Research map / Image generation / Fast diffusion sampling
Image Diffusion Preview with Consistency Solver
Fu-Yun Wang, Hao Zhou, Liangzhe Yuan, Sanghyun Woo, Boqing Gong, Bohyung Han, Ming-Hsuan Yang, Han Zhang, Yukun Zhu, Ting Liu, Long Zhao
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 |
|---|---|---|---|
| train solver coefficients with RL instead of trajectory distillation (ablation) | ConsistencySolver · COCO 2017 SD1.5, 5 steps | FID: 22.91 -> 20.39 | Improved |
| train solver coefficients with RL instead of trajectory distillation (ablation) | ConsistencySolver · COCO 2017 SD1.5, 8 steps | FID: 19.65 -> 18.82 | Improved |
| use learned ConsistencySolver instead of Multistep DPM-Solver | SD1.5 sampling · COCO 2017, 5 steps | FID: 25.87 -> 20.39 | Improved |
| use ConsistencySolver instead of Euler | FLUX.1-Kontext · KontextBench, 4 steps | Edit Reward: 0.61 -> 0.73 | Improved |
| apply solver trained on SD1.5 to SDXL without retraining | ConsistencySolver · COCO 2017 SDXL, 10 steps | FID: 26.32 -> 23.32 | Improved |
| use preview-then-refine workflow with 8-step ConsistencySolver | Diffusion Preview · LAION prompts | inference time (s): 6.35 -> 2.87 | Improved |
| use preview-then-refine workflow | Diffusion Preview · LAION prompts | average attempts: 6.00 -> 6.35 | Hurt |
| use DMD2 one-step distillation instead of the base model | DMD2 · GenEval prompts | prompts satisfied within 10 attempts: 57.0% of base (w/ GAN), 47.1% (w/o GAN) | Hurt |
| use ConsistencySolver instead of DMD2 for previews | ConsistencySolver · GenEval prompts | prompts satisfied within 10 attempts: 57.0% -> 94.2% of base | Improved |
| use solver order 4 instead of lower orders (ablation) | ConsistencySolver · COCO 2017 | structural fidelity | Improved |
| increase solver order from 4 to 5 (ablation) | ConsistencySolver · COCO 2017 | consistency | No clear effect |
| use depth-map reward (ablation) | ConsistencySolver RL · COCO 2017 | consistency | Improved |
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.
- UniPC — UniPC: A Unified Predictor-Corrector Framework for Fast Sampling of Diffusion Models (NeurIPS 2023) · 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
- A Simple Early Exiting Framework for Accelerated Sampling in Diffusion Models (ICML 2024 · similarity 0.88)
- SCott: Accelerating Diffusion Models with Stochastic Consistency Distillation (AAAI 2025 · similarity 0.88)
- AdaDiff: Adaptive Step Selection for Fast Diffusion Models (AAAI 2025 · similarity 0.88)
- Simplifying, Stabilizing and Scaling Continuous-time Consistency Models (ICLR 2025 · similarity 0.87)
- Towards Better Alignment: Training Diffusion Models with Reinforcement Learning Against Sparse Rewards (CVPR 2025 · similarity 0.87)
- Plug-and-Play Diffusion Distillation (CVPR 2024 · 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.
