●builderYou can apply this compression framework to reduce DiT inference costs in production image editing features like inpainting or generative fill.
●researcherProvides a concrete efficiency method for DiT-based editing pipelines with quantified latency baselines across mask ratio ranges.