Rethinking Cross-Layer Information Routing in Diffusion Transformers
Researchers have developed Diffusion-Adaptive Routing (DAR), a novel method to improve information flow in Diffusion Transformers (DiTs). By analyzing cross-layer information dynamics, they identified inefficiencies in traditional residual connections. DAR offers a learnable, timestep-adaptive aggregation that enhances training efficiency and model quality, achieving better FID scores on ImageNet with significantly fewer training iterations. AI
IMPACT Introduces a novel technique to enhance training efficiency and quality for diffusion models, potentially accelerating development of visual generation AI.