List your company
ResearchFormerly WizWand

Keyword Spotting on Speech Commands KS1 v1

98.8AccuracySSLAM
93.80895.10496.497.696Jul 13, 2022Jan 6, 2023Jul 3, 2023Dec 28, 2023Jun 22, 2024Dec 17, 2024Jun 13, 2025
Updated 8mo ago

Evaluation results

MethodLinks
SSLAM#Param=88M, Pre-training Data=AS2025.0698.8
ASIT#Param=86M, Pre-training Data=AS2025.0698.2
BEATSiter3+#Param=90M, Pre-training Data=AS, Supervised Fine-tuning=Included2025.0698.1
BEATSiter3#Param=90M, Pre-training Data=AS2025.0697.7
A-JEPA#Param=86M, Pre-training Data=AS2025.0697.7
Audio-MAE (global)Backbone=ViT-B, PT-Data=AS2022.0797.6
Audio-MAE#Param=86M, Pre-training Data=AS2025.0696.9
Audio-MAE (local)Backbone=ViT-B, PT-Data=AS2022.0796.9
SSAMBAModel size=base, Pretrained=true, Head learning rate=normal2024.0596.9
SSAMBAModel size=small, Pretrained=true, Head learning rate=larger2024.0596.4
HuBERTBackbone=Transformer, PT-Data=LS, Evaluation Protocol=Linear Evaluation2022.0796.3
wav2vec 2.0Backbone=Transformer, PT-Data=LS, Evaluation Protocol=Linear Evaluation2022.0796.2
SS-AST#Param=89M, Pre-training Data=AS+LS2025.0696
SS-ASTBackbone=ViT-B, PT-Data=AS+LS2022.0796
SSASTModel size=base, Pretrained=true, Head learning rate=normal2024.0596
MAE-AST#Param=86M, Pre-training Data=AS+LS2025.0695.8
MAE-ASTBackbone=ViT-B, PT-Data=AS+LS2022.0795.8
SSAMBAModel size=small, Pretrained=false, Head learning rate=larger2024.0595.7
SSAMBAModel size=base, Pretrained=false, Head learning rate=normal2024.0595.7
ASTBackbone=DeiT-B, PT-Data=IN2022.0795.5
SSASTModel size=small, Pretrained=true, Head learning rate=larger2024.0595.4
SSASTModel size=tiny, Pretrained=true, Head learning rate=larger2024.0594.8
SSAMBAModel size=tiny, Pretrained=false, Head learning rate=larger2024.0594.7
SSAMBAModel size=tiny, Pretrained=true, Head learning rate=larger2024.0594