Speech Deepfake Detection on Commercial APIs Challenge Set latest (test)
0.9601Average ScoreXLS-R-1B
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| XLS-R-1BTraining Strategy=DOSS-Weight, Backbone=XLS-R-1B2025.12 | 0.9601 | 0.9998 | 0.9992 | 0.9892 | 0.863 | 0.9978 | 0.9998 | 0.9996 | 0.9194 | 0.8732 | |
| XLS-R-300MTraining Strategy=DOSS-Weight, Backbone=XLS-R-300M2025.12 | 0.9281 | 0.9996 | 0.9984 | 0.9804 | 0.9212 | 0.9892 | 1 | 0.9998 | 0.905 | 0.5594 | |
| XLS-R-2BTraining Strategy=Naive Aggregation, Backbone=XLS-R-2B, Reference=Ge et al., 20252025.12 | 0.8631 | 0.9954 | 0.8542 | 0.992 | 0.819 | 0.9606 | 0.9896 | 0.9934 | 0.7648 | 0.3988 |