Research map / Image generation / Fast diffusion sampling
Diffusion posterior sampling
Solve inverse problems with diffusion priors through posterior sampling.
72 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2022–2026) · an approach in Fast diffusion sampling, keyword cluster “score · distribution · denoising diffusion”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
24 papers in 2023–24 and 46 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 5.2% to 3.5%.
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
- 2022 · Come-Closer-Diffuse-Faster: Accelerating Conditional Diffusion Models for Inverse Problems Through Stochastic Contraction (CVPR 2022 · 274 citations)
- 2023 · Diffusion Posterior Sampling for General Noisy Inverse Problems (ICLR 2023 · 154 citations)
- 2024 · A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · 37 citations)
- 2025 · Improving Diffusion Inverse Problem Solving with Decoupled Noise Annealing (CVPR 2025 · 20 citations)
- 2026 · Constrained Particle Seeking: Solving Diffusion Inverse Problems with Just Forward Passes (AAAI 2026 · 1 citation)
Earliest papers here
- Come-Closer-Diffuse-Faster: Accelerating Conditional Diffusion Models for Inverse Problems Through Stochastic Contraction (CVPR 2022 · 274 citations)
- Improving Diffusion Models for Inverse Problems using Manifold Constraints (NeurIPS 2022 · 99 citations)
- Diffusion Posterior Sampling for General Noisy Inverse Problems (ICLR 2023 · 154 citations)
What has been tried
45 papers here report 501 measured changes (a component added, removed or swapped, compared with a baseline): helped 69%, no clear effect 6%, hurt 22%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Use 100 optimization steps instead of 10 | ImageNet inpainting · PSNR (dB) | 18.83 -> 23.13 | Helped | A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · 37 citations) |
| Use Gaussian approximation rather than diffusion-score estimation for p(x0 | xt) | FFHQ, Gaussian deblurring · PSNR | 29.57 -> 29.40 | Hurt | Improving Diffusion Inverse Problem Solving with Decoupled Noise Annealing (CVPR 2025 · 20 citations) |
| Add the second-order trace correction (η = 0.02) | COCO 2017 validation, salt-pepper denoising · LPIPS | 0.404 -> 0.388 | Helped | Beyond First-Order Tweedie: Solving Inverse Problems using Latent Diffusion (CVPR 2024 · 19 citations) |
| Add decoder inversion D† to naïve DDIM inversion | Background-preserving editing (60 × 5 images) · NMSE (×10−3) | 30.4±4.8 -> 18.4±2.0 | Helped | On Exact Inversion of DPM-Solvers (CVPR 2024 · 16 citations) |
| use DDS-DDIM instead of DDS-DDPM in DPnP | Image reconstruction tasks · performance | No clear effect | Provably Robust Score-Based Diffusion Posterior Sampling for Plug-and-Play Image Reconstruction (NeurIPS 2024 · 14 citations) | |
| Replace stochastic encoding with stochastic resampling | LSUN-Bedroom and CelebA-HQ Gaussian deblurring and super-resolution · PSNR | Helped | Solving Inverse Problems with Latent Diffusion Models via Hard Data Consistency (ICLR 2024 · 13 citations) | |
| use 5 CG iterations per denoising step | Uniform 1D ×4 MRI, 49 NFE · PSNR[db] | 31.51 -> 34.61 | Helped | Decomposed Diffusion Sampler for Accelerating Large-Scale Inverse Problems (ICLR 2024 · 6 citations) |
| Add pretrained DMs for blur kernels and tilt maps | BID with turbulence · reconstruction quality | Helped | DMPlug: A Plug-in Method for Solving Inverse Problems with Diffusion Models (NeurIPS 2024 · 6 citations) | |
| Increase MMPS solver iterations from 1 to 5 | Motion deblur · PSNR | 22.12 -> 26.70 | Helped | Learning Diffusion Priors from Observations by Expectation Maximization (NeurIPS 2024 · 5 citations) |
| Set the midpoint schedule to η = 0.75 rather than η = 1 | Gaussian mixture, d = 200 · Sliced Wasserstein distance | Helped | Variational Diffusion Posterior Sampling with Midpoint Guidance (ICLR 2025 · 3 citations) | |
| Increase low-rank/matrix-free approximation rank from k=0 to k=1 | FFHQ 256×256-1K, SR ×4 · LPIPS | 0.254 -> 0.171 | Helped | Solving Inverse Problems via Diffusion Optimal Control (NeurIPS 2024 · 3 citations) |
| Use 1-step rather than 2-step CM mapping | LSUN-Bedroom Gaussian deblur · PSNR | 21.3 -> 22.4 | Helped | Consistency Posterior Sampling for Diverse Image Synthesis (CVPR 2025 · 2 citations) |
What reviewers will expect
Datasets and baselines named in the 28 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- ImageNet · 11 of 28 papers
- FFHQ · 10 of 28 papers
- DIV2K · 3 of 28 papers
- CelebA-HQ · 3 of 28 papers
- FFHQ 256x256 · 2 of 28 papers
- ImageNet 256x256 · 2 of 28 papers
- AFHQ · 2 of 28 papers
- CelebA · 2 of 28 papers
Baselines
- DPS · 14 of 28 papers
- DDNM · 9 of 28 papers
- PSLD · 8 of 28 papers
- DDRM · 7 of 28 papers
- ReSample · 7 of 28 papers
- DAPS · 7 of 28 papers
- DiffPIR · 6 of 28 papers
- PGDM · 4 of 28 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 70 papers; from OpenAlex and the papers' affiliations).
- NVIDIA · 5 papers
- Korea Advanced Institute of Science and Technology · 5 papers
- Google · 4 papers
- University of Michigan · 3 papers
- Georgia Institute of Technology · 2 papers
- California Institute of Technology · 2 papers
- Imperial College London · 2 papers
- University of Michigan - Ann Arbor · 2 papers
- Microsoft · 2 papers
- Amazon · 2 papers
Use it
Download BibTeX of 27 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)
- Denoising diffusion probabilistic models (97 papers)
- Guidance for conditional diffusion (123 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).