Graph property prediction on PCQM4M-LSC (val)
0.1224MAEEGT-Medium (18 layers)
Evaluation Results
| Method | Links | |
|---|---|---|
| EGT-Medium (18 layers)#Param=47.4M, Layers=182021.08 | 0.1224 | |
| Graphormer2025.07 | 0.1234 | |
| Graphormer#param.=47.1M2021.06 | 0.1234 | |
| Graphormer#Param=47.1M2021.08 | 0.1234 | |
| Graphormer + URPE-based Attention#Params=12.5M2022.05 | 0.1238 | |
| EGT-Small (6 layers)#Param=11.5M, Layers=62021.08 | 0.126 | |
| Ours (S, φ)Shorter schedule=Full iterations2021.10 | 0.1263 | |
| GraphormerSMALLmodel_size=Small2025.07 | 0.1264 | |
| Graphormer SMALL#param.=12.5M2021.06 | 0.1264 | |
| Graphormer-SMALL#Param=12.5M2021.08 | 0.1264 | |
| Graphormer#Params=12.5M2022.05 | 0.1264 | |
| Ours (S, φ)Shorter schedule=10% of full iterations2021.10 | 0.1294 | |
| Ours (S, φ)-SMALLModel size=SMALL, Shorter schedule=10% of full iterations2021.10 | 0.1376 | |
| GIN-VN#param.=6.7M2021.06 | 0.1395 | |
| GIN-VN#Params=6.7M2022.05 | 0.1395 | |
| GIN-VN2021.10 | 0.1396 | |
| GIN-VN#Param=6.7M2021.08 | 0.1396 | |
| OmniVec22025.07 | 0.1397 | |
| DeeperGCN-VN#param.=25.5M2021.06 | 0.1398 | |
| DeeperGCN-VN#Param=25.5M2021.08 | 0.1398 | |
| DeeperGCN-VN#Params=25.5M2022.05 | 0.1398 | |
| Graph Transformer2025.07 | 0.14 | |
| GT#param.=0.6M2021.06 | 0.14 | |
| GT#Param=0.6M2021.08 | 0.14 | |
| GT#Params=0.6M2022.05 | 0.14 | |
| Graph Transformer-Widemodel_variant=Wide2025.07 | 0.1408 | |
| GT-Wide#param.=83.2M2021.06 | 0.1408 | |
| GT (bigger model)#Param=83.2M2021.08 | 0.1408 | |
| GT-Wide#Params=83.2M2022.05 | 0.1408 | |
| GINE-VN#param.=13.2M2021.06 | 0.143 | |
| GINE-VN#Param=13.2M2021.08 | 0.143 | |
| GINE-VN#Params=13.2M2022.05 | 0.143 | |
| MLP (S)Shorter schedule=10% of full iterations2021.10 | 0.1464 | |
| GCN-VN#param.=4.9M2021.06 | 0.1485 | |
| GCN-VN#Params=4.9M2022.05 | 0.1485 | |
| GCN-VN2021.10 | 0.151 | |
| GCN-VN#Param=4.9M2021.08 | 0.151 | |
| GIN2021.10 | 0.1536 | |
| GIN#Param=3.8M2021.08 | 0.1536 | |
| GIN#param.=3.8M2021.06 | 0.1537 | |
| GIN#Params=3.8M2022.05 | 0.1537 | |
| GCN2021.10 | 0.1684 | |
| GCN#Param=2.0M2021.08 | 0.1684 | |
| GCN#param.=2.0M2021.06 | 0.1691 | |
| GCN#Params=2.0M2022.05 | 0.1691 | |
| MLP-FINGERPRINT2021.10 | 0.2044 | |
| Transformer + Laplacian PEShorter schedule=10% of full iterations2021.10 | 0.2162 | |
| Meta-Transformer-B16Fbackbone=B16F2025.07 | 0.8863 |