Automatic Speech Recognition on In-house English Conversational Speech (test)
11.1Agent WERIn-house SE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| In-house SECategory=Finetuning with Pseudo Labels +, No. of Params=94.4M2026.06 | 11.1 | 13.9 | 12.3 | |
| NeMoCategory=Finetuning with Pseudo Labels +, No. of Params=114M/121M*2026.06 | 11.6 | 15.2 | 13.2 | |
| In-house SECategory=Speech Encoder finetuning, No. of Params=94.4M2026.06 | 13 | 18.4 | 15.4 | |
| NeMoCategory=ASR finetuning, No. of Params=114M/121M*2026.06 | 13.2 | 17.8 | 15.2 | |
| SpeechBrain ModelCategory=Trained from Scratch, No. of Params=94.4M2026.06 | 18.7 | 30.9 | 24.1 | |
| Google TelephonyCategory=Commercial, No. of Params=-2026.06 | 25 | 29.7 | 27.2 | |
| MMSCategory=Speech Encoder finetuning, No. of Params=300M2026.06 | 25.2 | 30.4 | 27.3 | |
| MMSCategory=Speech Encoder finetuning, No. of Params=1B2026.06 | 26.6 | 32.4 | 28.9 | |
| Whisper-large-v3Category=SoTA Open Source, No. of Params=1.5B2026.06 | 27.5 | 30.3 | 28.7 | |
| NeMoCategory=SoTA Open Source, No. of Params=114M/121M*2026.06 | 27.7 | 31.5 | 29.5 |