AI RESEARCH

RobuQ: Pushing DiTs to W1.58A2 via Robust Activation Quantization

arXiv CS.CV

ArXi:2509.23582v2 Announce Type: replace Diffusion Transformers (DiTs) have recently emerged as a powerful backbone for image generation, nstrating superior scalability and performance over U-Net architectures. However, their practical deployment is hindered by substantial computational and memory costs. While Quantization-Aware