IMAGE GENERATION WITH DIFFUSION STATE SPACE MODELS

Summary from the TLDR AI newsletter

Diffusion State Space Models (DiS) are a new type of diffusion model that use a state space backbone instead of the traditional U-Net for image data. These models can handle long-range dependencies and are efficient in generating high-quality images with less computational effort.