List your company
ResearchFormerly WizWand

Speech Representation Evaluation on VoxCeleb 1 (test)

876.1SUPERB ScoreSpeech-FT
742.356777.078811.8846.522Feb 18, 2025
Updated 6mo ago

Evaluation results

MethodLinks
Speech-FTFine-tuning Source=English (mono-lingual)2025.02876.1
Pre-trainedNote=Original HuBERT model2025.02870.2
Speech-FTFine-tuning Source=Spanish (cross-lingual)2025.02868.26
Stable-FTFine-tuning Source=English (mono-lingual)2025.02867.96
Speech-FTFine-tuning Source=Chinese (cross-lingual)2025.02866.51
Stable-FT (20%)Fine-tuning Source=Chinese (cross-lingual), Snapshot=20% of total fine-tuning steps2025.02832.37
Stable-FTFine-tuning Source=Chinese (cross-lingual)2025.02789.88
Stable-FTFine-tuning Source=Spanish (cross-lingual)2025.02747.5