Jaihoon Kim

About Me

I am a second-year Ph.D. student in School of Computing at KAIST, where I am advised by Prof. Minhyuk Sung. I work on designing generative models and efficient inference algorithms for synthesizing diverse data, capturing the inherent structure, representations, and physical constraints of the underlying modality.

Past Experiences

Amazon Web Service (Santa Clara, United States): Research Scientist Intern
Samsung Electronics (Hwasung, South Korea): Cloud Engineer
Samsung Electronics (Hwasung, South Korea): Cloud Engineer Intern

News

[Sep. 2025] Two papers are accepted to NeurIPS 2025
[Sep. 2025] One paper is accepted to WACV 2026
[July. 2025] Internship at AWS
[Jan. 2025] One paper is accepted to ICLR 2025
[Sep. 2024] One paper is accepted to NeurIPS 2024
[Sep. 2024] Joined KAIST School of Computing as a Ph.D student
[Mar. 2023] Joined KAIST School of Computing as a Master student
[Feb. 2023] One paper is accepted to CVPR 2023
[July. 2021] Full time engineer at Samsung Electronics
[June. 2020] Internship at Samsung Electronics

Publications

NeurIPS

Inference-Time Scaling for Flow Models via Stochastic Generation and Rollover Budget Forcing

Jaihoon Kim^*, Taehoon Yoon^*, Jisung Hwang^*, Minhyuk Sung (^* Equal contribution.)

Neural Information Processing Systems (NeurIPS), 2025.

PDF Code Project Page

NeurIPS

Moment- and Power-Spectrum-Based Gaussianity Regularization for Text-to-Image Models

Jisung Hwang, Jaihoon Kim, Minhyuk Sung

Neural Information Processing Systems (NeurIPS), 2025.

PDF

WACV

Unconditional Priors Matter! Improving Conditional Generation of Fine-Tuned Diffusion Models

Prin Phunyaphibarn, Phillip Y. Lee, Jaihoon Kim, Minhyuk Sung

IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2026.

PDF Project Page

ICLR

StochSync: Stochastic Diffusion Synchronization for Image Generation in Arbitrary Spaces

Kyeongmin Yeo^*, Jaihoon Kim^*, Minhyuk Sung (^* Equal contribution.)

International Conference on Learning Representations (ICLR), 2025.

PDF Code Project Page

NeurIPS

SyncTweedies: A General Generative Framework Based on Synchronized Diffusions

Jaihoon Kim^*, Juil Koo^*, Kyeongmin Yeo^*, Minhyuk Sung (^* Equal contribution.)

Neural Information Processing Systems (NeurIPS), 2024.

PDF Code Project Page

CVPR

Paired-Point Lifting for Enhanced Privacy-Preserving Visual Localization

Jungwhan Lee, Jaihoon Kim, Chanhyuk Yun, Je Hyeong Hong

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023.

PDF Code

About Me

Past Experiences

News

Publications

Academic Experiences

Academic Service