Discrete Unit Quality Evaluation on LibriSpeech (960h)
6.31ABXSpidR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SpidRLayer=6, Units=Codebook, Zero-shot protocol=true2025.12 | 6.31 | 0.602 | |
| WavLM BaseLayer=11, Units=K-means, Zero-shot protocol=true2025.12 | 7.01 | 0.667 | |
| SpidRLayer=6, Units=K-means, Zero-shot protocol=true2025.12 | 7.2 | 0.636 | |
| HuBERTLayer=11, Units=K-means, Zero-shot protocol=true2025.12 | 7.32 | 0.637 | |
| DinoSRLayer=5, Units=Codebook, Implementation=re-implementation, Zero-shot protocol=true2025.12 | 7.73 | 0.614 | |
| DinoSRLayer=5, Units=Codebook, Zero-shot protocol=true2025.12 | 7.92 | 0.62 | |
| wav2vec 2.0Layer=6, Units=K-means, Zero-shot protocol=true2025.12 | 9.33 | 0.609 | |
| DinoSRLayer=5, Units=K-means, Implementation=re-implementation, Zero-shot protocol=true2025.12 | 10.54 | 0.591 | |
| DinoSRLayer=5, Units=K-means, Zero-shot protocol=true2025.12 | 10.87 | 0.558 |