ResearchBenchmarksArtifact Detection on Per-sample diagnostics Cross-class noiseArtifact Detection on Per-sample diagnostics Cross-class noiseFollow0.9969AUROCVGGish0.8578520.8939510.930050.966149May 7, 2026Evaluation resultsMethodMethodLinksAUROCSeparation RatioVGGishepsilon (ε)=0.05epsilon (ε)=0.052026.050.996963.82CLAPepsilon (ε)=0.05epsilon (ε)=0.052026.050.995110.04EnCodecepsilon (ε)=0.05epsilon (ε)=0.052026.050.964235.93PANNsepsilon (ε)=0.05epsilon (ε)=0.052026.050.896643.25AudioMAEepsilon (ε)=0.05epsilon (ε)=0.052026.050.863253.42