Audio Deepfake Detection on ASVspoof 5
4.41EERSFATNet-4
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SFATNet-4Data Condition=Clean (no codecs)2026.01 | 4.41 | 98.89 | |
| SFAT-Net 3Data Condition=Clean (no codecs)2026.01 | 8.85 | 96.69 | |
| Probing-Guided Layer SelectionBackbone=XLS-R, Backbone status=Frozen, Layers used=4, Trainable params=1.34M, Fusion=Concat2026.06 | 11.03 | — | |
| Wav2Vec2+AASIST(Baseline)Params (M)=317.842026.02 | 13.38 | — | |
| XLSR+BiCrossMambaParams (M)=318.212026.02 | 13.67 | — | |
| MLDG-LoRABackbone=W2V 2.0, Backbone status=LoRA, Layers used=All, Trainable params=3.59M, Fusion=Feature2026.06 | 14.1 | — | |
| XLSR+MambaParams (M)=3192026.02 | 14.4 | — | |
| HyperPotterParams (M)=317.872026.02 | 16.04 | — | |
| Wav2Vec2+AASISTParams (M)=317.842026.02 | 16.24 | — | |
| XLSR+SLSParams (M)=3402026.02 | 18.76 | — | |
| XLSR+Conformer+TCMParams (M)=3192026.02 | 18.85 | — |