About
Canyu Zhao
Hello, I'm a 3rd-year PHD Student at Zhejiang University, specializing in VLM, DIffusion Model and 3D Display.
Research Interests
Large Vision Language Model
Advancing the reasoning capabilities of Multimodal Large Language Models.
VLM/LLM
RL Post-training
Agent
Think with Image
Diffusion Model
Developing controllable generative models for high-fidelity image and video generation and editing.
Image Generation
Video Generation
Editing
Customization
3D Display
Glasses-free 3D display technologies and real-time rendering techniques.
3D Display
Featured Publications
Diception: A generalist diffusion model for visual perceptual tasks
Conference
NeurIPS SpotLight Award
2025
Freecustom: Tuning-free customized image generation for multi-concept composition
Conference
CVPR
2024
Omni-R1: Reinforcement Learning for Omnimodal Reasoning via Two-System Collaboration
Conference
NeurIPS
2025
FreerCustom: Training-Free Multi-Concept Customization for Image and Video Generation
Journal
International Journal of Computer Vision
2025
Active-O3: Empowering Multimodal Large Language Models with Active Perception via GRPO
Arxiv
Arxiv
2025
Tinker: Diffusion's Gift to 3D--Multi-View Consistent Editing From Sparse Inputs without Per-Scene Optimization
Arxiv
Arxiv
2025
StaMo: Unsupervised Learning of Generalizable Robot Motion from Compact State Representation
Arxiv
Arxiv
2025
NoTVLA: Narrowing of Dense Action Trajectories for Generalizable Robot Manipulation
Arxiv
Arxiv
2025
Bridge Thinking and Acting: Unleashing Physical Potential of VLM with Generalizable Action Expert
Arxiv
Arxiv
2025
Featured Projects
DreamFace: A Commercial AI Platform for Image and Video Generation
current
2025 - Present
Dreamface, Create Effortless AI Videos & Photos. Create avatar video, AI video, and AI photo with a single click. Participated in Video Generation Model development.