Controllable Generation
Developing generative models that enable fine-grained control over visual generation and editing through structured guidance, feedback-driven refinement, and physical priors.
Liangbing(Liam) Zhao
PhD Student · King Abdullah University of Science and Technology
Controllable Generation · World Models
loading scene…
01
About
Hi, I am Liangbing Zhao. You can also call me Liam. I am currently a first-year PhD student at King Abdullah University of Science and Technology (KAUST), where I also earned my Master’s degree. I am privileged to be advised by Prof. Mohamed Elhoseiny.
Previously, I completed my undergraduate studies under the supervision of Prof. Si Liu. I also spent time as a research intern at Meitu Inc. with Dr. Xuecheng Nie. For more details, please refer to my CV.
World Models
Building world models that learn physical dynamics through perception, interaction, and feedback, enabling prediction and planning in real-world environments.
02
News
PhysicEdit is accepted by ICML2026!
Thanks to the co-authors, especially Shengbo, one paper is accepted by Nature Communications.
ReflectionFlow and WikiAutoGen are accepted by ICCV2025!
ToddlerDiffusion is accepted by ICLR2025!
One paper is accepted by Electronics!
03
Selected publications
ICCV 2025
From reflection to perfection: Scaling inference-time optimization for text-to-image diffusion models via reflection tuning
Proceedings of the IEEE/CVF International Conference on Computer Vision, 2025
ICCV 2025
Wikiautogen: Towards multi-modal wikipedia-style article generation
Proceedings of the IEEE/CVF International Conference on Computer Vision, 2025
ICLR 2025
ToddlerDiffusion: Interactive Structured Image Generation with Cascaded Schrödinger Bridge
The Thirteenth International Conference on Learning Representations, 2025
Electronics 2023
Cross-Attention and Seamless Replacement of Latent Prompts for High-Definition Image-Driven Video Editing
Electronics, 2023
Get in touch.
feel free to contact me!