XYZFlow:Scaling Multi dimensional Shortcut Flows for Efficient Generative Modeling
2026-08-12 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors propose XYZFlow, a new method to generate high-quality images faster than previous techniques. Instead of relying on slow step-by-step processes, their approach uses a multidimensional way to make image creation paths easier to learn and follow. They improve the model by conditioning on the whole history of the image generation and by generating image patches in sequence with information from earlier patches. Their tests show that XYZFlow speeds up image generation significantly while maintaining good visual quality.
Diffusion modelsImage generationFlow matchingAutoregressive modelingMultidimensional scalingDenoisingNon-Markovian conditioningNext Shortcut PredictionFID score
Authors
Jinxiu Liu, Xuanming Liu, Kangfu Mei, Yandong Wen, Weiyang Liu
Abstract
High-fidelity image generation faces a trade-off between speed and quality. Diffusion models produce strong visuals but require costly iterative sampling. Existing efficient methods mainly distill pretrained models into few-step samplers, a challenging process that depends heavily on teacher-model quality. In this paper, we introduce XYZFlow, a framework that rethinks efficient generation through multidimensional scaling of flow matching. Unlike single-step mappings, XYZFlow enhances expressivity by making probability paths more identifiable and learnable through structured multidimensional conditioning. We view autoregressive modeling as implicit flow straightening, where richer context reduces trajectory ambiguity. XYZFlow realizes this idea through two orthogonal dimensions: temporal scaling, which uses non-Markovian conditioning on the full denoising history; and spatial scaling, enabled by Next Shortcut Prediction, which sequentially generates patches using preceding patches' denoising trajectories as priors. Experiments show that XYZFlow achieves state-of-the-art performance, with 7.2-8.5X teacher speedups and competitive FID, while Next Shortcut Prediction delivers superior quality-latency trade-offs over model scaling or step reduction.