Graph Classification on COLLAB (test)
85.5AccuracyGSN
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GSN2021.02 | 85.5 | — | — | — | |
| GSN-eSource=Bouritsas et al. [5]2026.05 | 85.5 | — | — | — | |
| PXGL-GNNlearning_mode=supervised2025.12 | 83.96 | — | — | — | |
| GNTK2021.02 | 83.6 | — | — | — | |
| GNTKSource=Bouritsas et al. [5]2026.05 | 83.6 | — | — | — | |
| LGANresidual=false2025.12 | 82.8 | — | — | — | |
| SANlearning_mode=supervised2025.12 | 82.73 | — | — | — | |
| LGAN-resresidual=true2025.12 | 82.7 | — | — | — | |
| GSN-vSource=Bouritsas et al. [5]2026.05 | 82.7 | — | — | — | |
| SignGT2023.10 | 82.4 | — | — | — | |
| S2GAElearning_mode=supervised2025.12 | 82.35 | — | — | — | |
| att-Pooling-NLSFsArchitecture=Pooling2024.06 | 82.19 | — | — | — | |
| Graphormer2023.10 | 81.8 | — | — | — | |
| ICLlearning_mode=supervised2025.12 | 81.45 | — | — | — | |
| PPGNcategory=k-WL-based2025.12 | 81.4 | — | — | — | |
| PPGNSource=Bouritsas et al. [5]2026.05 | 81.4 | — | — | — | |
| SubGNNlearning_mode=supervised2025.12 | 81.26 | — | — | — | |
| 3WLGNN2021.02 | 80.7 | — | — | — | |
| SUM-1-LAYERAggregation=SUM, Architecture=1-layer2018.10 | 80.6 | — | — | — | |
| CRAWL2021.02 | 80.4 | — | — | — | |
| GIN+BATCHNORMBackbone=GIN, Normalization=BatchNorm2020.09 | 80.2 | — | — | — | |
| GIN+GraphNormBackbone=GIN, Normalization=GraphNorm2020.09 | 80.2 | — | — | — | |
| SUM-MLP (GIN-0)Aggregation=SUM, Architecture=MLP2018.10 | 80.2 | — | — | — | |
| GIN2021.02 | 80.2 | — | — | — | |
| GINcategory=GNNs2025.12 | 80.2 | — | — | — | |
| GINSource=Bouritsas et al. [5]2026.05 | 80.2 | — | — | — | |
| DiffPool2024.06 | 80.16 | — | — | — | |
| GIN+LAYERNORMBackbone=GIN, Normalization=LayerNorm2020.09 | 80.1 | — | — | — | |
| SUM-MLP (GIN-ε)Aggregation=SUM, Architecture=MLP2018.10 | 80.1 | — | — | — | |
| ChebNetII2024.06 | 80.06 | — | — | — | |
| SAT2023.10 | 80.05 | — | — | — | |
| GIN+INSTANCENORMBackbone=GIN, Normalization=InstanceNorm2020.09 | 80 | — | — | — | |
| GCN2024.06 | 79.98 | — | — | — | |
| SAGNNlearning_mode=supervised2025.12 | 79.94 | — | — | — | |
| GraphTrans2023.10 | 79.81 | — | — | — | |
| WEGL2021.02 | 79.8 | — | — | — | |
| agg + I2026.05 | 79.74 | — | — | — | |
| GraphSAGE2023.10 | 79.7 | — | — | — | |
| GT2023.10 | 79.63 | — | — | — | |
| MEAN-MLPAggregation=MEAN, Architecture=MLP2018.10 | 79.2 | — | — | — | |
| att-Graph-level NLSFsArchitecture=Graph-level2024.06 | 79.06 | — | — | — | |
| MEAN-1-LAYER (GCN)Aggregation=MEAN, Architecture=1-layer2018.10 | 79 | — | — | — | |
| WL SUBTREEBackbone=WL SUBTREE2020.09 | 78.9 | — | — | — | |
| WL subtree2018.10 | 78.9 | — | — | — | |
| WL-Kernel2021.02 | 78.9 | — | — | — | |
| ICGMM^fSource=Castellana et al. [7]2026.05 | 78.9 | — | — | — | |
| WL kernelSource=Bouritsas et al. [5]2026.05 | 78.9 | — | — | — | |
| DiffPoollearning_mode=supervised2025.12 | 78.77 | — | — | — | |
| ICGMM^f_αγSource=Castellana et al. [7]2026.05 | 78.6 | — | — | — | |
| ARMA2024.06 | 78.34 | — | — | — | |
