Speech Recognition on LibriSpeech (100h)
6.2WER (%)Li-GRU
Evaluation Results
| Method | Links | |
|---|---|---|
| Li-GRUToolkit=PyTorch-Kaldi, Architecture=Li-GRU2018.11 | 6.2 | |
| GRUToolkit=PyTorch-Kaldi, Architecture=GRU2018.11 | 6.3 | |
| LSTMToolkit=PyTorch-Kaldi, Architecture=LSTM2018.11 | 6.4 | |
| MLPToolkit=PyTorch-Kaldi, Architecture=MLP2018.11 | 6.5 | |
| p-norm Kaldi baselineSubset=100h2018.11 | 6.5 | |
| Whisper-M + Llama 8B (4bit, r=4)Train=Joint, Decode=PF-S2T2026.06 | 6.8 | |
| Whisper-M + Llama 8B (4bit, r=4)Train=Joint, Decode=S2T2026.06 | 7 | |
| Whisper-M + Llama 8B (4bit, r=4)Train=S2T, Decode=S2T2026.06 | 7.3 | |
| Whisper-M + Llama 8B (4bit, r=4)Train=PF-S2T, Decode=PF-S2T2026.06 | 7.6 | |
| Whisper-MTrain=–, Decode=Beam2026.06 | 9.3 |