Graph Property Prediction on BACE
98.4ROC AUC<SOGk>
Evaluation Results
| Method | Links | |
|---|---|---|
| <SOGk>Backbone=LLaMA2-7B2026.02 | 98.4 | |
| S-CGIBDomain Adaptation=false2024.12 | 86.51 | |
| S-CGIBDomain Adaptation=true2024.12 | 86.46 | |
| LEAPPre-training Strategy=ContextPred, Shot Scenario=full-shot2025.12 | 86.3 | |
| SUPTsoftPre-training Strategy=ContextPred, Shot Scenario=full-shot2025.12 | 86.19 | |
| LEAPPre-training Strategy=AttrMasking, Shot Scenario=full-shot2025.12 | 86.16 | |
| GPF-plusPre-training Strategy=ContextPred, Shot Scenario=full-shot2025.12 | 86.07 | |
| GPF-plusPre-training Strategy=AttrMasking, Shot Scenario=full-shot2025.12 | 86 | |
| SUPThardPre-training Strategy=ContextPred, Shot Scenario=full-shot2025.12 | 85.94 | |
| GPFPre-training Strategy=ContextPred, Shot Scenario=full-shot2025.12 | 85.82 | |
| RELIEFPre-training Strategy=ContextPred, Shot Scenario=full-shot2025.12 | 85.77 | |
| LR-GMPBackbone=GIN, Pre-training Method=EdgePred2026.04 | 85.7 | |
| GPFPre-training Strategy=AttrMasking, Shot Scenario=full-shot2025.12 | 85.36 | |
| SUPTsoftPre-training Strategy=AttrMasking, Shot Scenario=full-shot2025.12 | 85.35 | |
| RELIEFPre-training Strategy=AttrMasking, Shot Scenario=full-shot2025.12 | 85.2 | |
| SUPThardPre-training Strategy=AttrMasking, Shot Scenario=full-shot2025.12 | 85.18 | |
| FTPre-training Strategy=ContextPred, Shot Scenario=full-shot2025.12 | 85.1 | |
| LEAPPre-training Strategy=Infomax, Shot Scenario=full-shot2025.12 | 84.35 | |
| SUPThardPre-training Strategy=Infomax, Shot Scenario=full-shot2025.12 | 84.27 | |
| GPF-plusPre-training Strategy=Infomax, Shot Scenario=full-shot2025.12 | 84.2 | |
| SUPTsoftPre-training Strategy=Infomax, Shot Scenario=full-shot2025.12 | 84.2 | |
| GPFPre-training Strategy=Infomax, Shot Scenario=full-shot2025.12 | 84.01 | |
| RELIEFPre-training Strategy=Infomax, Shot Scenario=full-shot2025.12 | 83.97 | |
| VGPBackbone=GIN, Pre-training Method=EdgePred2026.04 | 83.62 | |
| FTPre-training Strategy=AttrMasking, Shot Scenario=full-shot2025.12 | 83.19 | |
| FTPre-training Strategy=Infomax, Shot Scenario=full-shot2025.12 | 82.19 | |
| MGSSL2024.12 | 82.03 | |
| GPF-plusBackbone=GIN, Pre-training Method=EdgePred2026.04 | 81.75 | |
| GPFBackbone=GIN, Pre-training Method=EdgePred2026.04 | 81.57 | |
| MoAMa2024.12 | 81.32 | |
| GROVE2024.12 | 81.13 | |
| FTBackbone=GIN, Pre-training Method=EdgePred2026.04 | 80.9 | |
| GraphFP2024.12 | 80.28 | |
| EdgePredPre-Train Dataset=ZINC 2M2021.06 | 79.9 | |
| SimSGT2024.12 | 79.75 | |
| ContextPredPre-Train Dataset=ZINC 2M2021.06 | 79.6 | |
| AttrMaskingPre-Train Dataset=ZINC 2M2021.06 | 79.3 | |
| ContextPredPre-Train dataset=ZINC 2M2025.11 | 78.8 | |
| AD-GCL-FIXPre-Train Dataset=ZINC 2M2021.06 | 78.51 | |
| GraphCLPre-Train Dataset=ZINC 2M2021.06 | 78.47 | |
| ContextPred2024.12 | 78.39 | |
| LEAPPre-training Strategy=GCL, Shot Scenario=full-shot2025.12 | 78.34 | |
| SUPThardPre-training Strategy=GCL, Shot Scenario=full-shot2025.12 | 78.19 | |
| GPFPre-training Strategy=GCL, Shot Scenario=full-shot2025.12 | 78.01 | |
| GPF-plusPre-training Strategy=GCL, Shot Scenario=full-shot2025.12 | 78.01 | |
| SUPTsoftPre-training Strategy=GCL, Shot Scenario=full-shot2025.12 | 78 | |
| Infomax2024.12 | 77.8 | |
