I am PhD student at KAIST AI working with professor Jinwoo Shin. Previously, I interned at Google DeepMind working with Yinxiao Li and Feng Yang. I also closely worked with Kihyuk Sohn via Google University Relation program.

My research interests lie in probabilistic machine learning, generative modeling, and representation learning, with a focus on their applications to visual understanding and generation. Recently, I have been particularly interested in generative modeling for visual synthesis — including image, video, and 3D generation — as a step toward world modeling. I am also exploring applications in robotics, such as vision-language-action models. My previous works have focused on the fundamentals of training diffusion and flow models, as well as their adaptations to tasks such as 3D generation, personalization, and preference-based fine-tuning.

I plan to graduate in 2026 and am seeking industry research positions. Please feel free to reach out if you’re interested in potential collaborations or opportunities to work together.

Email: kyungmnlee (at) kaist (dot) ac (dot) kr | Curriculum Vitae

📝 Publications

• Preprint Decoupled MeanFlow: Turning Flow Models into Flow Maps for Accelerated Sampling, Kyungmin Lee, Sihyun Yu, Jinwoo Shin.

• Preprint Dual-Stream Diffusion for World-Model Augmented Vision-Language-Action Model, John Won, Kyungmin Lee, Huiwon Jang, Dongyoung Kim, Jinwoo Shin.

• Preprint Contrastive Representation Regularization for Vision-Language-Action Models, Taeyoung Kim, Jimin Lee, Myungkoo Koo, Dongyoung Kim, Kyungmin Lee, Changyeon Kim, Younggyo Seo, Jinwoo Shin.

• Preprint HAMLET: Switch your Vision-Language-Action Model into a History-Aware Policy, Myungkoo Koo, Daewon Choi, Taeyoung Kim, Kyungmin Lee, Changyeon Kim, Younggyo Seo, Jinwoo Shin. Project page

• Preprint Enhancing Motion Dynamics of Image-to-Video Models via Adaptive Low-Pass Guidance, Junesuk Choi, Kyungmin Lee, Sihyun Yu, Yisol Choi, Jinwoo Shin and Kimin Lee. Project page

• IJCAI 2025 StarFT: Robust Fine-tuning of Zero-shot Models via Spuriosity Alignment, Younghyun Kim*, Jongheon Jeong*, Sangkyung Kwak, Kyungmin Lee, Juho Lee and Jinwoo Shin

• CVPR 2025 Calibrated Multi-Preference Optimization for Aligning Diffusion Models, Kyungmin Lee, Xiaohang Li, Qifei Wang, Junfeng He, Junjie Ke, Ming-Hsuan Yang, Irfan Essa, Jinwoo Shin, Feng Yang and Yinxiao Li. Project page

• ICLR 2025 DiffusionGuard: A Robust Defense Against Malicious Diffusion-based Image Editing, Junesuk Choi, Kyungmin Lee, Jongheon Jeong, Saining Xie, Jinwoo Shin and Kimin Lee.

• NeurIPS 2024 Direct Consistency Optimization for Robust Customization of Text-to-Image Diffusion Models, Kyungmin Lee, Sangkyung Kwak, Kihyuk Sohn and Jinwoo Shin. Project page |

• ECCV 2024 Improving Diffusion Models for Authentic Virtual Try-on in the Wild, Yisol Choi, SangKyung Kwak, Kyungmin Lee, Hyungwon Choi and Jinwoo Shin. Project page | |

• CVPR 2024(Highlight) Discovering and Mitigating Visual Biases through Keyword Explanation, Younghyun Kim, Sangwoo Mo, Minkyu Kim, Kyungmin Lee, Jaeho Lee and Jinwoo Shin.

• ICLR 2024(Spotlight) DreamFlow: High-quality Text-to-3D generation by Approximating Probability Flow, Kyungmin Lee, Kihyuk Sohn and Jinwoo Shin. Project page

• NeurIPS 2023 Workshop Fine-tuning Protein Language Models by ranking protein fitness, Minji Lee, Kyungmin Lee and Jinwoo Shin.

• NeurIPS 2023Collaborative Score Distillation for Consistent Visual Editing, Subin Kim*, Kyungmin Lee*, Junesuk Choi, Jongheon Jeong, Kihyuk Sohn and Jinwoo Shin. Project page |

• NeurIPS 2023S-CLIP: Semi-supervised Vision-Language Pre-training using Few Specialist Captions, Sangwoo Mo, Minkyu Kim, Kyungmin Lee and Jinwoo Shin.

• NeurIPS 2023Slimmed Asymmetrical Contrastive Learning and Cross Distillation for Lightweight Model Training, Jian Meng, Li Yang, Kyungmin Lee, Jinwoo Shin, Deliang Fan and Jae-sun Seo.

• ICLR 2023(Spotlight) STUNT: Few-shot Tabular Learning with Self-generated Tasks from Unlabeled Tables, Jaehyun Nam, Jihoon Tack, Kyungmin Lee, Hankook Lee and Jinwoo Shin.

• NeurIPS 2022 RényiCL: Contrastive Representation Learning with Skew Rényi divergence, Kyungmin Lee and Jinwoo Shin.

• ECCV 2022 GCISG: Guided Causal Invariant Learning for Improved Syn-to-Real Generalization, Gilhyun Nam, Gyeongjae Choi and Kyungmin Lee.

• ICLR 2022 Representation Distillation by Prototypical Contrastive Predictive Coding, Kyungmin Lee

• IJCAI 2022 Pseudo-spherical Knowledge Distillation, Kyungmin Lee and Hyeongkeun Lee.

• ICLR 2021 Workshop Provable Defense by Denoised Smoothing with Learned Score function, Kyungmin Lee

💻 Work Experience

• 2024.07 - 2024.12, Google DeepMind, Student Researcher, Mountain View, CA, US.
• 2023.02 - 2024.03, Google Research, University Relation Program, Remote.
• 2019.04 - 2022.06, Agency for Defense Development, Researcher, Daejeon, Korea.

🤝 Academic Services

• Conference reviewer: NeurIPS, ICML, ICLR, CVPR, ICCV, ECCV, WACV, AISTATS
• Journal reviewer: TMLR, TPAMI

📖 Educations

• 2022.09 - 2026.08, KAIST, Ph.D. in Artificial Intelligence (expected).
• 2015.03 - 2019.02, KAIST, B.S. in Mathematics, Electrical and Computer Engineering (double major).