Bitext Mining on BUCC (full)
87.8F1 (Cosine Similarity)VMSST
Evaluation results
| Method | Links | ||
|---|---|---|---|
| VMSSTInitialization=Pretrained, Number of Layers=242022.12 | 87.8 | 92.5 | |
| BITRANSLATIONInitialization=Pretrained, Number of Layers=242022.12 | 85.9 | 91.9 | |
| VMSSTInitialization=Random, Number of Layers=242022.12 | 78.7 | 89 | |
| VMSST CONTRASTIVEInitialization=Pretrained, Number of Layers=242022.12 | 74.6 | 93.7 | |
| CONTRASTIVEInitialization=Pretrained, Number of Layers=242022.12 | 74.3 | 93.7 | |
| VMSSTInitialization=Random, Number of Layers=62022.12 | 73.7 | 85.9 | |
| BITRANSLATIONInitialization=Random, Number of Layers=242022.12 | 70.8 | 86.9 | |
| CONTRASTIVEInitialization=Random, Number of Layers=62022.12 | 66.2 | 84 | |
| VMSST CONTRASTIVEInitialization=Random, Number of Layers=62022.12 | 65.8 | 84.3 | |
| CONTRASTIVEInitialization=Random, Number of Layers=242022.12 | 64.3 | 83.7 | |
| BITRANSLATIONInitialization=Random, Number of Layers=62022.12 | 62.1 | 81.2 | |
| VMSST CONTRASTIVEInitialization=Random, Number of Layers=242022.12 | 58.5 | 84.1 |