Graph Classification on MNIST (test)
98.8AccuracyESA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ESA2024.02 | 98.8 | — | — | |
| NeuralWalker2024.06 | 98.692 | — | — | |
| EXPHORMERparameter_budget=~100K, runs=mean of 42024.01 | 98.55 | — | — | |
| Exphormer2023.08 | 98.55 | — | — | |
| Exphormer2024.06 | 98.55 | — | — | |
| GRITencoding=HDSE2023.08 | 98.424 | — | — | |
| CKGCNParameter limit=~ 100K2024.04 | 98.423 | — | — | |
| GREDparameter budget=≈ 500K2023.12 | 98.383 | — | — | |
| GraphGPSencoding=HDSE2023.08 | 98.367 | — | — | |
| EIGENFORMER (feat.)parameter_budget=~100K, runs=single, attention_mechanism=node and edge feature2024.01 | 98.362 | — | — | |
| Graph MLP-Mixerparameter budget=≈ 500K2023.12 | 98.32 | — | — | |
| SINC-GCN2025.11 | 98.28 | — | — | |
| GPS-GCCMBackbone=GPS2026.05 | 98.236 | — | — | |
| GPS2024.02 | 98.2 | — | — | |
| GPS+RWSEBase Model Architecture=GPS, Positional and Structural Encoding (PSE)=RWSE2025.04 | 98.19 | — | — | |
| EGT2022.05 | 98.173 | — | — | |
| EGT# Parameters=~ 100K2023.05 | 98.173 | — | — | |
| EGTparameter_budget=~100K, runs=mean of 42024.01 | 98.173 | — | — | |
| Graphormer-GD2023.08 | 98.173 | — | — | |
| EGTparameter budget=≈ 500K2023.12 | 98.173 | — | — | |
| EGTParameter limit=~ 100K2024.04 | 98.173 | — | — | |
| EGT2026.05 | 98.173 | — | — | |
| GPS+LapPEBase Model Architecture=GPS, Positional and Structural Encoding (PSE)=LapPE2025.04 | 98.16 | — | — | |
| GPS+GFSEBase Model Architecture=GPS, Positional and Structural Encoding (PSE)=GFSE2025.04 | 98.15 | — | — | |
| GPS-PCLBackbone=GPS2026.05 | 98.114 | — | — | |
| GRIT# Parameters=~ 100K2023.05 | 98.108 | — | — | |
| GRITparameter_budget=~100K, runs=mean of 42024.01 | 98.108 | — | — | |
| GRIT2023.08 | 98.108 | — | — | |
| GRITparameter budget=≈ 500K2023.12 | 98.108 | — | — | |
| GRITParameter limit=~ 100K2024.04 | 98.108 | — | — | |
| GRIT2024.06 | 98.108 | — | — | |
| GRIT2026.05 | 98.108 | — | — | |
| GATv22024.02 | 98.1 | — | — | |
| GPS2026.05 | 98.082 | — | — | |
| GPS+GPSEBase Model Architecture=GPS, Positional and Structural Encoding (PSE)=GPSE2025.04 | 98.08 | — | — | |
| GPS2022.05 | 98.051 | — | — | |
| GPS# Parameters=~ 100K2023.05 | 98.051 | — | — | |
| GPSparameter_budget=~100K, runs=mean of 42024.01 | 98.051 | — | — | |
| GraphGPS2023.08 | 98.051 | — | — | |
| GPSparameter budget=≈ 500K2023.12 | 98.051 | — | — | |
| GPSParameter limit=~ 100K2024.04 | 98.051 | — | — | |
| GPS2024.06 | 98.051 | — | — | |
| GPSBase Model Architecture=GPS, Positional and Structural Encoding (PSE)=None2025.04 | 98.05 | — | — | |
| GraphGPS2025.11 | 98.05 | — | — | |
| TF+GFSEBase Model Architecture=Transformer, Positional and Structural Encoding (PSE)=GFSE2025.04 | 98.03 | — | — | |
| GNASDepth=4, #Param=0.39M, Search=6.00 hr, Train=3.10 hr2021.03 | 98.01 | 0.1 | — | |
| PNA2024.02 | 98 | — | — | |
| CRaWl2022.05 | 97.944 | — | — | |
| CRAWLparameter budget=100k2021.02 | 97.944 | — | — | |
| CRaWl# Parameters=~ 100K2023.05 | 97.944 | — | — | |
