AI RESEARCH

Beyond Point-Wise Matching: Structural Representation Alignment for Accelerating Diffusion Transformers

arXiv CS.CV

ArXi:2605.16949v1 Announce Type: new Recent advances in Diffusion Transformers (DiTs) nstrate that aligning noisy latent states with well-trained semantic features-as pioneered by Representation Alignment (REPA)-can substantially accelerate