Audio Deepfake Detection on ASVspoof 2019
0.18EERW2V2-TCM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| W2V2-TCMParams=319M, Training Protocol=Protocol 1 (ASV19 train)2026.03 | 0.18 | — | |
| W2V-AASISTCategory=Finetuned frontend, #Param (million)=3172024.07 | 0.2 | 99.1 | |
| SLIMCategory=Finetuned frontend, #Param (million)=2532024.07 | 0.2 | 98.9 | |
| ConLLMPre-trained weights=GPT Weights2026.01 | 0.21 | — | |
| W2V2-AASISTParams=317M, Training Protocol=Protocol 1 (ASV19 train)2026.03 | 0.22 | — | |
| WLM-ASPCategory=Frozen frontend, #Param (million)=92024.07 | 0.3 | 98.3 | |
| SSL-FusionCategory=Frozen frontend, #Param (million)=102024.07 | 0.3 | 98.1 | |
| W2V-ASPCategory=Finetuned frontend, #Param (million)=3172024.07 | 0.3 | 98.4 | |
| CL+GL2026.01 | 0.33 | — | |
| MiO2026.01 | 0.41 | — | |
| mHuBERT-Iter1Params=100M, Training Protocol=Protocol 1 (ASV19 train)2026.03 | 0.49 | — | |
| HUB-ASPCategory=Frozen frontend, #Param (million)=92024.07 | 0.5 | 97.5 | |
| ConLLMPre-trained weights=None2026.01 | 0.5 | — | |
| mHuBERT-FinalParams=100M, Training Protocol=Protocol 1 (ASV19 train)2026.03 | 0.53 | — | |
| mHuBERT-Iter2Params=100M, Training Protocol=Protocol 1 (ASV19 train)2026.03 | 0.59 | — | |
| W2V-LCNNCategory=Frozen frontend, #Param (million)=N/A2024.07 | 0.6 | — | |
| FullCategory=SLIM variants (ours), #Param (million)=112024.07 | 0.6 | 96.9 | |
| WLM+HUBCategory=Frozen frontend, #Param (million)=92024.07 | 0.8 | 96.3 | |
| W2V+HUBCategory=Frozen frontend, #Param (million)=92024.07 | 0.9 | 95.6 | |
| WavLM-Base+Params=100M, Training Protocol=Protocol 1 (ASV19 train)2026.03 | 1.15 | — | |
| WavLM-BaseParams=100M, Training Protocol=Protocol 1 (ASV19 train)2026.03 | 1.74 | — | |
| W2V+WLMCategory=Frozen frontend, #Param (million)=92024.07 | 1.8 | 91.6 | |
| CQT-DCT-LCNN2026.01 | 1.84 | — | |
| W2V-LLGFCategory=Frozen frontend, #Param (million)=102024.07 | 2.3 | 93.6 | |
| DependencyCategory=SLIM variants (ours), #Param (million)=92024.07 | 2.8 | 89.7 | |
| RawNet2Category=Baseline, #Param (million)=42024.07 | 3 | 87.5 | |
| W2V-ASPCategory=Frozen frontend, #Param (million)=92024.07 | 3.3 | 85.8 | |
| Enc_style+lingCategory=SLIM variants (ours), #Param (million)=102024.07 | 3.5 | 83.4 | |
| LCNNCategory=Baseline, #Param (million)=42024.07 | 3.7 | 83.4 | |
| HuBERT-BaseParams=100M, Training Protocol=Protocol 1 (ASV19 train)2026.03 | 3.75 | — | |
| PS3DTCategory=Baseline, #Param (million)=N/A2024.07 | 4.5 | — | |
| Enc_lingCategory=SLIM variants (ours), #Param (million)=92024.07 | 5.9 | 76.4 | |
| Enc_styCategory=SLIM variants (ours), #Param (million)=92024.07 | 6.7 | 74 | |
| ProSDDTraining Setting=Trained on ASVspoof 20242026.04 | 19.04 | — | |
| AASISTTraining Setting=Trained on ASVspoof 20242026.04 | 23.16 | — | |
| RawNet2Training Setting=Trained on ASVspoof 20242026.04 | 24.75 | — | |
| XLSR-SLSTraining Setting=Trained on ASVspoof 20242026.04 | 27 | — |