Cross-modal and cross-lingual retrieval on FLEURS (test)
15.7Avg xSIM++ (26)MMS + SONAR-200
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MMS + SONAR-2002025.05 | 15.7 | 17.2 | |
| SONAR Speech2025.05 | 14.3 | — | |
| MMS + SONAR-752025.05 | 12.7 | 13.8 | |
| Speech-charSONAR-75 - PRETRConfiguration=PRETR2025.05 | 11.6 | 12.9 | |
| MMS + charSONAR-75Tokenization=chars2025.05 | 10.7 | 11.5 | |
| Speech-charSONAR-75 - PRETRConfiguration=PRETR, Fine-tuning=FLEURS train2025.05 | 10 | 10.7 | |
| Speech-charSONAR-75 - PRETRConfiguration=PRETR, Objective=DUAL2025.05 | 9.4 | 10.2 |