Research map / Image generation / Fast diffusion sampling
XYZFlow: Scaling Multidimensional Shortcut Flows for Efficient Generative Modeling
Jinxiu Liu, Xuanming Liu, Kangfu Mei, Yandong Wen, Weiyang Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add full-history (inter-patch trajectory) conditioning (ablation) | XYZFlow-B · ImageNet 256x256 class-conditional | FID: 3.51 -> 2.02 | Improved |
| add local-history (intra-patch temporal) conditioning (ablation) | XYZFlow-B · ImageNet 256x256 class-conditional | FID: 3.03 -> 2.25 | Improved |
| add shortcut (progressive step) schedule (ablation) | XYZFlow-B · ImageNet 256x256 class-conditional | FID and total steps: FID 2.05 -> 2.02; steps 20 -> 14 | No clear effect |
| add GAN discriminator loss (ablation) | XYZFlow-B · ImageNet 256x256 class-conditional | FID: 2.02 -> 1.63 | Improved |
| add GAN discriminator loss (ablation) | XYZFlow-L · ImageNet 256x256 class-conditional | FID: 1.79 -> 1.25 | Improved |
| use 8x8 cell size instead of 1x1 tokens (ablation) | XYZFlow-B · ImageNet 256x256 class-conditional | FID: 2.85 -> 2.02 | Improved |
| treat whole image as one cell (16x16) instead of 8x8 (ablation) | XYZFlow-B · ImageNet 256x256 class-conditional | FID: 2.02 -> 2.55 | Hurt |
| progressive 5->4->3->2 instead of constant 5 steps (ablation) | XYZFlow-B · ImageNet 256x256 class-conditional | FID and total steps: FID 1.63 -> 1.63; steps 20 -> 14 | No clear effect |
| aggressive 5->2->2->2 schedule instead of 5->4->3->2 (ablation) | XYZFlow-B · ImageNet 256x256 class-conditional | recall: 0.62 -> 0.58 | Hurt |
| 8->8->8->8 steps instead of 5->4->3->2 (ablation) | XYZFlow-B · ImageNet 256x256 class-conditional | FID: 1.63 -> 1.61; steps 14 -> 32 | No clear effect |
| use weaker DiT/XL-2 teacher instead of xAR (ablation) | XYZFlow-B · ImageNet 256x256 class-conditional | FID: 1.63 -> 1.74 (GAN) | Mixed |
| use stronger xAR teacher (xAR-B to xAR-H) (ablation) | XYZFlow-L · ImageNet 256x256 class-conditional | FID (w/ GAN): 1.32 -> 1.25 | No clear effect |
Closest papers by meaning
- Align Your Flow: Scaling Continuous-Time Flow Map Distillation (NeurIPS 2025 · similarity 0.90)
- Expanding Flow Maps (NeurIPS 2026 · similarity 0.89)
- Improving Progressive Generation with Decomposable Flow Matching (NeurIPS 2025 · similarity 0.89)
- Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation (NeurIPS 2026 · similarity 0.88)
- Cross Flow: One-Step Generation Across Latent and Pixel Spaces (NeurIPS 2026 · similarity 0.88)
- One Step Diffusion via Shortcut Models (ICLR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
