Streaming Automatic Speech Recognition on LibriSpeech (test-clean)
3.1WERTransducer
Evaluation results
| Method | Links | |
|---|---|---|
| Transducer2026.05 | 3.1 | |
| Chunkwise AlignerAlignment type=ground-truth, Delay=320ms2026.05 | 3.2 | |
| Chunkwise AlignerAlignment type=ground-truth, Delay=480ms2026.05 | 3.4 | |
| Chunkwise AlignerAlignment type=ground-truth, Delay=160ms2026.05 | 3.5 | |
| Chunkwise AlignerAlignment type=streaming CTC, Delay=0ms2026.05 | 3.6 | |
| Chunkwise AlignerAlignment type=ground-truth, Delay=0ms2026.05 | 3.9 | |
| CTC2026.05 | 4.1 |