Automatic Speech Recognition on LibriSpeech clean Speech Noise - Reverb (test)
19.8WERMask2Flow-TSE
Evaluation Results
| Method | Links | |
|---|---|---|
| Mask2Flow-TSEBackbone=Whisper medium.en, Params(M)=85.32026.03 | 19.8 | |
| No preprocessingBackbone=Whisper medium.en, Params(M)=N/A2026.03 | 21 | |
| Flow-onlyBackbone=Whisper medium.en, Params(M)=72.62026.03 | 21.9 | |
| Mask2Flow-TSEBackbone=Whisper small.en, Params(M)=85.32026.03 | 23.4 | |
| Flow-onlyBackbone=Whisper small.en, Params(M)=72.62026.03 | 25.7 | |
| MaskingBackbone=Whisper medium.en, Params(M)=12.72026.03 | 26.4 | |
| Mask2Flow-TSEBackbone=Whisper base.en, Params(M)=85.32026.03 | 28.1 | |
| ConVoiFilterBackbone=Whisper medium.en, Params(M)=49.92026.03 | 29.8 | |
| Flow-onlyBackbone=Whisper base.en, Params(M)=72.62026.03 | 31.1 | |
| MaskingBackbone=Whisper small.en, Params(M)=12.72026.03 | 33 | |
| Mask2Flow-TSEBackbone=Whisper tiny.en, Params(M)=85.32026.03 | 33.3 | |
| ConVoiFilterBackbone=Whisper small.en, Params(M)=49.92026.03 | 33.7 | |
| TSELMBackbone=Whisper medium.en, Params(M)=195.42026.03 | 36.5 | |
| Flow-onlyBackbone=Whisper tiny.en, Params(M)=72.62026.03 | 37.2 | |
| TSELMBackbone=Whisper small.en, Params(M)=195.42026.03 | 38.9 | |
| No preprocessingBackbone=Whisper small.en, Params(M)=N/A2026.03 | 40 | |
| ConVoiFilterBackbone=Whisper base.en, Params(M)=49.92026.03 | 40.8 | |
| TSELMBackbone=Whisper base.en, Params(M)=195.42026.03 | 41.4 | |
| MaskingBackbone=Whisper base.en, Params(M)=12.72026.03 | 41.9 | |
| Metis-TSEBackbone=Whisper medium.en, Params(M)=1425.02026.03 | 44.1 | |
| TSELMBackbone=Whisper tiny.en, Params(M)=195.42026.03 | 44.8 | |
| Metis-TSEBackbone=Whisper small.en, Params(M)=1425.02026.03 | 44.9 | |
| Metis-TSEBackbone=Whisper base.en, Params(M)=1425.02026.03 | 47.3 | |
| MaskingBackbone=Whisper tiny.en, Params(M)=12.72026.03 | 48.7 | |
| ConVoiFilterBackbone=Whisper tiny.en, Params(M)=49.92026.03 | 48.8 | |
| Metis-TSEBackbone=Whisper tiny.en, Params(M)=14252026.03 | 49.3 | |
| No preprocessingBackbone=Whisper base.en, Params(M)=N/A2026.03 | 49.4 | |
| No preprocessingBackbone=Whisper tiny.en, Params(M)=N/A2026.03 | 65.8 |