Research map / Image generation / Fast diffusion sampling
Guidance for conditional diffusion
Improve classifier-free and rectified guidance for conditional generation.
123 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2019–2026) · an approach in Fast diffusion sampling, keyword cluster “score · distribution · denoising diffusion”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 1 of 1 tracked benchmark.
30 papers in 2023–24 and 91 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 6.5% to 6.9%.
Open questions the data points to
- ImageNet 256x256 (FID): best went from 2.12 (up to 2024) to 1.91 in 2025 — Visual Generation Without Guidance (ICML 2025 · 0 citations)
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
- 2019 · Multi-Winner Contests for Strategic Diffusion in Social Networks (AAAI 2019 · 11 citations)
- 2022 · Planning with Diffusion for Flexible Behavior Synthesis (ICML 2022 · 60 citations)
- 2023 · DIFUSCO: Graph-based Diffusion Solvers for Combinatorial Optimization (NeurIPS 2023 · 34 citations)
- 2024 · Region-Disentangled Diffusion Model for High-Fidelity PPG-to-ECG Translation (AAAI 2024 · 39 citations)
- 2025 · Adaptive Guidance: Training-free Acceleration of Conditional Diffusion Models (AAAI 2025 · 6 citations)
- 2026 · Spatial-Temporal Feedback Diffusion Guidance for Controlled Traffic Imputation (AAAI 2026 · 2 citations)
Earliest papers here
- Multi-Winner Contests for Strategic Diffusion in Social Networks (AAAI 2019 · 11 citations)
- Planning with Diffusion for Flexible Behavior Synthesis (ICML 2022 · 60 citations)
- DIFUSCO: Graph-based Diffusion Solvers for Combinatorial Optimization (NeurIPS 2023 · 34 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.
ImageNet 256x256 · FID (lower is better)
4 papers, 2023–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2023 | 3.84 | Rethinking Conditional Diffusion Sampling with Progressive Guidance (NeurIPS 2023 · 0 citations) |
| 2024 | 2.12 | Elucidating the design space of classifier-guided diffusion generation (ICLR 2024 · 0 citations) |
| 2025 | 1.91 | Visual Generation Without Guidance (ICML 2025 · 0 citations) |
What has been tried
61 papers here report 646 measured changes (a component added, removed or swapped, compared with a baseline): helped 68%, no clear effect 8%, hurt 21%, 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 |
|---|---|---|---|---|
| Increase ROI window size from 16 to 32 | AFib detection · performance | Helped | Region-Disentangled Diffusion Model for High-Fidelity PPG-to-ECG Translation (AAAI 2024 · 39 citations) | |
| Condition the diffusion model on scene features | Language instruction following tasks · Task success rate | Hurt | Diffusion-ES: Gradient-free Planning with Diffusion for Autonomous and Instruction-guided Driving (CVPR 2024 · 37 citations) | |
| Use M = 2 or 3 generated images to estimate Diversity Reward instead of M = 5 | ImageNet-128x128 · Recall | Hurt | Training Diffusion Models Towards Diverse Image Generation with Reinforcement Learning (CVPR 2024 · 24 citations) | |
| Allow independent EMA lengths for the main and guiding models | ImageNet-512 · FID | Helped | Guiding a Diffusion Model with a Bad Version of Itself (NeurIPS 2024 · 22 citations) | |
| Train the pose readout head on more PascalVOC images (100, 1k, or 8.5k) | MSCOCO validation · PCK | Helped | Readout Guidance: Learning Control from Diffusion Features (CVPR 2024 · 14 citations) | |
| include higher noise levels in the guidance interval | ImageNet · FID | Hurt | Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models (NeurIPS 2024 · 10 citations) | |
| Increase recurrence from 1 to 2 | CIFAR10 label guidance · accuracy | 52.0 -> 66.8 | Helped | TFG: Unified Training-Free Guidance for Diffusion Models (NeurIPS 2024 · 8 citations) |
| Use DRaFT-LV instead of DRaFT-1 | HPDv2 averaged · HPSv2 score | 0.3467 -> 0.3536 | Helped | Directly Fine-Tuning Diffusion Models on Differentiable Rewards (ICLR 2024 · 7 citations) |
| use random augmentation with |T|=10 rather than no random augmentation | ImageNet text guidance · CLIP score | 0.541 -> 0.571 | Helped | Understanding and Improving Training-free Loss-based Diffusion Guidance (NeurIPS 2024 · 5 citations) |
| Increase σ for SEG | Text-conditional generation · FID | 53.423 -> 48.284 -> 41.784 -> 33.819 -> 29.325 -> 26.169 | Helped | Smoothed Energy Guidance: Guiding Diffusion Models with Reduced Energy Curvature of Attention (NeurIPS 2024 · 4 citations) |
| Use the continuous-time NELBO (T→∞) instead of finite-step training | Amazon Polarity · validation perplexity | Helped | Simple Guidance Mechanisms for Discrete Diffusion Models (ICLR 2025 · 3 citations) | |
| Use Flow Matching ODE rather than Memoryless FM at inference | Multi-prompt text-to-image · ClipScore | 29.58 ± 0.19 -> 30.59 ± 0.05 | Helped | Adjoint Matching: Fine-tuning Flow and Diffusion Generative Models with Memoryless Stochastic Optimal Control (ICLR 2025 · 3 citations) |
What reviewers will expect
Datasets and baselines named in the 50 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 50 papers
- FFHQ · 5 of 50 papers
- CIFAR10 · 4 of 50 papers
- QM9 · 3 of 50 papers
- COCO2017 · 2 of 50 papers
- WikiArt · 2 of 50 papers
- CelebA-HQ · 2 of 50 papers
- COCO 2014 · 2 of 50 papers
Baselines
- CFG · 8 of 50 papers
- DPS · 4 of 50 papers
- classifier-free guidance · 4 of 50 papers
- FreeDoM · 3 of 50 papers
- PAG · 3 of 50 papers
- Diffuser · 3 of 50 papers
- guidance distillation · 2 of 50 papers
- MPGD · 2 of 50 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 121 papers; from OpenAlex and the papers' affiliations).
- Tsinghua University · 5 papers
- Google · 4 papers
- University of Toronto · 4 papers
- Hong Kong University of Science and Technology · 4 papers
- Carnegie Mellon University · 4 papers
- NVIDIA · 4 papers
- Microsoft · 3 papers
- Stanford University · 3 papers
- Princeton University · 3 papers
- Meta · 2 papers
Use it
Download BibTeX of 25 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)
- 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).