Speech Recognition with Timestamps on AMI
64.8AASCrisperWhisper
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CrisperWhisper2026.04 | 64.8 | — | |
| Qwen-Audio2026.04 | 55.64 | 18.64 | |
| Mixed Training + Length AugLength Augmentation=true2026.04 | 46.17 | 0 | |
| Mixed Training + Length Aug + Reduced TFLength Augmentation=true, Reduced Teacher Forcing=true2026.04 | 44.61 | 0 | |
| Mixed Training + Length Aug + Timestamp RegLength Augmentation=true, Timestamp Regularization=true2026.04 | 43.48 | 0 | |
| Mixed Training2026.04 | 42.76 | 0 |