Bitext Mining on Tatoeba (full)
85.4AccuracyVMSST CONTRASTIVE
Evaluation results
| Method | Links | |
|---|---|---|
| VMSST CONTRASTIVEInitialization=Pretrained, Number of Layers=242022.12 | 85.4 | |
| CONTRASTIVEInitialization=Pretrained, Number of Layers=242022.12 | 85.1 | |
| VMSSTInitialization=Pretrained, Number of Layers=242022.12 | 81.1 | |
| BITRANSLATIONInitialization=Pretrained, Number of Layers=242022.12 | 78.2 | |
| VMSST CONTRASTIVEInitialization=Random, Number of Layers=242022.12 | 64.8 | |
| VMSST CONTRASTIVEInitialization=Random, Number of Layers=62022.12 | 63.2 | |
| CONTRASTIVEInitialization=Random, Number of Layers=62022.12 | 63.1 | |
| CONTRASTIVEInitialization=Random, Number of Layers=242022.12 | 62.7 | |
| VMSSTInitialization=Random, Number of Layers=242022.12 | 61.4 | |
| VMSSTInitialization=Random, Number of Layers=62022.12 | 58.7 | |
| BITRANSLATIONInitialization=Random, Number of Layers=242022.12 | 57.4 | |
| BITRANSLATIONInitialization=Random, Number of Layers=62022.12 | 53.3 |