Speech Recognition with Timestamps on VoxPopuli (VOXP)
116.37AASMixed Training + Length Aug
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Mixed Training + Length AugLength Augmentation=true2026.04 | 116.37 | 0.11 | |
| Mixed Training + Length Aug + Reduced TFLength Augmentation=true, Reduced Teacher Forcing=true2026.04 | 77.51 | 0.22 | |
| Mixed Training2026.04 | 76.96 | 0.49 | |
| Mixed Training + Length Aug + Timestamp RegLength Augmentation=true, Timestamp Regularization=true2026.04 | 73.72 | 0.33 | |
| CrisperWhisper2026.04 | 54.93 | — |