Research map / Image generation / Autoregressive and latent generation
High-resolution diffusion synthesis
Scale diffusion models to high resolutions with position extrapolation and efficient architectures.
140 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2022–2026) · an approach in Autoregressive and latent generation, keyword cluster “latent space · modeling · vae”
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.
38 papers in 2023–24 and 101 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 8.0% to 11.2%.
Open questions the data points to
- ImageNet 256×256 (FID): best went from 2.29 (up to 2024) to 1.35 in 2025 — Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models (CVPR 2025 · 53 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
- 2022 · High-Resolution Image Synthesis With Latent Diffusion Models (CVPR 2022 · 15,775 citations)
- 2023 · All Are Worth Words: A ViT Backbone for Diffusion Models (CVPR 2023 · 256 citations)
- 2024 · GenTron: Diffusion Transformers for Image and Video Generation (CVPR 2024 · 40 citations)
- 2025 · Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models (CVPR 2025 · 53 citations)
- 2026 · TIDE: Temporal-Aware Sparse Autoencoders for Interpretable Diffusion Transformers in Image Generation (AAAI 2026 · 1 citation)
Earliest papers here
- High-Resolution Image Synthesis With Latent Diffusion Models (CVPR 2022 · 15,775 citations)
- All Are Worth Words: A ViT Backbone for Diffusion Models (CVPR 2023 · 256 citations)
- High-Resolution Image Reconstruction With Latent Diffusion Models From Human Brain Activity (CVPR 2023 · 236 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 256×256 · FID (lower is better)
4 papers, 2023–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2023 | 2.29 | All Are Worth Words: A ViT Backbone for Diffusion Models (CVPR 2023 · 256 citations) |
| 2025 | 1.35 | Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models (CVPR 2025 · 53 citations) |
What has been tried
68 papers here report 796 measured changes (a component added, removed or swapped, compared with a baseline): helped 71%, no clear effect 3%, hurt 23%, 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 DINOv2 rather than no VF loss | ImageNet 256×256 · gFID | 29.35 -> 24.77 | Helped | Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models (CVPR 2025 · 53 citations) |
| replace adaLN-zero conditioning with cross-attention | T2I-CompBench · mean score | 34.32 -> 47.84 | Helped | GenTron: Diffusion Transformers for Image and Video Generation (CVPR 2024 · 40 citations) |
| add deep context fusion instead of shallow RGB-pixel fusion | UCF-101, 64×256² · FVD@512 | 467.0 -> 378.2 | Helped | Hierarchical Patch Diffusion Models for High-Resolution Video Generation (CVPR 2024 · 21 citations) |
| add the learned wavelet instead of using Biorthogonal 6-8 | DeepFashion3D · COV-CD | 64.52 -> 69.62 | Helped | UDiFF: Generating Conditional Unsigned Distance Fields with Optimal Wavelet Diffusion (CVPR 2024 · 17 citations) |
| Add the discriminative objective and teacher branch | ImageNet 256×256, 400k steps · FID | 62.84 -> 53.72 | Helped | SD-DiT: Unleashing the Power of Self-supervised Discrimination in Diffusion Transformer (CVPR 2024 · 16 citations) |
| Increase model size from 0.1B to 3.4B | Colorization · MSE | 0.609 -> 0.273 | Helped | LaVin-DiT: Large Vision Diffusion Transformer (CVPR 2025 · 15 citations) |
| add the optimized training strategy | ImageNet-256 · FID-50K | 5.21 -> 4.30 | Helped | Revisiting Non-Autoregressive Transformers for Efficient Image Synthesis (CVPR 2024 · 14 citations) |
| remove the truncation function | ShapeNet chair reconstruction · PSNR | 34.01 -> 29.39 | Hurt | DiffGS: Functional Gaussian Splatting Diffusion (NeurIPS 2024 · 11 citations) |
| Use sigmoid loss weighting instead of EDM-monotonic or other tested weightings | ImageNet512 · FID | Helped | Simpler Diffusion: 1.5 FID on ImageNet512 with Pixel-space Diffusion (CVPR 2025 · 10 citations) | |
| use multi-resolution diffusion loss instead of a single-resolution loss | ImageNet 256×256 · FID | Helped | Matryoshka Diffusion Models (ICLR 2024 · 10 citations) | |
| Use cosine-based masking scheduling instead of piecewise scheduling | ImageNet-1K · MLT | 8.37 -> 6.92 | Helped | LMD: Faster Image Reconstruction with Latent Masking Diffusion (AAAI 2024 · 9 citations) |
| infer embeddings from the low-resolution image instead of using ground-truth embeddings | TCGA-BRCA 4× super-resolution · LPIPS | 0.250 -> 0.229 | Helped | ZoomLDM: Latent Diffusion Model for Multi-scale Image Generation (CVPR 2025 · 7 citations) |
What reviewers will expect
Datasets and baselines named in the 49 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 · 13 of 49 papers
- FFHQ · 4 of 49 papers
- ImageNet-1K · 4 of 49 papers
- MSCOCO · 4 of 49 papers
- CelebA-HQ · 3 of 49 papers
- AFHQ · 3 of 49 papers
- MJHQ-30K · 3 of 49 papers
- TCGA-BRCA · 2 of 49 papers
Baselines
- DiT-XL/2 · 7 of 49 papers
- REPA · 6 of 49 papers
- DemoFusion · 5 of 49 papers
- HiDiffusion · 4 of 49 papers
- MaskDiT · 4 of 49 papers
- DiT · 4 of 49 papers
- ToMeSD · 3 of 49 papers
- AccDiffusion · 3 of 49 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 139 papers; from OpenAlex and the papers' affiliations).
- Tsinghua University · 9 papers
- Nanjing University · 5 papers
- National University of Singapore · 4 papers
- The Chinese University of Hong Kong · 4 papers
- Google · 4 papers
- ByteDance · 4 papers
- Huawei · 4 papers
- Peking University · 4 papers
- Stanford University · 3 papers
- Korea Advanced Institute of Science and Technology · 3 papers
Use it
Download BibTeX of 28 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
- Generative latent space tooling (129 papers)
- Latent diffusion with learned tokenizers (114 papers)
- Hierarchical VQ-VAE generation (62 papers)
- Next-scale and token autoregressive images (90 papers)
- Accelerating autoregressive image generation (22 papers)
Elsewhere in the topic
- Classical generative image synthesis (147 papers)
- Retrieval and matching with generative features (140 papers)
- Generalizable fake image detection (119 papers)
- Synthetic training data from generators (111 papers)
- Compositional scene generation (105 papers)
- Robust diffusion watermarking (79 papers)
- Neural text-to-speech synthesis (72 papers)
- Equivariant diffusion for 3D molecules (71 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).