Audio Deepfake Detection on ITW
98.82ACCXLS-R-300M
Evaluation results
| Method | Links | ||
|---|---|---|---|
| XLS-R-300M#Params=317M, #Hours=12k, Training Protocol=DOSS-Weight2025.12 | 98.82 | — | |
| XLS-R-2B#Params=2.2B, #Hours=74k, Training Protocol=Naive Aggregation2025.12 | 98.7 | — | |
| XLS-R-1B#Params=965M, #Hours=74k, Training Protocol=Naive Aggregation2025.12 | 98.62 | — | |
| XLS-R-1B#Params=965M, #Hours=12k, Training Protocol=DOSS-Weight2025.12 | 98.58 | — | |
| MMS-300M#Params=317M, #Hours=74k, Training Protocol=Naive Aggregation2025.12 | 97.31 | — | |
| MMS-1B#Params=965M, #Hours=74k, Training Protocol=Naive Aggregation2025.12 | 96.66 | — | |
| LCNN-LSTM2026.01 | 66 | 62 | |
| SALMONNmode=multi-prompt2026.01 | 66 | 62 | |
| LCNN2026.01 | 65 | 63 | |
| Mesonet2026.01 | 53 | 53 | |
| ResNet182026.01 | 49 | 46 | |
| RawPC2026.01 | 45 | 43 | |
| CRNNSpoof2026.01 | 41 | 39 | |
| RawGAT-ST2026.01 | 37 | 38 | |
| RawNet22026.01 | 33 | 33 |