Multilingual Automatic Speech Recognition on Common Voice Exp. 3 (test)
15.4WER (US)Initial model
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Initial model2026.01 | 15.4 | 45.7 | 27.6 | 29.8 | 44.8 | 32.64 | — | |
| Sep. Model2026.01 | 15.4 | 26.3 | 13.6 | 12.2 | 22.1 | 17.9 | 0 | |
| FTA2026.01 | 19.9 | 30.8 | 18 | 18.5 | 29 | 23.22 | -0.7 | |
| SVRMemory size (|M|)=502026.01 | 21.2 | 31.1 | 17.1 | 16.4 | 27.2 | 22.6 | -1.2 | |
| SVRMemory size (|M|)=202026.01 | 21.9 | 31.5 | 17.1 | 16.5 | 27.6 | 22.93 | -1.8 | |
| SVRMemory size (|M|)=1t2026.01 | 23.5 | 31.5 | 17.2 | 16.2 | 27.5 | 23.15 | -1.7 | |
| CLRL-Tuning2026.01 | 24.8 | 43.5 | 23.3 | 21.5 | 31.8 | 28.99 | -4.1 | |
| ERMemory size (|M|)=2002026.01 | 25.9 | 34 | 17.5 | 14.6 | 23.2 | 23.04 | -5.6 | |
| KDMemory size (|M|)=2002026.01 | 27.2 | 32.2 | 16.1 | 14.2 | 23.4 | 22.61 | -5.3 | |
| ERMemory size (|M|)=202026.01 | 33.9 | 39.6 | 18.2 | 15.6 | 22.8 | 26 | -9.7 | |
| KDMemory size (|M|)=202026.01 | 44.3 | 39.8 | 17.2 | 14.3 | 22.2 | 27.55 | -12 | |
| Fine-Tuning2026.01 | 44.6 | 47.1 | 18 | 14.6 | 22.1 | 29.29 | -14.2 | |
| LWF2026.01 | 50.7 | 62.3 | 19.3 | 14 | 20.8 | 33.39 | -19.9 |