Audio Deepfake Detection on ASVspoof DF 2021 (evaluation)
7.88EERDual-Granularity Orthogonal Disentanglement (Full Model)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Dual-Granularity Orthogonal Disentanglement (Full Model)Model Scale=≤5M parameters2026.06 | 7.88 | 0.2689 | |
| WavLM-MLPModel Scale=>300M parameters, Pre-training=SSL-based2026.06 | 7.95 | 0.2708 | |
| Cosine OnlyModel Scale=≤5M parameters2026.06 | 8.23 | 0.2762 | |
| DG-AggModel Scale=≤5M parameters2026.06 | 8.26 | 0.2814 | |
| Wav2Vec2-AASISTModel Scale=>300M parameters, Pre-training=SSL-based2026.06 | 8.54 | 0.2876 | |
| GRL BaselineModel Scale=≤5M parameters2026.06 | 8.91 | 0.2953 | |
| Res-TSSDNetModel Scale=≤5M parameters2026.06 | 9.81 | 0.3105 | |
| RawGAT-STModel Scale=≤5M parameters2026.06 | 10.75 | 0.3218 | |
| AASISTModel Scale=≤5M parameters2026.06 | 12.83 | 0.3624 | |
| RawNet2Model Scale=≤5M parameters2026.06 | 15.14 | 0.4198 | |
| LCNN BaselineModel Scale=≤5M parameters2026.06 | 15.3 | 0.4482 | |
| LFCC-LCNNModel Scale=≤5M parameters2026.06 | 15.62 | 0.4567 | |
| LFCC-GMMModel Scale=≤5M parameters2026.06 | 22.38 | 0.5765 |