Seoul National University undergraduate publishes at NeurIPS, CVPR and ICML in generative visual computing
Lee Hyun-soo, an undergraduate in Seoul National University's Department of Electrical and Computer Engineering, has authored papers accepted at NeurIPS, CVPR, ECCV, WACV and ICML on diffusion-based image, 3D and pose generation.
Seoul National University’s College of Engineering announced on Aug. 7, 2026 that Lee Hyun-soo, an undergraduate student in the Department of Electrical and Computer Engineering, has published a series of papers at major international conferences in generative visual computing.
According to the college, Lee was co-first author of “Conditional Score Guidance for Text-Driven Image-to-Image Translation” at NeurIPS 2023, which proposed a conditional score guidance method. At CVPR 2026 he presented “Low-Resolution Editing is All You Need for High-Resolution Editing,” which showed that patch optimization synchronized with a low-resolution editing result can enable editing at resolutions above 1K, and was co-first author of “Image-Guided Geometric Stylization of 3D Meshes,” a technique for deforming the geometric shape of 3D objects.
His earlier CVPR 2025 paper, “SyncSDE: A Probabilistic Framework for Diffusion Synchronization,” identified in probabilistic terms where correlations should be introduced between diffusion generation processes, the college said. At ECCV 2026 he presented “Accelerated Likelihood Maximization for Diffusion-based Versatile Content Generation,” which handles 3D texture generation and human motion completion, while “Point2Pose” at WACV 2026 recast point-cloud-based 3D pose estimation as a conditional generation problem. At ICML 2026 he presented “Calibrated Test-Time Guidance for Bayesian Inference,” proposing an estimation method for generating samples from Bayesian posterior distributions.
Lee is pursuing a double major in mathematical sciences alongside electrical and computer engineering, and has carried out research in the computer vision laboratory of Professor Han Bo-hyung and the 3D vision laboratory of Professor Kim Young-min, the college said. He also conducted research with Professor Stephan Mandt while on exchange at the University of California, Irvine, and is currently working on representation learning at Seoul National University’s machine perception and reasoning laboratory.
The college said his goal is to develop intelligent systems that connect generative modeling with representation learning.