lab (MLILAB), KAIST
- Multi-source warping-based talking-head video-to-video generation
- Counterfactual explanation of error cases with diffusion models
AITRICS
- Prediction model for the early detection of cachexia
- Counterfactual trajectory prediction on clinical time-series data
SK hynix
- Memory device operation characteristics analysis
Academic Services
- Conference Reviewer: ICLR 2027
Education
Korea Advanced Institute of Science and Technology (KAIST)
- Ph.D. in Artificial IntelligenceSep. 2025 - PresentAdvisor: Eunho Yang
- M.S. in Artificial IntelligenceSep. 2023 - Aug. 2025Advisor: Eunho Yang
- B.S. in School of ComputingMar. 2018 - Aug. 2023 (GPA 3.81/4.3)
Skills
- Languages
- English, Korean
- Programming Languages
- Python (PyTorch Libraries), C/C++, JavaScript, R, MATLAB, LaTeX
- Developer Tools
- CUDA, Bash
Extra Activities
- Leading an ML/DL Paper Reading Study 2021
- CS101 (Introduction to Programming) Teaching Assistant 2020 - 2023
- CS101 (Introduction to Programming) Freshman Tutoring Program Tutor 2019 - 2020
- KAIST Undergraduate Student Council (Executive, 2020) 2018 - 2022
- KAIST School of Computing Student Council (Executive, 2020) 2019 - 2020