Audio-Visual Speech Recognition on LRS2 Noise-free
2.34Overall Error RateLlama-AVSR
Evaluation Results
| Method | Links | |
|---|---|---|
| Llama-AVSRTraining=Clean2026.06 | 2.34 | |
| VIB-AVSRTraining=Noisy2026.06 | 2.38 | |
| VIB-AVSRTraining=Clean2026.06 | 2.42 | |
| Llama-AVSRTraining=Noisy2026.06 | 2.72 |