Research map / Image generation / Fast diffusion sampling
Denoising diffusion probabilistic models
Develop DDPM variants, implicit samplers, and bridge formulations for generation and denoising.
97 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Fast diffusion sampling, keyword cluster “score · distribution · denoising diffusion”
State History Benchmarks Tried Baselines Checklist Preprints Who Export
Where it stands
Few recent reports The tracked benchmarks have at most one report from 2025–26 that could beat the old record, too few to say whether they are still moving.
37 papers in 2023–24 and 48 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 8.0% to 3.6%.
Open questions the data points to
- Papers report DDIM on CIFAR-10 (FID) anywhere from 4.67 to 11.14: comparisons against it depend on whose number you take.
- Papers report Soft Diffusion on CIFAR10 (FID) anywhere from 3.86 to 4.64: comparisons against it depend on whose number you take.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2017 · Learning to Generate Samples from Noise through Infusion Training (ICLR 2017 · 11 citations)
- 2020 · Denoising Diffusion Probabilistic Models (NeurIPS 2020 · 5,610 citations)
- 2021 · Improved Denoising Diffusion Probabilistic Models (ICML 2021 · 411 citations)
- 2022 · RePaint: Inpainting Using Denoising Diffusion Probabilistic Models (CVPR 2022 · 1,653 citations)
- 2023 · gDDIM: Generalized denoising diffusion implicit models (ICLR 2023 · 27 citations)
- 2024 · Boosting Diffusion Models with Moving Average Sampling in Frequency Domain (CVPR 2024 · 30 citations)
- 2025 · Effective Cloud Removal for Remote Sensing Images by an Improved Mean-Reverting Denoising Model with Elucidated Design Space (CVPR 2025 · 17 citations)
- 2026 · Back to Basics: Let Denoising Generative Models Denoise (CVPR 2026 · 4 citations)
Earliest papers here
- Learning to Generate Samples from Noise through Infusion Training (ICLR 2017 · 11 citations)
- Denoising Diffusion Probabilistic Models (NeurIPS 2020 · 5,610 citations)
- Improved Denoising Diffusion Probabilistic Models (ICML 2021 · 411 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
CIFAR-10 · FID (lower is better)
13 papers, 2021–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 4.04 | Denoising Diffusion Implicit Models (ICLR 2021 · 102 citations) |
| 2023 | 1.9 | Stable Target Field for Reduced Variance Score Estimation in Diffusion Models (ICLR 2023 · 5 citations) |
What has been tried
68 papers here report 748 measured changes (a component added, removed or swapped, compared with a baseline): helped 67%, no clear effect 6%, hurt 22%, mixed 5%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add pixel-space moving average | ImageNet 128×128 · FID | 23.00 -> 21.88 | Helped | Boosting Diffusion Models with Moving Average Sampling in Frequency Domain (CVPR 2024 · 30 citations) |
| Add cloudy images as conditioning inputs | Sen2 MTC New · PSNR | 12.81 -> 18.26 | Helped | Effective Cloud Removal for Remote Sensing Images by an Improved Mean-Reverting Denoising Model with Elucidated Design Space (CVPR 2025 · 17 citations) |
| corrupt ground-truth boxes and project them as event queries instead of corrupting event latents | URBAN-SED test · Event-F1 (%) | 41.3 -> 37.29 | Hurt | DiffSED: Sound Event Detection with Denoising Diffusion (AAAI 2024 · 10 citations) |
| Increase soft-mask Gaussian kernel standard deviation from 0.5 to larger values | Single-step assimilation and forecasts from assimilated data · RMSE | 94.6 -> 90.1 | Helped | DiffDA: a Diffusion model for weather-scale Data Assimilation (ICML 2024 · 10 citations) |
| use resnoise prediction instead of x₀ prediction | ISTD original · PSNR | 29.67 -> 30.09 | Helped | Resfusion: Denoising Diffusion Probabilistic Models for Image Restoration Based on Prior Residual Noise (NeurIPS 2024 · 10 citations) |
