Research map / Graph learning / Graph self-supervised learning
Graph domain adaptation
Transfer knowledge across graphs with spectral regularization and foundation models.
82 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Graph self-supervised learning, keyword cluster “detection · anomaly · ood”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 1 of 2 tracked benchmarks.
21 papers in 2023–24 and 46 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 6.1% to 7.9%.
Open questions the data points to
- PubMed (accuracy): best went from 77.19 (up to 2024) to 78.28 in 2025 — AutoGFM: Automated Graph Foundation Model with Adaptive Architecture Customization (ICML 2025)
What looks settled
- Cora (accuracy): best up to 2024 85.4, best in 2025–26 80.32 from 3 papers, at most 0.2% better, if at all.
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 · Transfer Learning for Deep Learning on Graph-Structured Data (AAAI 2017 · 58 citations)
- 2018 · Graph Convolutional Networks With Argument-Aware Pooling for Event Detection (AAAI 2018 · 397 citations)
- 2019 · Learning Transferable Graph Exploration (NeurIPS 2019 · 3 citations)
- 2020 · Open Graph Benchmark: Datasets for Machine Learning on Graphs (NeurIPS 2020 · 484 citations)
- 2021 · Uncertain Graph Neural Networks for Facial Action Unit Detection (AAAI 2021 · 79 citations)
- 2022 · SIGMA: Semantic-Complete Graph Matching for Domain Adaptive Object Detection (CVPR 2022 · 232 citations)
- 2023 · Non-IID Transfer Learning on Graphs (AAAI 2023 · 29 citations)
- 2024 · Data-Augmented Curriculum Graph Neural Architecture Search under Distribution Shifts (AAAI 2024 · 17 citations)
- 2025 · Behavior Importance-Aware Graph Neural Architecture Search for Cross-Domain Recommendation (AAAI 2025 · 4 citations)
- 2026 · Graph Domain Adaptation via Homophily-Agnostic Reconstructing Structure (AAAI 2026 · 1 citation)
Earliest papers here
- Transfer Learning for Deep Learning on Graph-Structured Data (AAAI 2017 · 58 citations)
- Graph Convolutional Networks With Argument-Aware Pooling for Event Detection (AAAI 2018 · 397 citations)
- Learning Transferable Graph Exploration (NeurIPS 2019 · 3 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.
Cora · accuracy (higher is better)
5 papers, 2022–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 85.4 | Knowledge Distillation Improves Graph Structure Augmentation for Graph Neural Networks (NeurIPS 2022 · 1 citation) |
PubMed · accuracy (higher is better)
4 papers, 2024–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2024 | 77.19 | GFT: Graph Foundation Model with Transferable Tree Vocabulary (NeurIPS 2024 · 8 citations) |
| 2025 | 78.28 | AutoGFM: Automated Graph Foundation Model with Adaptive Architecture Customization (ICML 2025) |
What has been tried
55 papers here report 571 measured changes (a component added, removed or swapped, compared with a baseline): helped 69%, no clear effect 7%, hurt 18%, mixed 6%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add embedding-guided data generation and curriculum weighting | Spurious-Motif (b = 0.7) · accuracy | 72.44±17.13 -> 87.68±6.12 | Helped | Data-Augmented Curriculum Graph Neural Architecture Search under Distribution Shifts (AAAI 2024 · 17 citations) |
| add MMD alignment | D→A · Macro-F1 | 58.50 -> 67.09 | Helped | Rethinking Propagation for Unsupervised Graph Domain Adaptation (AAAI 2024 · 17 citations) |
| add multimodal graph feature decorrelation (MGFD) rather than single feature decorrelation (SFD) | Tencent, Amazon review, and Recipe Multi-OOD · accuracy | GCN: 61.00±1.25 -> 55.26±2.24; 52.80±9.20 -> 61.85±3.51; 61.65±5.05 -> 70.63±2.10; MGAT: 60.06±7.99 -> 65.66±1.41; 56.55 | Mixed | Multimodal Graph Neural Architecture Search under Distribution Shifts (AAAI 2024 · 9 citations) |
