Automatic Speech Recognition on Common Voice (ENG, DEU, ESP) and CGN (NL, VL) OWSM v3.2 (test)
13.4WER (ENG)Initial model
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Initial modelParams=–2026.06 | 13.4 | 11.3 | 11.3 | 45.7 | 37.9 | 22.48 | — | |
| Separate ModelParams=244.8M2026.06 | 13.4 | 11.3 | 11.3 | 22.4 | 13.6 | 14.38 | 0 | |
| CSSVDParams=8.9M2026.06 | 14.9 | 14.4 | 13.4 | 28.6 | 20.4 | 18.33 | -1.9 | |
| LoRA + FTAParams=9.3M2026.06 | 16.4 | 20.6 | 15.1 | 27.2 | 19 | 19.64 | -3.6 | |
| BiLoRAParams=9.3M2026.06 | 19.4 | 34.7 | 17.8 | 30 | 16 | 23.58 | -9.7 | |
| EWC-LoRAParams=9.3M2026.06 | 23.9 | 60.5 | 23.9 | 25.8 | 17 | 30.21 | -18.6 | |
| Full Fine-TuningParams=244.8M2026.06 | 24.3 | 57.3 | 21.7 | 27.7 | 13.6 | 28.94 | -18.2 | |
| OPLoRAParams=9.3M2026.06 | 33.8 | 81.6 | 36.3 | 30.2 | 15.7 | 39.53 | -30.3 | |
| SSVDParams=9.0M2026.06 | 40.8 | 85.3 | 43.1 | 37.4 | 15.9 | 44.49 | -36.4 | |
| LoRAParams=9.3M2026.06 | 41.8 | 88.4 | 44.5 | 28.9 | 14.7 | 43.66 | -35.7 | |
| MiLoRAParams=9.3M2026.06 | 47.4 | 91.7 | 48.5 | 30.2 | 14.1 | 46.39 | -39.6 |