Automatic Speech Recognition on Urdu (test)
17.2WERWhisper Large v3
Evaluation results
| Method | Links | |
|---|---|---|
| Whisper Large v3Number of Parameters=1.5B, Fine-tuning=on labeled data2025.12 | 17.2 | |
| CP2Initialization=W2V Large init., Number of Parameters=300M, Fine-tuning=on labeled data2025.12 | 20.6 | |
| CP1Initialization=XLS-R init., Number of Parameters=300M, Fine-tuning=on labeled data2025.12 | 22.2 | |
| CSInitialization=from scratch, Number of Parameters=95M, Fine-tuning=on labeled data2025.12 | 25 | |
| Seamless Large v2Number of Parameters=2.3B, Fine-tuning=on labeled data2025.12 | 31.9 | |
| W2V LargeNumber of Parameters=300M, Fine-tuning=on labeled data2025.12 | 39.3 |