Semantic Textual Similarity on Multilingual STS-B (val)
77.48Spearman CorrelationVisualCSE
Evaluation results
| Method | Links | |
|---|---|---|
| VisualCSELanguage=Russian, Backbone=BERT-base2022.09 | 77.48 | |
| VisualCSELanguage=French, Backbone=BERT-base2022.09 | 72.52 | |
| SimCSELanguage=Russian, Backbone=BERT-base2022.09 | 72.5 | |
| SimCSELanguage=French, Backbone=BERT-base2022.09 | 70.31 | |
| VisualCSELanguage=Chinese, Backbone=BERT-base2022.09 | 70.05 | |
| VisualCSELanguage=German, Backbone=BERT-base2022.09 | 69.87 | |
| SimCSELanguage=Chinese, Backbone=BERT-base2022.09 | 67.98 | |
| SimCSELanguage=German, Backbone=BERT-base2022.09 | 67.34 |