Research map / Image generation / Fast diffusion sampling
Distributional control of diffusion
Steer pretrained generators for diversity, minority sampling, and distribution-level guidance.
109 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
Quiet None of the tracked benchmarks has a report from 2025–26.
34 papers in 2023–24 and 56 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 7.4% to 4.2%.
Open questions the data points to
- Papers report PixelCNN++ on CIFAR-10 (FID) anywhere from 54.25 to 65.9: comparisons against it depend on whose number you take.
No longer reported by papers here: CIFAR-10 (FID, last in 2023).
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 · Parallel Multiscale Autoregressive Density Estimation (ICML 2017 · 43 citations)
- 2018 · Learning Generative ConvNets via Multi-Grid Modeling and Sampling (CVPR 2018 · 51 citations)
- 2021 · Diffusion Models Beat GANs on Image Synthesis (NeurIPS 2021 · 2,151 citations)
- 2022 · Riemannian Score-Based Generative Modelling (NeurIPS 2022 · 26 citations)
- 2023 · ProlificDreamer: High-Fidelity and Diverse Text-to-3D Generation with Variational Score Distillation (NeurIPS 2023 · 159 citations)
- 2024 · Consistent3D: Towards Consistent High-Fidelity Text-to-3D Generation with Deterministic Sampling Prior (CVPR 2024 · 18 citations)
- 2025 · Normalizing Flows are Capable Generative Models (ICML 2025 · 4 citations)
- 2026 · Rectified Noise: A Generative Model Using Positive-incentive Noise (AAAI 2026 · 4 citations)
Earliest papers here
- Parallel Multiscale Autoregressive Density Estimation (ICML 2017 · 43 citations)
- Learning Generative ConvNets via Multi-Grid Modeling and Sampling (CVPR 2018 · 51 citations)
- Autoregressive Quantile Networks for Generative Modeling (ICML 2018 · 31 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)
7 papers, 2018–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2018 | 49.5 | Autoregressive Quantile Networks for Generative Modeling (ICML 2018 · 31 citations) |
| 2021 | 12.81 | Synthesize-It-Classifier: Learning a Generative Classifier Through Recurrent Self-Analysis (CVPR 2021 · 1 citation) |
| 2023 | 10.69 | TR0N: Translator Networks for 0-Shot Plug-and-Play Conditional Generation (ICML 2023 · 0 citations) |
What has been tried
65 papers here report 711 measured changes (a component added, removed or swapped, compared with a baseline): helped 69%, no clear effect 6%, hurt 22%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Use a random time step schedule | Text-to-3D prompts in Figure 6 · geometry and texture quality | Hurt | Consistent3D: Towards Consistent High-Fidelity Text-to-3D Generation with Deterministic Sampling Prior (CVPR 2024 · 18 citations) | |
| add the k-d tree to accelerate vicinal-loss neighbor search | Uneven · speedup | 2.7 | Helped | CcDPM: A Continuous Conditional Diffusion Probabilistic Model for Inverse Design (AAAI 2024 · 12 citations) |
| Set entropy regularization strength λ to 1 | Michelangelo-style statue of dog reading news on a cellphone · texture realism | Hurt | Taming Mode Collapse in Score Distillation for Text-to-3D Generation (CVPR 2024 · 12 citations) | |
| Add decomposed unconditional and conditional single-hand modeling | InterHand2.6M test, two-hand synthesis · FHID | 2.09 -> 1.00 | Helped | InterHandGen: Two-Hand Interaction Generation via Cascaded Reverse Diffusion (CVPR 2024 · 11 citations) |
| add the physical stability checker | ShapeNet-v2, mean of chair, table, airplane, and sofa · PSS | 0.755 -> 0.775 | Helped | GPLD3D: Latent Diffusion of 3D Shape Generative Models by Enforcing Geometric and Physical Priors (CVPR 2024 · 9 citations) |
| replace quickshift with felzenszwalb segmentation | CUB-200-2011 (probability) · accuracy | 55.47% -> 59.33% | Helped | Fully Exploiting Every Real Sample: SuperPixel Sample Gradient Model Stealing (CVPR 2024 · 8 citations) |
| Add a negative prompt to CSD | Texture synthesis · visual quality | Helped | Text-to-3D with Classifier Score Distillation (ICLR 2024 · 8 citations) | |
| apply guidance with weight 2 instead of guidance 0 | ImageNet 64x64 · FID | 25.3 -> 5.7 | Helped | Normalizing Flows are Capable Generative Models (ICML 2025 · 4 citations) |
| Replace resampled Gaussian noise with DDIM-inversion-inferred noise | Text-to-3D · visual generation quality | Helped | Score Distillation via Reparametrized DDIM (NeurIPS 2024 · 4 citations) | |
| Vary the negative-prompt descriptors | Text-to-image generation · Visual quality | No clear effect | Rethinking Score Distillation as a Bridge Between Image Distributions (NeurIPS 2024 · 4 citations) | |
| Train with multi-view images and augmentations instead of single-view images without augmentations | ShapeNet chairs subset · F-Score | 0.321 -> 0.452 | Helped | DI-PCG: Diffusion-based Efficient Inverse Procedural Content Generation for High-quality 3D Asset Creation (CVPR 2025 · 3 citations) |
| add model adaptation | CelebA-HQ · LPIPST | 1.45 ± 0.05 -> 1.07 ± 0.07 | Helped | IMPUS: Image Morphing with Perceptually-Uniform Sampling Using Diffusion Models (ICLR 2024 · 3 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 |
|---|---|---|---|---|
| PixelCNN++ | CIFAR-10 | FID | 54.25 – 65.9 | 2 papers, e.g. ICLR 2021 and ICML 2018 |
What reviewers will expect
Datasets and baselines named in the 25 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- CelebA · 5 of 25 papers
- ImageNet · 4 of 25 papers
- CIFAR-10 · 4 of 25 papers
- COCO · 3 of 25 papers
- Pascal VOC · 2 of 25 papers
- MNIST · 2 of 25 papers
Baselines
- Glow · 2 of 25 papers
- VDM · 2 of 25 papers
- NFDM · 2 of 25 papers
- EDM · 2 of 25 papers
- iDDPM · 2 of 25 papers
- ADM · 2 of 25 papers
- BigGAN · 2 of 25 papers
- CADS · 2 of 25 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 90 papers; from OpenAlex and the papers' affiliations).
- Korea Advanced Institute of Science and Technology · 6 papers
- Meta · 4 papers
- Tsinghua University · 4 papers
- Johns Hopkins University · 3 papers
- University of Oxford · 3 papers
- Google · 3 papers
- Massachusetts Institute of Technology · 3 papers
- Technology · 3 papers
- Zhejiang University · 2 papers
- University of Texas at Austin · 2 papers
Use it
Download BibTeX of 36 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)
- Denoising diffusion probabilistic models (97 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).