15 papers
SATB-VR: Training Few-Step Video Restoration Diffusion Model using SNR-Aware Trajectory Blending
Haoran Bai, Xiaoxu Chen, Xiaoyu Liu +4
While diffusion models excel in video restoration, their reliance on extensive iterative steps limits efficiency. Conversely, aggressive single-step distillation often compromises…
TextLDM: Language Modeling with Continuous Latent Diffusion
Jiaxiu Jiang, Jingjing Ren, Wenbo Li +10
Diffusion Transformers (DiT) trained with flow matching in a VAE latent space have unified visual generation across images and videos. A natural next step toward a single architect…
Auto-FlexSwitch: Efficient Dynamic Model Merging via Learnable Task Vector Compression
Junqi Gao, Dazhi Zhang, Zhichang Guo +3
Model merging has attracted attention as an effective path toward multi-task adaptation by integrating knowledge from multiple task-specific models. Among existing approaches, dyna…
Physics Consistency and Latent Dynamics in Spatiotemporal Physics Field Generation
Peimian Du, Jiabin Liu, Xiaowei Jin +2
Data-driven models for spatiotemporal physical field generation, such as flow and acoustic fields, often deviate from governing equations and lack interpretability in latent tempor…
Bridging Geometry-Coherent Text-to-3D Generation with Multi-View Diffusion Priors and Gaussian Splatting
Feng Yang, Wenliang Qian, Wangmeng Zuo +1
Score Distillation Sampling (SDS) leverages pretrained 2D diffusion models to advance text-to-3D generation but neglects multi-view correlations, being prone to geometric inconsist…
RoomEditor++: A Parameter-Sharing Diffusion Architecture for High-Fidelity Furniture Synthesis
Qilong Wang, Xiaofan Ming, Zhenyi Lin +4
Virtual furniture synthesis, which seamlessly integrates reference objects into indoor scenes while maintaining geometric coherence and visual realism, holds substantial promise fo…