| Increase $\gamma$ from 2 to 4 | Toy data, unconditional generation · Mean distance | 65.81± 17.86 -> 35.99± 7.58 | Helped | SPD-DDPM: Denoising Diffusion Probabilistic Models in the Symmetric Positive Definite Space (AAAI 2024 · 9 citations) |
| Increase Euler step ratio s from 0 (ODE-only) toward a tuned value | Edges→Handbags and DIODE · FID | Helped | Denoising Diffusion Bridge Models (ICLR 2024 · 4 citations) | |
| use stochastic rather than deterministic sampling (η = 1) | LJSpeech · MCD | 1.991 ± 0.486 -> 1.862 ± 0.443 | Helped | Cauchy Diffusion: A Heavy-tailed Denoising Diffusion Probabilistic Model for Speech Synthesis (AAAI 2025 · 3 citations) |
| optimize noise with only frequency matching, without the Meta-Network | OxfordPets, 16-shot · top-1 accuracy | Helped | Noise Matters: Optimizing Matching Noise for Diffusion Classifiers (NeurIPS 2025 · 3 citations) | |
| reverse the default scaling scheme | ImageNet validation · AIC | 18.9 -> 16.2 | Hurt | Denoising Diffusion Path: Attribution Noise Reduction with An Auxiliary Diffusion Model (NeurIPS 2024 · 3 citations) |
| add camera-setting concatenation | Noise synthesis · AKLD | 0.169 -> 0.137 | Helped | Realistic Noise Synthesis with Diffusion Models (AAAI 2025 · 2 citations) |
| Add shortest-path graph modeling and its loss to DDIM | Not reported · Not reported | No clear effect | Optimizing for the Shortest Path in Denoising Diffusion Model (CVPR 2025 · 2 citations) |
Baseline numbers that disagree
The same baseline on the same benchmark and metric, reported more than 5% apart by different papers here. Training budgets, data splits and reimplementations differ, so state which number you compare with.
| Baseline | Benchmark | Metric | Reported | Papers |
|---|---|---|---|---|
| DDIM | CIFAR-10 | FID | 4.67 – 11.14 | 2 papers, e.g. ICLR 2023 and CVPR 2025 |
| Soft Diffusion | CIFAR10 | FID | 3.86 – 4.64 | 2 papers, e.g. NeurIPS 2023 and ICLR 2023 |
What reviewers will expect
Datasets and baselines named in the 29 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- CIFAR-10 · 8 of 29 papers
- ImageNet · 4 of 29 papers
- CelebA-HQ · 3 of 29 papers
- FFHQ · 3 of 29 papers
- CelebA · 2 of 29 papers
- VoiceBank+DEMAND · 2 of 29 papers
- ImageNet 256x256 · 2 of 29 papers
- MS-COCO · 2 of 29 papers
Baselines
- DDIM · 5 of 29 papers
- DDNM · 3 of 29 papers
- PriorGrad · 2 of 29 papers
- DPM-solver · 2 of 29 papers
- ECM · 2 of 29 papers
- DDRM · 2 of 29 papers
- DPS · 2 of 29 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 85 papers; from OpenAlex and the papers' affiliations).
- Google · 6 papers
- University of Oxford · 5 papers
- Massachusetts Institute of Technology · 4 papers
- University of Electronic Science and Technology of China · 3 papers
- New York University · 3 papers
- Carnegie Mellon University · 3 papers
- Iowa State University · 2 papers
- Zhejiang University · 2 papers
- Imperial College London · 2 papers
- Tencent · 2 papers
Use it
Download BibTeX of 37 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Diffusion theory and design space (252 papers)
- Distributional control of diffusion (109 papers)
- Guidance for conditional diffusion (123 papers)
- Diffusion posterior sampling (72 papers)
- Diffusion theory and learned samplers (121 papers)
- Diffusion models as perception tools (198 papers)
Elsewhere in the topic
- Training-free diffusion acceleration (257 papers)
- Masked diffusion language model decoding (147 papers)
- Fast diffusion ODE solvers (120 papers)
- Flow matching distillation (95 papers)
- Distillation for few-step generation (84 papers)
- Rectified flow straightening (70 papers)
- Discrete diffusion solvers (64 papers)
- Guidance and alignment for flow matching (34 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).