List your company
ResearchFormerly WizWand

Audio Deepfake Detection on ASVspoof LA 2021 (EER)

0.79EERMamBo-1-Mamba2-N2
-0.54728.478917.50526.5311Sep 25, 2025Oct 24, 2025Nov 23, 2025Dec 22, 2025Jan 21, 2026Feb 19, 2026Mar 21, 2026
Updated 4mo ago

Evaluation results

MethodLinks
MamBo-1-Mamba2-N2Params (M)=317.482026.010.79
MamBo-2-Hydra-N1Params (M)=316.682026.010.8
MamBo-3-Hydra-N3Params (M)=319.372026.010.81
Wav2Vec+AASISTBackbone=Wav2Vec, Head=AASIST2026.020.82
XLSR-Conformer-NACLLoss Function=WCE2025.090.89
OCKD2026.020.9
XLSR-MambaParams (M)=319.332026.010.93
XLSR-MambaLoss Function=WCE2025.090.93
Fake-Mamba (L)Params (M)=319.722026.010.97
XLSR-ConformerLoss Function=WCE2025.090.97
MamBo-4-Hydra-N1Params (M)=318.022026.010.98
XLSR-Conformer+TCMParams (M)=319.772026.011.03
XLSR-Conformer-TCMLoss Function=WCE, experimental_setting=Original2025.091.03
OC+ACS2026.021.3
ASC+OCLoss Function=WCE2025.091.3
XLSR-Conformer-TCMLoss Function=WCE, experimental_setting=Reproduced2025.091.37
XLSR-ConformerParams (M)=319.742026.011.38
Wav2Vec+AASIST2Backbone=Wav2Vec, Head=AASIST22026.021.61
ResembleAI-2BTraining Protocol=Protocol 2 (multi-dataset), Model Scale=2000M, System Category=SOTA systems2026.031.64
XLSR-Conformer-TCMLoss Function=OC-Softmax2025.091.75
XLSR-Nes2NetXLoss Function=WCE, experimental_setting=Original2025.092
XLSR-Nes2NetXLoss Function=QAMO2025.092.29
XLS-R + HierConBackbone=XLS-R 300M, Head=HierCon2026.022.46
XLSR-Conformer-TCMLoss Function=QAMO2025.092.53
RawMambaParams (M)=0.7192026.012.84
XLSR-SLSParams (M)=341.492026.012.87
FTDKD2026.022.96
RawBMambaParams (M)=0.7192026.013.28
WavLM+ASPBackbone=WavLM, Head=ASP2026.023.31
XLSR-Nes2NetXLoss Function=OC-Softmax2025.093.36
BiCrossMamba-STParams (M)=0.5162026.013.39
WavLM+AttMBackbone=WavLM, Head=AttM2026.023.5
Wav2Vec+LinearBackbone=Wav2Vec, Head=Linear2026.023.63
XLS-R + SLSBackbone=XLS-R, Head=SLS2026.023.88
XLSR-Nes2NetXLoss Function=WCE, experimental_setting=Reproduced2025.093.9
DF-Arena 500MTraining Protocol=Protocol 2 (multi-dataset), Model Scale=500M, System Category=SOTA systems2026.034.23
WavLM+MFABackbone=WavLM, Head=MFA2026.025.08
W2V2-AASISTTraining Protocol=Protocol 2 (multi-dataset), Model Scale=317M, System Category=SOTA systems2026.036.01
MoLEXTraining Protocol=Protocol 2 (multi-dataset), Model Scale=376M, System Category=SOTA systems2026.036.32
WavLM ECAPA2026.036.67
W2V2-TCMTraining Protocol=Protocol 2 (multi-dataset), Model Scale=319M, System Category=SOTA systems2026.036.95
mHuBERT-Iter2Training Protocol=Protocol 2 (multi-dataset), Model Scale=100M, System Category=RAPTOR (ours)2026.037.02
DF-Arena 100M-V1Training Protocol=Protocol 2 (multi-dataset), Model Scale=100M, System Category=SOTA systems2026.037.61
WavLM-Base+Training Protocol=Protocol 2 (multi-dataset), Model Scale=100M, System Category=RAPTOR (ours)2026.038.66
mHuBERT-FinalTraining Protocol=Protocol 2 (multi-dataset), Model Scale=100M, System Category=RAPTOR (ours)2026.039.16
SNAP2026.039.46
mHuBERT-Iter1Training Protocol=Protocol 2 (multi-dataset), Model Scale=100M, System Category=RAPTOR (ours)2026.0311.16
HuBERT-BaseTraining Protocol=Protocol 2 (multi-dataset), Model Scale=100M, System Category=RAPTOR (ours)2026.0311.96
WavLM-BaseTraining Protocol=Protocol 2 (multi-dataset), Model Scale=100M, System Category=RAPTOR (ours)2026.0313.07
WavLM2026.0320.02
Wav2Vec 2.0 ECAPA2026.0326.6
RawNet22026.0329.71
AASIST2026.0334.22