| GraphCL2024.12 | 77.8 | |
| AttrMaskingPre-Train dataset=ZINC 2M2025.11 | 77.8 | |
| EdgePredPre-Train dataset=ZINC 2M2025.11 | 77.3 | |
| SpecMatch-CLPre-Train dataset=ZINC 2M2025.11 | 76.93 | |
| GraphLoG2024.12 | 76.6 | |
| RELIEFPre-training Strategy=GCL, Shot Scenario=full-shot2025.12 | 76.03 | |
| AttrMasking2024.12 | 75.95 | |
| GTTok.=Lap, Mode=Pad.2026.05 | 75.94 | |
| InfoGraphPre-Train Dataset=ZINC 2M2021.06 | 75.9 | |
| FTPre-training Strategy=GCL, Shot Scenario=full-shot2025.12 | 75.7 | |
| GTTok.=Adj, Mode=Trunc.2026.05 | 75.26 | |
| DeepSet2026.05 | 75.22 | |
| SimGRACEPre-Train dataset=ZINC 2M2025.11 | 75 | |
| JOAO2024.12 | 74.94 | |
| GTTok.=Lap, Mode=Trunc.2026.05 | 74.91 | |
| GTTok.=Adj, Mode=Pad.2026.05 | 74.88 | |
| GTTok.=Comb., Mode=Trunc.2026.05 | 74.78 | |
| GraphCLPre-Train dataset=ZINC 2M2025.11 | 74.6 | |
| JOAOv22024.12 | 74.38 | |
| EdgePred2024.12 | 74.29 | |
| JOAOPre-Train dataset=ZINC 2M2025.11 | 73.2 | |
| GIN2026.05 | 72.73 | |
| GTTok.=Comb., Mode=Pad.2026.05 | 72.49 | |
| No Pre-TrainPre-Train dataset=None2025.11 | 72.4 | |
| GTTok.=RW, Mode=Trunc.2026.05 | 70.87 | |
| GPPTBackbone=GIN, Pre-training Method=EdgePred2026.04 | 70.85 | |
| GTTok.=RW, Mode=Pad.2026.05 | 70.61 | |
| LEAPPre-training Strategy=ContextPred, Shot Scenario=50-shot2025.12 | 70.33 | |
| GPPT w/o olBackbone=GIN, Pre-training Method=EdgePred2026.04 | 70.31 | |
| No Pre-TrainPre-Train Dataset=None2021.06 | 70.1 | |
| RELIEFPre-training Strategy=ContextPred, Shot Scenario=50-shot2025.12 | 70.07 | |
| SUPThardPre-training Strategy=ContextPred, Shot Scenario=50-shot2025.12 | 69.64 | |
| GIMLETBackbone=Non-LLM2026.02 | 69.6 | |
| LEAPPre-training Strategy=AttrMasking, Shot Scenario=50-shot2025.12 | 69.45 | |
| RELIEFPre-training Strategy=AttrMasking, Shot Scenario=50-shot2025.12 | 69.37 | |
| GPF-plusPre-training Strategy=ContextPred, Shot Scenario=50-shot2025.12 | 69.18 | |
| SUPThardPre-training Strategy=AttrMasking, Shot Scenario=50-shot2025.12 | 68.98 | |
| GPF-plusPre-training Strategy=AttrMasking, Shot Scenario=50-shot2025.12 | 68.9 | |
| SUPTsoftPre-training Strategy=ContextPred, Shot Scenario=50-shot2025.12 | 68.75 | |
| GPFPre-training Strategy=ContextPred, Shot Scenario=50-shot2025.12 | 68.42 | |
| GPFPre-training Strategy=AttrMasking, Shot Scenario=50-shot2025.12 | 68.26 | |
| LLaMA3-3BProtocol=Few-shot (Random)2026.02 | 68.2 | |
| GraphPromptBackbone=GIN, Pre-training Method=EdgePred2026.04 | 67.7 | |
| RELIEFPre-training Strategy=Infomax, Shot Scenario=50-shot2025.12 | 67.56 | |
| LEAPPre-training Strategy=Infomax, Shot Scenario=50-shot2025.12 | 67.31 | |
| SUPTsoftPre-training Strategy=AttrMasking, Shot Scenario=50-shot2025.12 | 67.05 | |
| FTPre-training Strategy=AttrMasking, Shot Scenario=50-shot2025.12 | 66.86 | |
| SUPThardPre-training Strategy=Infomax, Shot Scenario=50-shot2025.12 | 66.8 | |
| FTPre-training Strategy=ContextPred, Shot Scenario=50-shot2025.12 | 66.74 | |
| SUPTsoftPre-training Strategy=Infomax, Shot Scenario=50-shot2025.12 | 66.44 | |
| GPF-plusPre-training Strategy=Infomax, Shot Scenario=50-shot2025.12 | 66.23 | |
| FTPre-training Strategy=Infomax, Shot Scenario=50-shot2025.12 | 65.65 | |
| GPFPre-training Strategy=Infomax, Shot Scenario=50-shot2025.12 | 65.59 |