Semantic Textual Similarity on Multilingual STS low-resource languages
81.4STS Score (Afr)Synth - Prompting
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Synth - PromptingBackbone=XLM-R2026.03 | 81.4 | 77.9 | 82.3 | 81.4 | 38.3 | 47.4 | 69.9 | 68.4 | |
| Synth - XL-LoRABackbone=XLM-R2026.03 | 81 | 78 | 84.3 | 84 | 47.8 | 58.4 | 72.8 | 72.3 | |
| Synth - PromptingBackbone=MM-BERT2026.03 | 80.6 | 76.8 | 76.3 | 76.6 | 41.3 | 45.2 | 68 | 66.4 | |
| Synth - Adapter CompositionBackbone=XLM-R2026.03 | 80.4 | 77.6 | 84.8 | 83 | 46.4 | 49.5 | 74.7 | 70.9 | |
| Synth - XL-LoRABackbone=MM-BERT2026.03 | 80.4 | 79.5 | 84.6 | 83.5 | 49.1 | 56.1 | 72.6 | 72.3 | |
| Synth - Adapter CompositionBackbone=MM-BERT2026.03 | 79.9 | 77.6 | 82.2 | 81.1 | 46.2 | 42.1 | 72.6 | 68.8 | |
| Cross LingualBackbone=XLM-R2026.03 | 78 | 77.4 | 81.6 | 80.9 | 47.1 | 48.2 | 79.9 | 70.4 | |
| Cross LingualBackbone=MM-BERT2026.03 | 78 | 77.7 | 79.8 | 73.6 | 49.3 | 29.4 | 79.6 | 66.8 | |
| UnsupervisedBackbone=MM-BERT2026.03 | 75.5 | 58 | 68.6 | 62.1 | 44.7 | 25.5 | 58 | 56.1 | |
| UnsupervisedBackbone=XLM-R2026.03 | 74.8 | 69.7 | 77.1 | 76.2 | 39.2 | 45.4 | 70.9 | 64.7 | |
| Base EncoderBackbone=MM-BERT2026.03 | 72 | 63.8 | 70.7 | 66.8 | 52.2 | 21.7 | 62 | 58.5 | |
| Base EncoderBackbone=XLM-R2026.03 | 56.2 | 52.7 | 55.7 | 46.3 | 46.7 | 4.1 | 60.8 | 46.1 |