Graph Classification on MNIST
99.2AccuracyARMA
Evaluation results
| Method | Links | |
|---|---|---|
| ARMAGNN layer=ARMA2019.01 | 99.2 | |
| CayleyNetGNN layer=CayleyNet2019.01 | 99.18 | |
| ChebyshevGNN layer=Chebyshev2019.01 | 99.14 | |
| ESA+tuned=true2024.02 | 98.917 | |
| EXPHORMER2023.03 | 98.55 | |
| GCNGNN layer=GCN2019.01 | 98.48 | |
| LRGA + GatedGCNParameter Budget=486K2020.06 | 98.47 | |
| LRGA + GATParameter Budget=476K2020.06 | 98.41 | |
| LRGA + GCNParameter Budget=463K2020.06 | 98.34 | |
| Graph-ViT + GFSAGFSA=true2023.12 | 98.26 | |
| GatedGCNParameter Budget=500K2020.06 | 98.24 | |
| TIGT2024.02 | 98.23 | |
| TAU-GRITTokenizer=TAU, Backbone=GRIT2025.10 | 98.213 | |
| LRGA + GatedGCNParameter Budget=93K2020.06 | 98.2 | |
| Graph-ViTGFSA=false2023.12 | 98.2 | |
| EGT#Param=≈100K2021.08 | 98.173 | |
| EGT2024.02 | 98.173 | |
| EGT2025.10 | 98.173 | |
| EGT2023.03 | 98.17 | |
| GPS + GFSAGFSA=true2023.12 | 98.14 | |
| GRIT2024.02 | 98.108 | |
| GRIT2025.10 | 98.108 | |
| GPS2024.02 | 98.051 | |
| GraphGPS2025.10 | 98.051 | |
| GraphGPS2023.03 | 98.05 | |
| GPSGFSA=false2023.12 | 98.05 | |
| CRaW12024.02 | 97.944 | |
| PNA2023.03 | 97.94 | |
| CRaWI2023.03 | 97.94 | |
| PNA2024.02 | 97.94 | |
| Graphormer#Param=≈100K2021.08 | 97.905 | |
| Cy2C-GNNs2024.02 | 97.772 | |
| PNA#Param=≈100K2021.08 | 97.69 | |
| LRGA + GCNParameter Budget=91K2020.06 | 97.63 | |
| FGN2026.06 | 97.6 | |
| LoGoGNN2026.06 | 97.5 | |
| LRGA + GATParameter Budget=90K2020.06 | 97.47 | |
| GatedGCNedge features=none2022.05 | 97.4 | |
| Performer + GRF++Graph Features=GRF++2025.10 | 97.38 | |
| GatedGCNParameter Budget=104K2020.06 | 97.34 | |
| GatedGCN#Param=≈100K2021.08 | 97.34 | |
| GatedGCN2023.03 | 97.34 | |
| GatedGCN2024.02 | 97.34 | |
| GraphSage#Param=≈100K2021.08 | 97.312 | |
| SAGEReference=[36]2024.07 | 97.31 | |
| GatedGCN2026.06 | 97.3 | |
| GraphSAGE2026.06 | 97.3 | |
| GHC2024.07 | 97.24 | |
| PHM-GNNedge features=none2022.05 | 97.2 | |
| Performer + GRFGraph Features=GRF2025.10 | 97.14 | |
| Performer2025.10 | 97.02 | |
| GHM2024.07 | 96.74 | |
| GATParameter Budget=441K2020.06 | 96.5 | |
| tGNNedge features=none2022.05 | 96.5 | |
| GIN2023.03 | 96.49 | |
| GIN#Param=≈100K2021.08 | 96.485 | |
| GIN2024.02 | 96.485 | |
| No PromptingQuantization Framework=FP32, Model Architecture=GIN2026.01 | 96.4 | |
| GPF-LoRAPQuantization Framework=A2Q, Model Architecture=GIN2026.01 | 96.4 | |
| GPF-plusQuantization Framework=A2Q, Model Architecture=GIN2026.01 | 95.9 | |
| No PromptingQuantization Framework=FP32, Model Architecture=GAT2026.01 | 95.8 | |
| No PromptingQuantization Framework=A2Q, Model Architecture=GIN2026.01 | 95.7 | |
| GATedge features=none2022.05 | 95.6 | |
| GAT2023.03 | 95.54 | |
| GATReference=[36]2024.07 | 95.54 | |
| GAT#Param=≈100K2021.08 | 95.535 | |
| GAT2024.02 | 95.535 | |
| GATParameter Budget=110K2020.06 | 95.53 | |
| GAT2026.06 | 95.5 | |
| MLP2026.06 | 95.3 | |
| ALL-INTraining Scope=Trained on all datasets, Node Properties=Standard2026.05 | 95.22 | |
| DiffPooledge features=none2022.05 | 95 | |
| ALL-IN-SPECIALIZEDTraining Scope=Trained per dataset, Node Properties=Standard2026.05 | 94.77 | |
| ALL-INTraining Scope=Trained on all datasets, Node Properties=0 props2026.05 | 94.57 | |
| MLPedge features=none2022.05 | 94.5 | |
| ALL-IN-SPECIALIZEDTraining Scope=Trained per dataset, Node Properties=0 props2026.05 | 94.03 | |
| GPF-LoRAPQuantization Framework=DQ, Model Architecture=GAT2026.01 | 94 | |
| GINedge features=none2022.05 | 93.9 | |
| GPF-plusQuantization Framework=DQ, Model Architecture=GAT2026.01 | 93.1 | |
| GPF-LoRAPQuantization Framework=A2Q, Model Architecture=GAT2026.01 | 93.1 | |
| GPF-LoRAPQuantization Framework=DQ, Model Architecture=GIN2026.01 | 92.9 | |
| MLP2024.07 | 92.53 | |
| No PromptingQuantization Framework=A2Q, Model Architecture=GAT2026.01 | 92.5 | |
| No PromptingQuantization Framework=DQ, Model Architecture=GIN2026.01 | 92.4 | |
| GPF-LoRAPQuantization Framework=DQ, Model Architecture=GCN2026.01 | 92.4 | |
| GPF-plusQuantization Framework=A2Q, Model Architecture=GAT2026.01 | 92.4 | |
| GPF-plusQuantization Framework=DQ, Model Architecture=GIN2026.01 | 91.9 | |
| GPF-LoRAPQuantization Framework=QAT, Model Architecture=GIN2026.01 | 91.8 | |
| GPF-LoRAPQuantization Framework=A2Q, Model Architecture=GCN2026.01 | 91.8 | |
| GCNParameter Budget=504K2020.06 | 91.39 | |
| GPF-plusQuantization Framework=QAT, Model Architecture=GIN2026.01 | 91.3 | |
| No PromptingQuantization Framework=QAT, Model Architecture=GIN2026.01 | 91 | |
| GPF-plusQuantization Framework=A2Q, Model Architecture=GCN2026.01 | 91 | |
| No PromptingQuantization Framework=FP32, Model Architecture=GCN2026.01 | 90.9 | |
| GPF-plusQuantization Framework=DQ, Model Architecture=GCN2026.01 | 90.9 | |
| GCN2023.03 | 90.71 | |
| GCN#Param=≈100K2021.08 | 90.705 | |
| GCN2024.02 | 90.705 | |
| GCNParameter Budget=101K2020.06 | 90.7 | |
| GCNImplementation=Local Baseline2024.07 | 90.6 |