List your company
ResearchFormerly WizWand

Synthetic Sound Detection on ESDD 2026 (dev)

0EERBEATs
-19109.25237.5365.75Dec 9, 2025
Updated 8mo ago

Evaluation results

MethodLinks
BEATsBackbone=BEATs, Back-end=MHFA2025.120
BEATs (FT-AS2M-CPT1)Backbone=BEATs, Back-end=MHFA, Pre-training=AS2M-CPT12025.120
BEATs (FT-AS2M-CPT2)Backbone=BEATs, Back-end=MHFA, Pre-training=AS2M-CPT22025.120
MAE++ BaseBackbone=MAE++, Back-end=MHFA, Model Size=Base2025.120
EAT BaseBackbone=EAT, Back-end=MHFA, Model Size=Base2025.120
EAT Base_FT(AS2M)Backbone=EAT, Back-end=MHFA, Model Size=Base, Pre-training=AS2M Fine-tuned2025.120
EAT Base_FT(AS2M)*Backbone=EAT, Back-end=MHFA-DSU, Model Size=Base, Pre-training=AS2M Fine-tuned, DSU (Feature Domain Augmentation)=true2025.120
EAT LargeBackbone=EAT, Back-end=MHFA, Model Size=Large2025.120
EAT Large_FT(AS2M)Backbone=EAT, Back-end=MHFA, Model Size=Large, Pre-training=AS2M Fine-tuned2025.120
Fusion (11+12)Fusion Strategy=Systems 11 and 122025.120
Fusion (11+12+14)Fusion Strategy=Systems 11, 12, and 142025.120
BEATs+AASISTBackbone=BEATs, Back-end=AASIST2025.1210
Dasheng-0.6BBackbone=Dasheng, Back-end=MHFA, Model Size=0.6B2025.1227
Dasheng-1.2BBackbone=Dasheng, Back-end=MHFA, Model Size=1.2B2025.1233
AASISTBackbone=AASIST2025.1292
WavLM Base +Backbone=WavLM, Back-end=MHFA, Model Size=Base2025.12475