| CRaW1parameter_budget=~100K, runs=mean of 42024.01 | 97.944 | — | — | |
| CRaW1Parameter limit=~ 100K2024.04 | 97.944 | — | — | |
| CRaWL2024.06 | 97.944 | — | — | |
| PNA2022.05 | 97.94 | — | — | |
| PNA2021.02 | 97.94 | — | — | |
| DGN# Parameters=~ 100K2023.05 | 97.94 | — | — | |
| PNAparameter_budget=~100K, runs=mean of 42024.01 | 97.94 | — | — | |
| PNA2023.08 | 97.94 | — | — | |
| PNAParameter limit=~ 100K2024.04 | 97.94 | — | — | |
| PNA2026.05 | 97.94 | — | — | |
| SIR-GCN2025.11 | 97.9 | — | — | |
| TF+RWSEBase Model Architecture=Transformer, Positional and Structural Encoding (PSE)=RWSE2025.04 | 97.81 | — | — | |
| TF+GPSEBase Model Architecture=Transformer, Positional and Structural Encoding (PSE)=GPSE2025.04 | 97.78 | — | — | |
| GAT2024.02 | 97.5 | — | — | |
| GateGCN+GFSEBase Model Architecture=GateGCN, Positional and Structural Encoding (PSE)=GFSE2025.04 | 97.44 | — | — | |
| Graph ViT/MLP-Mixer2023.08 | 97.422 | — | — | |
| GatedGCNDepth=4, #Param=0.10M, Search=-, Train=3.50 hr2021.03 | 97.34 | 0.14 | — | |
| GatedGCN2022.05 | 97.34 | — | — | |
| GatedGCN2021.02 | 97.34 | — | — | |
| GatedGCN# Parameters=~ 100K2023.05 | 97.34 | — | — | |
| GatedGCNparameter_budget=~100K, runs=mean of 42024.01 | 97.34 | — | — | |
| GatedGCN2023.08 | 97.34 | — | — | |
| GatedGCNparameter budget=≈ 500K2023.12 | 97.34 | — | — | |
| GatedGCNParameter limit=~ 100K2024.04 | 97.34 | — | — | |
| GatedGCN2024.06 | 97.34 | — | — | |
| GateGCNBase Model Architecture=GateGCN, Positional and Structural Encoding (PSE)=None2025.04 | 97.34 | — | — | |
| GatedGCN2026.05 | 97.34 | — | — | |
| GraphSAGE2021.02 | 97.312 | — | — | |
| GraphSageDepth=4, #Param=0.10M, Search=-, Train=3.13 hr2021.03 | 97.31 | 0.1 | — | |
| GraphSAGE2025.11 | 97.31 | — | — | |
| DropGIN2024.02 | 97.3 | — | — | |
| Transformer (TF)Base Model Architecture=Transformer, Positional and Structural Encoding (PSE)=None2025.04 | 97.29 | — | — | |
| PNA2025.11 | 97.19 | — | — | |
| EIGENFORMERparameter_budget=~100K, runs=single2024.01 | 97.142 | — | — | |
| GateGCN+LapPEBase Model Architecture=GateGCN, Positional and Structural Encoding (PSE)=LapPE2025.04 | 97.1 | — | — | |
| TF+LapPEBase Model Architecture=Transformer, Positional and Structural Encoding (PSE)=LapPE2025.04 | 96.95 | — | — | |
| GateGCN+GPSEBase Model Architecture=GateGCN, Positional and Structural Encoding (PSE)=GPSE2025.04 | 96.94 | — | — | |
| GIN2024.02 | 96.9 | — | — | |
| GateGCN+RWSEBase Model Architecture=GateGCN, Positional and Structural Encoding (PSE)=RWSE2025.04 | 96.84 | — | — | |
| GINDepth=4, #Param=0.10M, Search=-, Train=1.41 hr2021.03 | 96.49 | 0.25 | — | |
| GIN2025.11 | 96.49 | — | — | |
| GIN2022.05 | 96.485 | — | — | |
| GIN2021.02 | 96.485 | — | — | |
| GIN# Parameters=~ 100K2023.05 | 96.485 | — | — | |
| GINparameter_budget=~100K, runs=mean of 42024.01 | 96.485 | — | — | |
| GIN2023.08 | 96.485 | — | — | |
| GINparameter budget=≈ 500K2023.12 | 96.485 | — | — | |
| GINParameter limit=~ 100K2024.04 | 96.485 | — | — | |
| GIN2024.06 | 96.485 | — | — | |
| GIN2026.05 | 96.485 | — | — |