Speech Deepfake Detection on FakeOrReal
3.04EERXLSR-AASIST
Evaluation results
| Method | Links | ||
|---|---|---|---|
| XLSR-AASISTTraining Strategy=DPDA training + PCGrad2025.09 | 3.04 | — | |
| XLSR-Conformer-TCMLoss Function=QAMO2025.09 | 3.45 | — | |
| XLSR-Nes2NetXLoss Function=OC-Softmax2025.09 | 3.97 | — | |
| XLSR-Conformer-TCMTraining Strategy=DPDA training + PCGrad2025.09 | 4.47 | — | |
| XLSR-AASISTTraining Strategy=DPDA training2025.09 | 4.6 | — | |
| XLSR-MambaTraining Strategy=DPDA training + PCGrad2025.09 | 4.86 | — | |
| XLSR-Nes2NetXLoss Function=QAMO2025.09 | 4.9 | — | |
| XLSR-SLS2025.09 | 5.07 | — | |
| XLSR-AASIST-SAM2025.09 | 5.18 | — | |
| XLSR-Conformer-TCMTraining Strategy=DPDA training2025.09 | 5.31 | — | |
| XLSR-MambaTraining Strategy=DPDA training2025.09 | 5.39 | — | |
| XLSR-Conformer-TCMLoss Function=OC-Softmax2025.09 | 5.65 | — | |
| XLSR-Conformer-TCMLoss Function=WCE, experimental_setting=Reproduced2025.09 | 5.7 | — | |
| XLSR-Nes2NetXLoss Function=WCE, experimental_setting=Original2025.09 | 6.31 | — | |
| XLSR-Nes2Net-X2025.09 | 6.31 | — | |
| XLSR-MambaLoss Function=WCE2025.09 | 6.71 | — | |
| XLSR-MambaTraining Strategy=Baseline2025.09 | 6.71 | — | |
| XLSR-AASISTTraining Strategy=Baseline2025.09 | 7.46 | — | |
| XLSR-Nes2NetXLoss Function=WCE, experimental_setting=Reproduced2025.09 | 10.12 | — | |
| XLSR-Conformer-TCMLoss Function=WCE, experimental_setting=Original2025.09 | 10.68 | — | |
| XLSR-Conformer-TCMTraining Strategy=Baseline2025.09 | 10.68 | — | |
| XLS-R-1BTraining Strategy=DOSS-Weight, #Params=965M, #Hours=12k2025.12 | 13 | — | |
| SFATNet-4Data Condition=Clean (no codecs)2026.01 | 20.33 | 85.03 | |
| SFAT-Net 3Data Condition=Clean (no codecs)2026.01 | 21.08 | 81.01 | |
| XLS-R-300MTraining Strategy=DOSS-Weight, #Params=317M, #Hours=12k2025.12 | 25 | — | |
| SOTATraining Strategy=Existing Benchmarks2025.12 | 97 | — | |
| MMS-1BTraining Strategy=Naive Aggregation, #Params=965M, #Hours=74k2025.12 | 173 | — | |
| XLS-R-2BTraining Strategy=Naive Aggregation, #Params=2.2B, #Hours=74k2025.12 | 173 | — | |
| MMS-300MTraining Strategy=Naive Aggregation, #Params=317M, #Hours=74k2025.12 | 608 | — | |
| XLS-R-1BTraining Strategy=Naive Aggregation, #Params=965M, #Hours=74k2025.12 | 1,215 | — |