Speech Recognition on MCV Seen Accents 100H (test)
4.4Error Rate (AUS)MOE-CTC
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MOE-CTCEncoder size=Large, Parameter size=131.90M2026.02 | 4.4 | 7.8 | 6.1 | 3.2 | 5.9 | 5.5 | |
| ACCENT-MOEEncoder size=Large, Parameter size=123.48M2026.02 | 4.9 | 7.9 | 6.8 | 3.6 | 6.1 | 5.9 | |
| MoEEncoder size=Large, Parameter size=123.48M2026.02 | 5.3 | 8.8 | 7.3 | 4 | 6.8 | 6.4 | |
| FastConformerEncoder size=Large, Parameter size=115.60M2026.02 | 5.5 | 9.4 | 7.8 | 3.9 | 7.7 | 6.9 | |
| Inter-CTCEncoder size=Large, Parameter size=115.60M2026.02 | 5.6 | 9.2 | 7.9 | 4.1 | 7.5 | 6.9 | |
| MOE-CTCEncoder size=Medium, Parameter size=32.58M2026.02 | 5.9 | 8.7 | 8.4 | 4.3 | 7.2 | 6.9 | |
| ACCENT-MOEEncoder size=Medium, Parameter size=28.37M2026.02 | 6.3 | 9.3 | 8.5 | 4.7 | 8 | 7.4 | |
| MoEEncoder size=Medium, Parameter size=28.37M2026.02 | 7 | 10.3 | 9.7 | 5 | 9.9 | 8.4 | |
| FastConformerEncoder size=Medium, Parameter size=26.39M2026.02 | 7.3 | 11.4 | 10.3 | 5.7 | 10 | 8.9 | |
| Inter-CTCEncoder size=Medium, Parameter size=26.39M2026.02 | 7.3 | 11.2 | 10.4 | 5.1 | 9.8 | 8.8 | |
| MOE-CTCEncoder size=Small, Parameter size=16.62M2026.02 | 8.2 | 10.5 | 11.4 | 6.6 | 10.1 | 9.4 | |
| ACCENT-MOEEncoder size=Small, Parameter size=13.72M2026.02 | 8.5 | 11.2 | 12.2 | 7.5 | 10.8 | 10 | |
| MoEEncoder size=Small, Parameter size=13.72M2026.02 | 9.8 | 12.7 | 12.9 | 8.6 | 11.5 | 11.1 | |
| Inter-CTCEncoder size=Small, Parameter size=12.78M2026.02 | 11.3 | 14.3 | 15.1 | 11 | 13.2 | 13 | |
| FastConformerEncoder size=Small, Parameter size=12.78M2026.02 | 11.6 | 15.5 | 15.1 | 11.4 | 13.1 | 13.3 |