Research map / Image generation / GANs
Generative adversarial network theory
Analyzes and stabilizes GAN training through kernels, Langevin dynamics and structure preservation.
62 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in GANs, keyword cluster “gans · mode · samples”
State History Benchmarks Tried Baselines Preprints 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.
4 papers in 2023–24 and 2 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 6.9% to 8.0%.
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
- 2016 · f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization (NeurIPS 2016)
- 2017 · Wasserstein Generative Adversarial Networks (ICML 2017 · 4,925 citations)
- 2018 · Generative Modeling Using the Sliced Wasserstein Distance (CVPR 2018 · 179 citations)
- 2019 · Improving Generalization and Stability of Generative Adversarial Networks (ICLR 2019 · 110 citations)
- 2020 · A Closer Look at the Optimization Landscapes of Generative Adversarial Networks (ICLR 2020 · 31 citations)
- 2021 · Partition-Guided GANs (CVPR 2021 · 13 citations)
- 2022 · Manifold Learning Benefits GANs (CVPR 2022 · 18 citations)
- 2023 · Non-adversarial training of Neural SDEs with signature kernel scores (NeurIPS 2023 · 7 citations)
- 2024 · Neuroevolution of a Multi-Generator GAN (Student Abstract) (AAAI 2024 · 3 citations)
- 2025 · Discrete Distribution Networks (ICLR 2025 · 0 citations)
- 2026 · Who Said Neural Networks Aren't Linear? (ICML 2026 · 0 citations)
Earliest papers here
- f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization (NeurIPS 2016)
- Wasserstein Generative Adversarial Networks (ICML 2017 · 4,925 citations)
- MMD GAN: Towards Deeper Understanding of Moment Matching Network (NeurIPS 2017 · 414 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)
5 papers, 2019–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2019 | 17.2 | Adaptive Density Estimation for Generative Models (NeurIPS 2019 · 4 citations) |
| 2021 | 8.93 | Partition-Guided GANs (CVPR 2021 · 13 citations) |
What has been tried
58 papers here report 574 measured changes (a component added, removed or swapped, compared with a baseline): helped 63%, no clear effect 11%, 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 the 1G2CGAN strategy of mutating the last two convolutional layers | MNIST · image quality | Hurt | Neuroevolution of a Multi-Generator GAN (Student Abstract) (AAAI 2024 · 3 citations) | |
| reduce K from 512 to 64 | FFHQ-64x64 · FID | 43.1 -> 47.0 | Hurt | Discrete Distribution Networks (ICLR 2025 · 0 citations) |
| Use the transport cost function rather than removing cost | CIFAR-10 · recall | 0.13 -> 0.62 | Helped | Analyzing and Improving Optimal-Transport-based Adversarial Networks (ICLR 2024 · 0 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 |
|---|---|---|---|---|
| DCGAN | CIFAR-10 | FID | 33.71 – 37.1 | 2 papers, e.g. AAAI 2020 and ICLR 2025 |
Recent preprints
Loading recent arXiv preprints…
Use it
Download BibTeX of 24 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
- Stabilized GAN training (144 papers)
- GAN latent optimization and evaluation (52 papers)
- GANs for text generation (21 papers)
Elsewhere in the topic
- Large-scale GAN latent analysis (108 papers)
- GAN-based image manipulation (90 papers)
- Conditional GANs (25 papers)
- GAN vocoders and audio synthesis (13 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).