| IGNSource=Bouritsas et al. [5]2026.05 | 78.3 | — | — | — | |
| GPRGNN2024.06 | 77.93 | — | — | — | |
| WL2017.07 | 77.82 | 1.45 | — | — | |
| DGCNNlearning_mode=supervised2025.12 | 77.59 | — | — | — | |
| E-CGMMSource=Castellana et al. [7]2026.05 | 77.45 | — | — | — | |
| ChebNet2024.06 | 77.4 | — | — | — | |
| CGMMSource=Castellana et al. [7]2026.05 | 77.32 | — | — | — | |
| GRAPHSAGElearning_mode=supervised2025.12 | 76.37 | — | — | — | |
| CayleyNet2024.06 | 76.33 | — | — | — | |
| WL2024.06 | 76.23 | — | — | — | |
| GINlearning_mode=supervised2025.12 | 75.83 | — | — | — | |
| GAT2023.10 | 75.8 | — | — | — | |
| GINSource=Errica et al. [16]2026.05 | 75.6 | — | — | — | |
| F-GINSupervision=Fully-supervised, Downstream Classifier=Linear2021.06 | 74.82 | — | — | — | |
| GAT2024.06 | 74.27 | — | — | — | |
| SpecMatch-CLtraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 74.26 | — | — | — | |
| AWLBackbone=AWL2020.09 | 73.9 | — | — | — | |
| AWL2018.10 | 73.9 | — | — | — | |
| GraphSAGESource=Errica et al. [16]2026.05 | 73.9 | — | — | — | |
| APPNP2024.06 | 73.85 | — | — | — | |
| DGCNN2021.02 | 73.8 | — | — | — | |
| DGCNNcategory=GNNs2025.12 | 73.8 | — | — | — | |
| DGCNNSource=Bouritsas et al. [5]2026.05 | 73.8 | — | — | — | |
| DGCNN2017.07 | 73.76 | 0.49 | — | — | |
| DGCNNBackbone=DGCNN2020.09 | 73.7 | — | — | — | |
| DGCNN2018.10 | 73.7 | — | — | — | |
| SAGE2024.06 | 73.61 | — | — | — | |
| MSSGCLtraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 73.48 | — | — | — | |
| Graphlet2017.07 | 73.42 | 2.43 | — | — | |
| AD-GCL-FIXSupervision=Unsupervised, Downstream Classifier=Linear2021.06 | 73.32 | — | — | — | |
| AD-GCL-OPTSupervision=Unsupervised, Downstream Classifier=Linear2021.06 | 73.32 | — | — | — | |
| Deep GK2017.07 | 73.09 | 0.25 | — | — | |
| SAGPool2023.10 | 73.08 | — | — | — | |
| PATCHYSAN2018.10 | 72.6 | — | — | — | |
| PSCNk=102017.07 | 72.6 | 2.15 | — | — | |
| PATCHY-SANcategory=GNNs2025.12 | 72.6 | — | — | — | |
| GK2024.06 | 72.32 | — | — | — | |
| CuCotraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 72.3 | — | — | — | |
| NAD-GCL-OPTSupervision=Unsupervised, Downstream Classifier=Linear2021.06 | 72.04 | — | — | — | |
| GCN2023.10 | 71.92 | — | — | — | |
| SimGRACEtraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 71.72 | — | — | — | |
| NAD-GCL-FIXSupervision=Unsupervised, Downstream Classifier=Linear2021.06 | 71.56 | — | — | — | |
| GraphCLtraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 71.36 | — | — | — | |
| 2D CNN2017.07 | 71.33 | 1.96 | — | — | |
| GraphCLSupervision=Unsupervised, Downstream Classifier=Linear2021.06 | 71.26 | — | — | — | |
| DGCNNSource=Errica et al. [16]2026.05 | 71.2 | — | — | — | |
| SortPool2023.10 | 71.18 | — | — | — | |
| graph2vectraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 71.1 | — | — | — | |
| InfoGraphtraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 70.65 | — | — | — | |
| InfoGraphSupervision=Unsupervised, Downstream Classifier=Linear2021.06 | 70.35 | — | — | — |