Singing Voice Deepfake Detection on CtrSVDD
6.23EERSingGraph
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| SingGraph2026.05 | 6.23 | — | — | — | |
| W2V2-MusicDETFront-end=W2V2 (Frozen), Back-end=MusicDET2026.05 | 6.74 | — | — | — | |
| W2V2-AASISTFront-end=W2V2 (Frozen), Back-end=AASIST2026.05 | 8.23 | — | — | — | |
| Alethia-LargeEvaluation Protocol=Zero-shot testing2026.04 | 10.8 | 91.3 | 92.5 | 84.1 | |
| AASISTBack-end=AASIST2026.05 | 11.54 | — | — | — | |
| W2V2-ResNetFront-end=W2V2 (Frozen), Back-end=ResNet2026.05 | 11.72 | — | — | — | |
| MERT-AASISTFront-end=MERT (Frozen), Back-end=AASIST2026.05 | 12.55 | — | — | — | |
| W2V2-LCNNFront-end=W2V2 (Frozen), Back-end=LCNN2026.05 | 12.78 | — | — | — | |
| W2V-1BEvaluation Protocol=Zero-shot testing2026.04 | 13.2 | 89.7 | 90.8 | 83.1 | |
| CtrSVDD Baseline*Evaluation Protocol=In-domain finetuning2026.04 | 13.8 | — | — | — | |
| Alethia-BaseEvaluation Protocol=Zero-shot testing2026.04 | 16.7 | 89.8 | 94 | 65.2 | |
| WavLM-AASISTFront-end=WavLM (Frozen), Back-end=AASIST2026.05 | 18.74 | — | — | — | |
| W2V-300MEvaluation Protocol=Zero-shot testing2026.04 | 22.4 | 84.9 | 88.6 | 63.7 | |
| Spec-ResNetFront-end=Spectrogram, Back-end=ResNet2026.05 | 22.59 | — | — | — | |
| WavLM-LargeEvaluation Protocol=Zero-shot testing2026.04 | 22.6 | 89.8 | 97.7 | 43.5 | |
| HuBERT-LargeEvaluation Protocol=Zero-shot testing2026.04 | 25.7 | 88.9 | 97.3 | 39.7 |