Graph Classification on IMDB-B (test)
77.8AccuracyGSN-e
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GSN-eSource=Bouritsas et al. [5]2026.05 | 77.8 | — | |
| PXGL-GNNlearning_mode=supervised2025.12 | 77.35 | — | |
| ChebNetII2024.06 | 77.09 | — | |
| GNTKSource=Bouritsas et al. [5]2026.05 | 76.9 | — | |
| GSN2024.02 | 76.8 | — | |
| GSN-vSource=Bouritsas et al. [5]2026.05 | 76.8 | — | |
| att-Pooling-NLSFsArchitecture=Pooling2024.06 | 76.78 | — | |
| LGANresidual=false2025.12 | 76.7 | — | |
| TOGL2024.02 | 76.65 | — | |
| GraphSAGE2024.02 | 76.6 | — | |
| LGAN-resresidual=true2025.12 | 76.6 | — | |
| DiffPool-TIPTIP=true2024.02 | 76.4 | — | |
| GIN+GraphNormBackbone=GIN, Normalization=GraphNorm2020.09 | 76 | — | |
| GraphTCL2025.12 | 75.8 | — | |
| S2GAElearning_mode=supervised2025.12 | 75.77 | — | |
| CW Networkscategory=k-WL-based2025.12 | 75.6 | — | |
| WKPI-kCvariant=kC2019.04 | 75.4 | — | |
| SANlearning_mode=supervised2025.12 | 75.27 | — | |
| MinCutPool-TIPTIP=true2024.02 | 75.2 | — | |
| GIN+BATCHNORMBackbone=GIN, Normalization=BatchNorm2020.09 | 75.1 | — | |
| SUM-MLP (GIN-0)Aggregation=SUM, Architecture=MLP2018.10 | 75.1 | — | |
| GIN2019.04 | 75.1 | — | |
| GINcategory=GNNs2025.12 | 75.1 | — | |
| GINSource=Bouritsas et al. [5]2026.05 | 75.1 | — | |
| GIN+INSTANCENORMBackbone=GIN, Normalization=InstanceNorm2020.09 | 74.8 | — | |
| TopoGCL2025.12 | 74.67 | — | |
| SAGNNlearning_mode=supervised2025.12 | 74.53 | — | |
| AWLBackbone=AWL2020.09 | 74.5 | — | |
| GIN+LAYERNORMBackbone=GIN, Normalization=LayerNorm2020.09 | 74.5 | — | |
| AWL2018.10 | 74.5 | — | |
| Graclus2024.02 | 74.45 | — | |
| SUM-MLP (GIN-ε)Aggregation=SUM, Architecture=MLP2018.10 | 74.3 | — | |
| att-Graph-level NLSFsArchitecture=Graph-level2024.06 | 74.26 | — | |
| DMONPool-TIPTIP=true2024.02 | 74.25 | — | |
| GCN2024.02 | 74.2 | — | |
| 1-2-3 GNNcategory=k-WL-based2025.12 | 74.2 | — | |
| SUM-1-LAYERAggregation=SUM, Architecture=1-layer2018.10 | 74.1 | — | |
| MEAN-1-LAYER (GCN)Aggregation=MEAN, Architecture=1-layer2018.10 | 74 | — | |
| GRAPHSAGElearning_mode=supervised2025.12 | 73.86 | — | |
| WL SUBTREEBackbone=WL SUBTREE2020.09 | 73.8 | — | |
| MinCutPoolTIP=false2024.02 | 73.8 | — | |
| WL subtree2018.10 | 73.8 | — | |
| WL kernelSource=Bouritsas et al. [5]2026.05 | 73.8 | — | |
| MEAN-MLPAggregation=MEAN, Architecture=MLP2018.10 | 73.7 | — | |
| DiffPoollearning_mode=supervised2025.12 | 73.55 | — | |
| DMONPoolTIP=false2024.02 | 73.5 | — | |
| Natural GNSource=Bouritsas et al. [5]2026.05 | 73.5 | — | |
| GMT2021.02 | 73.48 | — | |
| GPFPre-training Strategy=GCC (E2E), Tuning Strategy=GPF2022.09 | 73.4 | — | |
