General Information
- Full Name
- Liangbing Zhao
- Languages
- Chinese, English
Education
2019
B.S.
Beihang University
- Supervised by Professor Si Liu.
- GPA 3.78/4.0
- Core Courses
- Data Structures[C++] (97/100), Python Programming (95/100)
- Mathematical Analysis I (93/100), Linear Algebra (94/100), Mathematical Analysis II (99/100), Discrete Mathematics II (96/100)
- Computer Organization (91/100), Operating Systems (91/100), Object-Oriented Programming (91/100)
Research Experience
2023 - 2023
ToddlerDiffusion——Flash Interpretable Controllable Diffusion Model
KAUST
- Introduced an inherently interpretable and controllable diffusion model that systematically generates a chain of interpretable stages, generating sketches, palettes, and finally RGB images.
- Leveraged bridge architecture to reduce training and sampling steps to 10-20, but getting the same quality results as previous state-of-the-art methods.
- Under review for CVPR2024.
2022-2023
Cross-Attention and Seamless Replacement of Latent Prompts for High-Definition Image-Driven Video Editing
Meitu Inc.
- Introduced a novel image-driven video editing task to generate customized video from an input image.
- Proposed a LIVE(Latent prompt-based Image-driven Video Editing)framework to deal with customized video editing task. Devised a training and inference pipeline specific to the video editing task maintaining cross-frame consistency while ensuring that the new object is coherent with the sequence’s motion.
- Accepted by Electronics.
Honors and Awards
2020
- Special Prize for Learning Merit Scholarship
- Beihang University Excellent Student
Academic Interests
Controllable Generation
- Developing generative models that enable fine-grained control over visual generation and editing through structured guidance, feedback-driven refinement, and physical priors.
World Models
- Building world models that learn physical dynamics through perception, interaction, and feedback, enabling prediction and planning in real-world environments.