Pairwise Token Consistency on LibriSpeech clean-100 1.0 (train)
0.83Unigram Consistencywav2tok 2.0
Evaluation results
| Method | Links | ||
|---|---|---|---|
| wav2tok 2.0Tokens=2562026.06 | 0.83 | 0.75 | |
| wav2tok 2.0Tokens=5122026.06 | 0.81 | 0.71 | |
| wav2tokTokens=2562026.06 | 0.8 | 0.72 | |
| wav2tok 2.0Tokens=10242026.06 | 0.79 | 0.68 | |
| wav2tokTokens=5122026.06 | 0.77 | 0.65 | |
| BEST-STDTokens=2562026.06 | 0.74 | 0.59 | |
| BEST-STDTokens=5122026.06 | 0.73 | 0.57 | |
| wav2tokTokens=10242026.06 | 0.73 | 0.61 | |
| BEST-STDTokens=10242026.06 | 0.7 | 0.53 | |
| Speech TokenizerTokens=10242026.06 | 0.44 | 0.24 | |
| WavLM-BaseTokens=5122026.06 | 0.33 | 0.16 | |
| WavLM-BaseTokens=10002026.06 | 0.28 | 0.14 | |
| HuBERT-BaseTokens=5122026.06 | 0.25 | 0.08 | |
| HuBERT-BaseTokens=10002026.06 | 0.22 | 0.1 | |
| EncodecTokens=10242026.06 | 0.12 | 0.06 |