| SpecMatch-CLtraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 73.35 | — | |
| delta-2-LWLcategory=k-WL-based2025.12 | 73.3 | — | |
| ICLlearning_mode=supervised2025.12 | 73.29 | — | |
| ARMA2024.06 | 73.27 | — | |
| GCN2021.02 | 73.26 | — | |
| MAX-MLPAggregation=MAX, Architecture=MLP2018.10 | 73.2 | — | |
| DiffPoolTIP=false2024.02 | 73.15 | — | |
| ChebNet2024.06 | 73.14 | — | |
| MSSGCLtraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 73.14 | — | |
| InfoGraphtraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 73.03 | — | |
| P-WL-UC2019.04 | 73 | — | |
| TREE-G2022.07 | 73 | 2.4 | |
| PPGNcategory=k-WL-based2025.12 | 73 | — | |
| ICGMM^f_αγSource=Castellana et al. [7]2026.05 | 73 | — | |
| PPGNSource=Bouritsas et al. [5]2026.05 | 73 | — | |
| GCN2024.06 | 72.96 | — | |
| APPNP2024.06 | 72.94 | — | |
| WL2022.07 | 72.8 | 1.9 | |
| GIN2021.02 | 72.78 | — | |
| CGMMSource=Castellana et al. [7]2026.05 | 72.7 | — | |
| GIN2024.02 | 72.65 | — | |
| MinCutPool2021.02 | 72.65 | — | |
| TopK2024.02 | 72.6 | — | |
| TREE-Gpropagation depth (d)=0, active subset distance (a)=02022.07 | 72.6 | 2.4 | |
| FTPre-training Strategy=GCC (E2E), Tuning Strategy=FT2022.09 | 72.6 | — | |
| 2-WLcategory=Kernels2025.12 | 72.6 | — | |
| GINlearning_mode=supervised2025.12 | 72.52 | — | |
| TREE-Gactive subset distance (a)=02022.07 | 72.5 | 2.7 | |
| GPFPre-training Strategy=GCC (MoCo), Tuning Strategy=GPF2022.09 | 72.5 | — | |
| HaarPool2021.02 | 72.46 | — | |
| CuCotraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 72.33 | — | |
| MAX-1-LAYER (GraphSAGE)Aggregation=MAX, Architecture=1-layer2018.10 | 72.3 | — | |
| WLtraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 72.3 | — | |
| DGCNNlearning_mode=supervised2025.12 | 72.19 | — | |
| ASAP2021.02 | 72.06 | — | |
| IGNSource=Bouritsas et al. [5]2026.05 | 72 | — | |
| RetGK2019.04 | 71.9 | — | |
| ICGMM^fSource=Castellana et al. [7]2026.05 | 71.8 | — | |
| FTPre-training Strategy=GCC (MoCo), Tuning Strategy=FT2022.09 | 71.7 | — | |
| WL2017.07 | 71.6 | 5.16 | |
| SubGNNlearning_mode=supervised2025.12 | 71.58 | — | |
| GATv22022.07 | 71.5 | 1.2 | |
| CayleyNet2024.06 | 71.45 | — | |
| WL2024.06 | 71.42 | — | |
| agg + I2026.05 | 71.4 | — | |
| SimGRACEtraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 71.26 | — | |
| GIN2022.07 | 71.2 | 3.4 | |
| GINSource=Errica et al. [16]2026.05 | 71.2 | — | |
| GraphCLtraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 71.14 | — | |
| graph2vectraining_regime=unsupervised, evaluation_protocol=10-fold cross-validation with linear SVM2025.11 | 71.1 | — | |
| DiffPool2024.06 | 71.09 | — |