Speech Recognition with Timestamps on LibriSpeech Clean
12.44AASMixed Training + Length Aug + Reduced TF
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Mixed Training + Length Aug + Reduced TFLength Augmentation=true, Reduced Teacher Forcing=true2026.04 | 12.44 | 4 | |
| Mixed Training + Length Aug + Timestamp RegLength Augmentation=true, Timestamp Regularization=true2026.04 | 12.61 | 8 | |
| Mixed Training + Length AugLength Augmentation=true2026.04 | 13.37 | 0 | |
| Qwen-Audio2026.04 | 16.67 | 57 | |
| CrisperWhisper2026.04 | 30.2 | — | |
| Mixed Training2026.04 | 33.75 | 42 |