Research map / Image generation / Fast diffusion sampling
Phased DMD: Few-step Distribution Matching Distillation via Score Matching within Subintervals
Xiangyu Fan, Zesong Qiu, Zhuguanyu Wu, Fanzhou Wang, Zhiqian Lin, Tianxiang Ren, Dahua Lin, Ruihao Gong, Lei Yang
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 |
|---|---|---|---|
| replace DMD2 stochastic gradient truncation with Phased DMD | Wan2.2-T2V-A14B distilled · T2V video generation (220 prompts) | optical flow: 3.23 -> 9.30 | Improved |
| replace DMD2 with Phased DMD | Wan2.2-T2V-A14B distilled · T2V video generation (220 prompts) | FVD vs base: 763.1 -> 700.9 | Improved |
| replace DMD2 with Phased DMD | Wan2.2-I2V-A14B distilled · I2V video generation (220 prompts) | optical flow: 7.87 -> 9.84 | Improved |
| replace DMD2 with Phased DMD | Wan2.1-T2V-14B distilled · T2I diversity | DINOv3 similarity: 0.826 -> 0.782 | Improved |
| replace DMD2 with Phased DMD | Wan2.2-T2V-A14B distilled · T2I diversity | LPIPS: 0.447 -> 0.481 | Improved |
| replace DMD2 with Phased DMD | Qwen-Image-20B distilled · T2I diversity | DINOv3 / LPIPS: 0.941 -> 0.958 DINOv3; 0.309 -> 0.322 LPIPS | Mixed |
| replace vanilla few-step DMD with Phased DMD | Wan2.1-T2V-14B distilled · T2I diversity | DINOv3 similarity: 0.825 -> 0.782 | Improved |
| use correct subinterval score-matching objective instead of biased target (ablation) | 1D toy flow model · toy 4-point distribution | sampling trajectory match | Improved |
| sample t from (tk,1) rather than (tk,tk-1) (ablation) | Phased DMD · ablation in supplement | quality | Improved |
Closest papers by meaning
- Improved Distribution Matching Distillation for Fast Image Synthesis (NeurIPS 2024 · similarity 0.92)
- CoDMD: Copula-aware Distribution Matching Distillation for Fast Video Generation (NeurIPS 2026 · similarity 0.90)
- Continuous-Time Distribution Matching for Few-Step Diffusion Distillation (NeurIPS 2026 · similarity 0.89)
- One-step Diffusion with Distribution Matching Distillation (CVPR 2024 · similarity 0.89)
- Distribution Matching Distillation without Fake Score Network (NeurIPS 2026 · similarity 0.89)
- OSV: One Step is Enough for High-Quality Image to Video Generation (CVPR 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.
