Automatic Speech Recognition on ASCEND (test)
17.8MERGround-Truth
Evaluation results
| Method | Links | |
|---|---|---|
| Ground-TruthTraining Data=Real recordings, Domain=Ground-truth2026.01 | 17.8 | |
| CosyVoice (SEAME-FT) gen.Training Data=Synthetic speech, Adaptation=SEAME fine-tuned, Backbone=Whisper-small2026.01 | 19.1 | |
| CosyVoice zero-shot gen.Training Data=Synthetic speech, Adaptation=None (zero-shot), Backbone=Whisper-small2026.01 | 25.2 |