Research map / Image generation / Fast diffusion sampling
SenseFlow: Scaling Distribution Matching for Flow-based Text-to-Image Distillation
Xingtong Ge, Xin Zhang, Tongda Xu, Yi Zhang, Xinjie Zhang, Yan Wang, Jun Zhang
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 implicit distribution alignment (IDA) to DMD with ISG removed (ablation) | DMD on SD 3.5 Large · COCO-5K | FID-T: 43.84 -> 17.00 | Improved |
| add intra-segment guidance (ISG) (ablation) | SenseFlow on SD 3.5 Large · COCO-5K | FID-T: 17.00 -> 13.38 | Improved |
| add ISG (ablation) | SenseFlow on SD 3.5 Large · COCO-5K | ImageReward: 1.0186 -> 1.1713 | Improved |
| add IDA on top of TTUR 5:1 (ablation) | DMD on SD 3.5 Large · Training-hours FID curves | FID stability | Improved |
| replace diffusion-based discriminator with VFM-based discriminator (ablation) | DMD2-SDXL · COCO-5K | ImageReward: 0.9324 -> 0.9744 | Improved |
| replace diffusion-based discriminator with VFM-based discriminator (ablation) | DMD2-SDXL · COCO-5K | FID-T: 15.04 -> 18.55 | Hurt |
| increase TTUR ratio to 20:1 without IDA (ablation) | DMD on SD 3.5 Large · Training-hours FID curves | FID stability | No clear effect |
| use stochastic vs Euler sampler for distilled model | SenseFlow FLUX.1 dev · GenEval | overall: 0.6471 (stochastic) vs 0.6420 (Euler) | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Hyper-SDXL — Hyper-SD: Trajectory Segmented Consistency Model for Efficient Image Synthesis (NeurIPS 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- On the Scalability of Diffusion-based Text-to-Image Generation (CVPR 2024 · similarity 0.88)
- Improved Distribution Matching Distillation for Fast Image Synthesis (NeurIPS 2024 · similarity 0.88)
- Continuous-Time Distribution Matching for Few-Step Diffusion Distillation (NeurIPS 2026 · similarity 0.88)
- InstaFlow: One Step is Enough for High-Quality Diffusion-Based Text-to-Image Generation (ICLR 2024 · similarity 0.87)
- One-step Diffusion with Distribution Matching Distillation (CVPR 2024 · similarity 0.87)
- Guided Score identity Distillation for Data-Free One-Step Text-to-Image Generation (ICLR 2025 · similarity 0.87)
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.
