Automatic Speech Recognition on FM
5.8WERMedASR
Evaluation results
| Method | Links | |
|---|---|---|
| MedASRLanguage Model=6-gram SentencePiece LM, Decoding=beam search2026.05 | 5.8 | |
| MedASRLanguage Model=None, Decoding=greedy2026.05 | 6.7 | |
| Gemini 2.5 Pro2026.05 | 14.6 | |
| Whisper (Large-v3)2026.05 | 32.5 |