Perception
- Video understanding
- Computational imaging
Perception and Reasoning through Intelligent Synthesis and Modeling
Four pillars
of intelligence
Reasoning
- Multimodal intelligence
- Agents & alignment
Learning
- Generative modeling
- World models
Acting
- Controllable generation
- Real-world agents
Welcome to the PRISM Lab at Korea University, led by Hyungjin Chung. We develop generative models and agents—and better ways to guide and align them—so that intelligent systems can reason, create, and act in complex real-world settings.
We believe that better models create better intelligence, while better guidance and alignment turn that intelligence into breakthroughs. Our work connects foundational research with computational imaging, medical AI, AI4Science, physical AI, and industrial applications.
Our core research areas include:
- Generative AI and agents
- Video generation and world models
- Video understanding
- Inverse problems and computational imaging
- Medical AI and AI4Science
Explore our research directions →
We are looking for motivated students and postdoctoral researchers to join the lab. Please see the contact page for details.
News
View allTwo papers, InverseCrafter and ReDesign, are accepted to ECCV 2026.
One paper, Contrastive CFG, is accepted to ICML 2026.
Two papers, ReDirector and Video Parallel Scaling, are accepted to CVPR 2026.