Research map / Graph learning / GNN architectures
Simple and scalable GNNs
Replace or simplify GNNs with label propagation, linear models, and subsampling.
131 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2018–2026) · an approach in GNN architectures, keyword cluster “gnn · explanations · trained”
State History Benchmarks Tried Baselines Checklist Preprints Who Export
Where it stands
Leveling off Several papers from 2025–26 still report on 1 tracked benchmark, but the best result has barely moved since 2024.
45 papers in 2023–24 and 35 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 9.5% to 6.9%.
Open questions the data points to
- Papers report GIN on MUTAG (accuracy) anywhere from 74 to 93.5: comparisons against it depend on whose number you take.
- Papers report GAT on Pubmed (accuracy) anywhere from 78.1 to 86.83: comparisons against it depend on whose number you take.
What looks settled
- Squirrel (accuracy (%)): best up to 2024 65.25, best in 2025–26 43.61 from 2 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
- 2018 · Mean-field theory of graph neural networks in graph partitioning (NeurIPS 2018 · 28 citations)
- 2019 · Adaptive GNN for Image Analysis and Editing (NeurIPS 2019 · 5 citations)
- 2020 · Design Space for Graph Neural Networks (NeurIPS 2020 · 166 citations)
- 2021 · Graph Neural Networks with Heterophily (AAAI 2021 · 218 citations)
- 2022 · Revisiting Heterophily For Graph Neural Networks (NeurIPS 2022 · 70 citations)
- 2023 · A critical look at the evaluation of GNNs under heterophily: Are we really making progress? (ICLR 2023 · 21 citations)
- 2024 · Dynamic Spiking Graph Neural Networks (AAAI 2024 · 28 citations)
- 2025 · Modality-Independent Graph Neural Networks with Global Transformers for Multimodal Recommendation (AAAI 2025 · 20 citations)
- 2026 · Beyond Fixed Depth: Adaptive Graph Neural Networks for Node Classification Under Varying Homophily (AAAI 2026 · 1 citation)
Earliest papers here
- Mean-field theory of graph neural networks in graph partitioning (NeurIPS 2018 · 28 citations)
- Adaptive GNN for Image Analysis and Editing (NeurIPS 2019 · 5 citations)
- Design Space for Graph Neural Networks (NeurIPS 2020 · 166 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)
18 papers, 2020–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 83.9 | Scalable Graph Neural Networks via Bidirectional Propagation (NeurIPS 2020 · 53 citations) |
| 2021 | 89.77 | Combining Label Propagation and Simple Models out-performs Graph Neural Networks (ICLR 2021 · 127 citations) |
| 2023 | 94.66 | Empowering Graph Representation Learning with Test-Time Graph Transformation (ICLR 2023 · 10 citations) |
Citeseer · accuracy (higher is better)
15 papers, 2020–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 72.9 | Scalable Graph Neural Networks via Bidirectional Propagation (NeurIPS 2020 · 53 citations) |
| 2021 | 79.61 | Graph Differentiable Architecture Search with Structure Learning (NeurIPS 2021 · 18 citations) |
| 2024 | 80.49 | DFA-GNN: Forward Learning of Graph Neural Networks by Direct Feedback Alignment (NeurIPS 2024 · 1 citation) |
| 2025 | 81.25 | GRAIN: Multi-Granular and Implicit Information Aggregation Graph Neural Network for Heterophilous Graphs (AAAI 2025 · 3 citations) |
Pubmed · accuracy (higher is better)
14 papers, 2020–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 80.6 | Scalable Graph Neural Networks via Bidirectional Propagation (NeurIPS 2020 · 53 citations) |
| 2021 | 90.52 | Graph Differentiable Architecture Search with Structure Learning (NeurIPS 2021 · 18 citations) |
Cornell · accuracy (%) (higher is better)
11 papers, 2021–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 91.36 | Adaptive Universal Generalized PageRank Graph Neural Network (ICLR 2021 · 92 citations) |
| 2022 | 95.9 | Revisiting Heterophily For Graph Neural Networks (NeurIPS 2022 · 70 citations) |
PROTEINS · accuracy (higher is better)
10 papers, 2020–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 82.25 | Memory-Based Graph Networks (ICLR 2020 · 11 citations) |
Squirrel · accuracy (%) (higher is better)
10 papers, 2021–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 49.93 | Adaptive Universal Generalized PageRank Graph Neural Network (ICLR 2021 · 92 citations) |
| 2022 | 59.11 | Graph Pointer Neural Networks (AAAI 2022 · 39 citations) |
| 2024 | 65.25 | Graph Neural Networks with Soft Association between Topology and Attribute (AAAI 2024 · 18 citations) |
What has been tried
105 papers here report 1,106 measured changes (a component added, removed or swapped, compared with a baseline): helped 67%, no clear effect 8%, hurt 19%, 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 |
|---|---|---|---|---|
| Vary λ from smaller to larger values | Dynamic node classification · accuracy | Mixed | Dynamic Spiking Graph Neural Networks (AAAI 2024 · 28 citations) | |
