Acoustic Discriminability (ABX) on 5 Languages (sw, ta, th, tr, uk) (dev)
3.58Triphone ABX (WS)MAUBERT PHONE + MPR
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| MAUBERT PHONE + MPRLayer=12, # units=768, Evaluation Protocol=supervised FT (10 h)2025.12 | 3.58 | 4.49 | 1.79 | 2.3 | 2.88 | 3.35 | 3.07 | |
| MAUBERT FEAT + MPRLayer=11, # units=768, Evaluation Protocol=supervised FT (10 h)2025.12 | 3.65 | 4.38 | 1.83 | 2.28 | 3.17 | 3.44 | 3.13 | |
| HuBERT-base + MPRLayer=11, # units=768, Evaluation Protocol=supervised FT (10 h)2025.12 | 4.26 | 4.98 | 2.05 | 2.62 | 3.94 | 4.3 | 3.69 | |
| MAUBERT-FEAT + K-means (L9)Layer=10, # units=768, Evaluation Protocol=self-supervised FT (10 h)2025.12 | 4.72 | 5.5 | 2.58 | 3.31 | 5.08 | 5.59 | 4.46 | |
| MAUBERT-PHONE + phone freq.Layer=10, # units=768, Evaluation Protocol=self-supervised FT (10 h)2025.12 | 4.77 | 5.78 | 2.49 | 3.17 | 4.82 | 5.26 | 4.38 | |
| HuBERT-base + PRLayer=ws, # units=768, Evaluation Protocol=supervised FT (10 h)2025.12 | 4.87 | 6.13 | 2.3 | 3.09 | 3.65 | 4.17 | 4.04 | |
| MAUBERT-FEAT + feat. freq.Layer=9, # units=768, Evaluation Protocol=self-supervised FT (10 h)2025.12 | 4.88 | 5.65 | 2.63 | 3.28 | 5.24 | 5.66 | 4.56 | |
| MAUBERT-PHONE + K-means (phone)Layer=10, # units=768, Evaluation Protocol=self-supervised FT (10 h)2025.12 | 4.88 | 5.83 | 2.7 | 3.4 | 5.29 | 5.79 | 4.65 | |
| MAUBERT-PHONE + all phonesLayer=10, # units=768, Evaluation Protocol=self-supervised FT (10 h)2025.12 | 4.88 | 5.84 | 2.49 | 3.16 | 4.85 | 5.28 | 4.42 | |
| MAUBERT-PHONE + K-means (proj)Layer=10, # units=768, Evaluation Protocol=self-supervised FT (10 h)2025.12 | 4.91 | 5.71 | 2.66 | 3.32 | 4.93 | 5.55 | 4.51 | |
| MAUBERT-FEAT + K-means (feat)Layer=9, # units=768, Evaluation Protocol=self-supervised FT (10 h)2025.12 | 5 | 5.81 | 2.69 | 3.41 | 5.29 | 5.69 | 4.65 | |
| MAUBERT-FEAT + phone freq.Layer=9, # units=768, Evaluation Protocol=self-supervised FT (10 h)2025.12 | 5.01 | 5.9 | 2.62 | 3.35 | 5.21 | 5.62 | 4.62 | |
| MAUBERT PHONELayer=proj, # units=1024, Evaluation Protocol=Zero-shot2025.12 | 5.42 | 6.46 | 2.96 | 3.79 | 5.49 | 6.12 | 5.04 | |
| MAUBERT FEATLayer=9, # units=768, Evaluation Protocol=Zero-shot2025.12 | 5.49 | 6.52 | 2.95 | 3.81 | 5.97 | 6.47 | 5.2 | |
| HuBERT-base + K-means (L11)Layer=10, # units=768, Evaluation Protocol=self-supervised FT (10 h)2025.12 | 5.71 | 6.64 | 3.15 | 4.09 | 7.13 | 7.58 | 5.72 | |
| WavLM-base+Layer=7, # units=768, Evaluation Protocol=Zero-shot2025.12 | 5.94 | 6.97 | 3.22 | 4.18 | 7.34 | 8.01 | 5.94 | |
| WavLM-largeLayer=24, # units=1024, Evaluation Protocol=Zero-shot2025.12 | 5.94 | 7 | 3.19 | 4.14 | 7.92 | 8.24 | 6.07 | |
| XEUSLayer=18, # units=1024, Evaluation Protocol=Zero-shot2025.12 | 6.14 | 7.15 | 3.58 | 4.52 | 9.28 | 9.45 | 6.69 | |
| HuBERT-baseLayer=11, # units=768, Evaluation Protocol=Zero-shot2025.12 | 6.77 | 8.18 | 3.77 | 4.92 | 8.55 | 9.19 | 6.9 | |
| mHuBERT-147Layer=7, # units=768, Evaluation Protocol=Zero-shot2025.12 | 7.37 | 8.64 | 3.7 | 4.8 | 9 | 9.51 | 7.17 | |
| MMS-1BLayer=34, # units=1280, Evaluation Protocol=Zero-shot2025.12 | 9.37 | 10.74 | 4.76 | 6.02 | 10.53 | 11.37 | 8.8 | |
| MFCCLayer=-, # units=39, Evaluation Protocol=Zero-shot2025.12 | 20 | 29 | 13.23 | 22.36 | 18.05 | 26.33 | 21.49 |