Automatic Speech Recognition on SUPERB 16kHz
5.89WERMSRHuBERT
Evaluation results
| Method | Links | |
|---|---|---|
| MSRHuBERTFine-tuning protocol=Single-rate, Pre-training sampling rate=Multi-rate2026.03 | 5.89 | |
| Re. 16kHz HuBERT BaseFine-tuning protocol=Single-rate, Resampling in fine-tuning=Yes, Pre-training sampling rate=16kHz2026.03 | 6.03 | |
| Re. 16kHz HuBERT Base (w/o resampling in fine-tuning)Fine-tuning protocol=Single-rate, Resampling in fine-tuning=No, Pre-training sampling rate=16kHz2026.03 | 6.03 | |
| Re. 24kHz HuBERT BaseFine-tuning protocol=Single-rate, Resampling in fine-tuning=Yes, Pre-training sampling rate=24kHz2026.03 | 6.15 | |
| Re. 48kHz HuBERT BaseFine-tuning protocol=Single-rate, Resampling in fine-tuning=Yes, Pre-training sampling rate=48kHz2026.03 | 6.37 | |
| HuBERT BaseFine-tuning protocol=Single-rate, Resampling in fine-tuning=Yes, Pre-training sampling rate=16kHz2026.03 | 6.41 | |
| MSRHuBERTFine-tuning protocol=Mixed-rate, Pre-training sampling rate=Multi-rate2026.03 | 6.54 | |
| Re. 16kHz HuBERT BaseFine-tuning protocol=Mixed-rate, Resampling in fine-tuning=Yes, Pre-training sampling rate=16kHz2026.03 | 6.89 | |
| Re. 24kHz HuBERT BaseFine-tuning protocol=Mixed-rate, Resampling in fine-tuning=Yes, Pre-training sampling rate=24kHz2026.03 | 6.95 | |
| Re. 48kHz HuBERT BaseFine-tuning protocol=Mixed-rate, Resampling in fine-tuning=Yes, Pre-training sampling rate=48kHz2026.03 | 7.28 | |
| HuBERT BaseFine-tuning protocol=Mixed-rate, Resampling in fine-tuning=Yes, Pre-training sampling rate=16kHz2026.03 | 7.42 | |
| Re. 16kHz HuBERT Base (w/o resampling in fine-tuning)Fine-tuning protocol=Mixed-rate, Resampling in fine-tuning=No, Pre-training sampling rate=16kHz2026.03 | 7.69 |