| add Sampling-based Global Transformer (SGT) to MIRF | Baby, Sports, and Clothing · NDCG@20 | Helped | Modality-Independent Graph Neural Networks with Global Transformers for Multimodal Recommendation (AAAI 2025 · 20 citations) | |
| add soft association between attribute and topology | Squirrel · accuracy | 35.23 -> 65.25 | Helped | Graph Neural Networks with Soft Association between Topology and Attribute (AAAI 2024 · 18 citations) |
| Add bi-level optimization | D&D · validation loss | Helped | Learning to Reweight for Generalizable Graph Neural Network (AAAI 2024 · 18 citations) | |
| Use product embeddings rather than directly learning inventory attention weights | SS-std production learning · MAP | 0.771 (0.005) -> 0.790 (0.005) | Helped | Learning Production Functions for Supply Chains with Graph Neural Networks (AAAI 2025 · 4 citations) |
| Add a second set of feature functions after graph representation aggregation | graph tasks · performance | No clear effect | The Intelligible and Effective Graph Neural Additive Network (NeurIPS 2024 · 4 citations) | |
| use static or dynamic random node permutation instead of natural ordering | large node prediction datasets · Accuracy | Flickr: 55.55 ± 0.25, 55.73 ± 0.27, 55.80 ± 0.38; Arxiv: 73.10 ± 0.24, 73.08 ± 0.28, 73.12 ± 0.22; Reddit: 96.65 ± 0.05, | No clear effect | A Scalable and Effective Alternative to Graph Transformers (AAAI 2025 · 3 citations) |
| Set aggregation parameter α to 0.2 rather than other tested values | Cora, Pubmed, Chameleon, Cornell · accuracy | Helped | GRAIN: Multi-Granular and Implicit Information Aggregation Graph Neural Network for Heterophilous Graphs (AAAI 2025 · 3 citations) | |
| add edge feature integration | ogbg-ppa · accuracy | 0.7310 to 0.5083 | Helped | Can Classic GNNs Be Strong Baselines for Graph-level Tasks? Simple Architectures Meet Excellence (ICML 2025 · 3 citations) |
| replace MMD with Wasserstein distance | eight graph classification benchmarks · accuracy (%) | MUTAG 92.1±5.9 -> 90.8±4.6; PROTEINS 82.6±1.2 -> 82.1±0.9; IMDB-M 52.9±1.8 -> 53.1±0.9 | Mixed | Graph Classification via Reference Distribution Learning: Theory and Practice (NeurIPS 2024 · 3 citations) |
| Add initial compensation based on LDC | Cora and Texas, 10 and 20 layers · node classification accuracy | Helped | Improving Distinguishability of Class for Graph Neural Networks (AAAI 2024 · 2 citations) | |
| Use hyperspherical S^32 instead of hyperbolic H^32 | Computers · accuracy (%) | 89.27{ ±0.19} -> 87.84{ ±0.77} | Hurt | Spiking Graph Neural Network on Riemannian Manifolds (NeurIPS 2024 · 2 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 |
|---|---|---|---|---|
| GIN | MUTAG | accuracy | 74 – 93.5 | 4 papers, e.g. NeurIPS 2022 and AAAI 2020 |
| GAT | Pubmed | accuracy | 78.1 – 86.83 | 2 papers, e.g. ICLR 2022 and NeurIPS 2021 |
| GCN | ogbn-products | accuracy | 71.35 – 78.97 | 2 papers, e.g. ICLR 2023 and ICLR 2022 |
| ResGCN | ogbn-arxiv | accuracy | 66.32 – 72.46 | 2 papers, e.g. NeurIPS 2024 and ICML 2021 |
| APPNP | PubMed | accuracy (%) | 80.1 – 85.02 | 2 papers, e.g. NeurIPS 2020 and ICLR 2021 |
What reviewers will expect
Datasets and baselines named in the 24 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 · 6 of 24 papers
- PubMed · 6 of 24 papers
- CiteSeer · 5 of 24 papers
- ogbn-arxiv · 4 of 24 papers
- Photo · 4 of 24 papers
- Squirrel · 4 of 24 papers
- Peptides-func · 3 of 24 papers
- Peptides-struct · 3 of 24 papers
Baselines
- GCN · 5 of 24 papers
- GAT · 5 of 24 papers
- GraphGPS · 4 of 24 papers
- Exphormer · 3 of 24 papers
- Graphormer · 3 of 24 papers
- MLP · 3 of 24 papers
- GOAT · 2 of 24 papers
- SAN · 2 of 24 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 80 papers; from OpenAlex and the papers' affiliations).
- The Chinese University of Hong Kong · 6 papers
- National University of Singapore · 4 papers
- Nanyang Technological University · 4 papers
- Peking University · 4 papers
- Harbin Institute of Technology · 3 papers
- Hebei University of Technology · 3 papers
- Jilin University · 3 papers
- Microsoft · 3 papers
- Tsinghua University · 3 papers
- Northwestern Polytechnical University · 3 papers
Use it
Download BibTeX of 40 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
- GNN generalization theory (187 papers)
- Robustness, privacy, and fairness of GNNs (94 papers)
- Physics-inspired GNNs (81 papers)
- GNN explainability (64 papers)
- Graph structure learning (201 papers)
Elsewhere in the topic
- Expressive equivariant message passing (109 papers)
- Theory of graph convolutional networks (83 papers)
- Deep graph convolution and simplification (82 papers)
- Graph rewiring and scalable message passing (72 papers)
- Weisfeiler-Leman expressivity hierarchies (66 papers)
- Structural GNN variants (65 papers)
- Oversmoothing and heterophily (60 papers)
- Substructure-aware GNN expressivity (50 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).