| Replace computation trees with subgraphs | Node classification · accuracy | 75.51 -> 74.23 | Hurt | GFT: Graph Foundation Model with Transferable Tree Vocabulary (NeurIPS 2024 · 8 citations) |
| Add GSKS to the FedSSP model without GSKS or PGPA | SM · accuracy | 74.12 -> 77.83 | Helped | FedSSP: Federated Graph Learning with Spectral Knowledge and Personalized Preference (NeurIPS 2024 · 8 citations) |
| Remove source-domain data | Four Amazon tasks · performance | average performance drop of 5.8% | Hurt | Behavior Importance-Aware Graph Neural Architecture Search for Cross-Domain Recommendation (AAAI 2025 · 4 citations) |
| Replace random-walk sampling with k-hop sampling | Six cross-domain tasks · Micro-F1 | Mixed | Smoothness Really Matters: A Simple Yet Effective Approach for Unsupervised Graph Domain Adaptation (AAAI 2025 · 3 citations) | |
| Add prototypical Mixup | Citations, AC→D / AD→C / CD→A · Micro F1 | Helped | TRACI: A Data-centric Approach for Multi-Domain Generalization on Graphs (AAAI 2025 · 3 citations) | |
| add personalized edge generators | Motif/Basis · accuracy | 60.20±1.75 -> 75.72±1.85 | Helped | Virtual Nodes Can Help: Tackling Distribution Shifts in Federated Graph Learning (AAAI 2025 · 3 citations) |
| aggregate independent, local, and group energies | CoraML, N(0,1) far-OOD · AUC-ROC | 1.4 -> 89.8 | Helped | Energy-based Epistemic Uncertainty for Graph Neural Networks (NeurIPS 2024 · 3 citations) |
| Remove the shared token | Most datasets · accuracy | Hurt | Multi-Domain Graph Foundation Models: Robust Knowledge Transfer via Topology Alignment (ICML 2025 · 2 citations) | |
| Inject a prompt graph rather than set τ=1.0 (no prompt graph injected) | 1-shot node classification; datasets vary · accuracy | Mixed | One Prompt Fits All: Universal Graph Adaptation for Pretrained Models (NeurIPS 2025 · 2 citations) |
What reviewers will expect
Datasets and baselines named in the 31 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- Cora · 7 of 31 papers
- CiteSeer · 5 of 31 papers
- ACMv9 · 4 of 31 papers
- PubMed · 4 of 31 papers
- Citationv1 · 3 of 31 papers
- DBLPv7 · 3 of 31 papers
- Airport · 2 of 31 papers
- ARXIV · 2 of 31 papers
Baselines
- GCN · 16 of 31 papers
- GAT · 11 of 31 papers
- UDAGCN · 7 of 31 papers
- GIN · 7 of 31 papers
- GCOPE · 5 of 31 papers
- DANE · 4 of 31 papers
- GRADE · 4 of 31 papers
- StruRW · 4 of 31 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 67 papers; from OpenAlex and the papers' affiliations).
- University of Illinois, Urbana Champaign · 5 papers
- Tsinghua University · 5 papers
- Beihang University · 5 papers
- University of California, Los Angeles · 4 papers
- Nanjing University · 3 papers
- Peking University · 3 papers
- National University of Singapore · 3 papers
- Zhejiang University · 3 papers
- University of Science and Technology of China · 3 papers
- Georgia Institute of Technology · 3 papers
Use it
Download BibTeX of 35 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
- Graph anomaly detection (101 papers)
- Graph out-of-distribution generalization (58 papers)
Elsewhere in the topic
- Semi-supervised node representation learning (214 papers)
- Heterophily and imbalance on graphs (178 papers)
- Diffusion-based graph generation (138 papers)
- LLMs and graph foundation models (117 papers)
- Molecular graph pretraining (101 papers)
- Brain and biological graph learning (78 papers)
- Multi-view graph clustering (70 papers)
- Graph contrastive learning analysis (61 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).