Automatic Speech Recognition on LibriSpeech (WER, CER)
3.44Word Error Rate (WER)Whisper Small
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Whisper Small2026.03 | 3.44 | 1.24 | |
| SW2VStreaming=true2026.03 | 4.11 | 1.98 | |
| WavLM-LargeStreaming=false, Variant=Full2026.03 | 4.23 | 2.74 | |
| W2V2Streaming=false, Variant=Full2026.03 | 4.73 | 2.17 | |
| W2V-BERT 2.0Streaming=false, Layer=17th2026.03 | 4.94 | 2.92 | |
| DAC-8Streaming=false2026.03 | 5 | 2.6 | |
| JHCodec-M-8Streaming=true2026.03 | 5.53 | 3.04 | |
| NanoCodecStreaming=false2026.03 | 7.26 | 4.03 | |
| Mimi-32Streaming=true2026.03 | 8.75 | 5.48 |