Link Prediction on Wiki (test)
39.52P@1GraphFormers
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GraphFormersTraining=Two-stage progressive learning, Symmetry=Unidirectional-simplified2021.05 | 39.52 | 62.3 | 52.2 | |
| PLM+MaxText Encoder=UniLM-base, Aggregator=Pooling-and-Concat (Max)2021.05 | 37.12 | 60.71 | 50.22 | |
| PLM+AttText Encoder=UniLM-base, Aggregator=Pooling-and-Concat (Attention)2021.05 | 37.09 | 60.67 | 50.18 | |
| PLM+MeanText Encoder=UniLM-base, Aggregator=Pooling-and-Concat (Mean)2021.05 | 36.64 | 60.37 | 49.8 | |
| PLMText Encoder=UniLM-base, Input=Center node textual feature2021.05 | 34.66 | 57.99 | 47.12 | |
| PLM+GATText Encoder=UniLM-base, Aggregator=GAT2021.05 | 30.06 | 54.3 | 42.7 | |
| IFTNType=Simplified text encoder baseline2021.05 | 18.38 | 42.76 | 29.45 | |
| TNVEType=Simplified text encoder baseline2021.05 | 17.86 | 42.74 | 29.33 |