Speech-to-text translation on English-to-Chinese (En->Zh)
49.1BLEU ScoreVIRBA
Evaluation Results
| Method | Links | |
|---|---|---|
| VIRBABackbone=Qwen2.5-Omni2025.08 | 49.1 | |
| VIRBABackbone=Qwen2-Audio2025.08 | 48.4 | |
| Kimi-Audio2025.08 | 48.2 | |
| Qwen2.5-Omni2025.08 | 48 | |
| Group-DPO2025.08 | 47.3 | |
| Step-Audio-R12025.08 | 47.1 | |
| AF32025.08 | 46.8 | |
| SeamlessM4T-L V22025.08 | 46.2 | |
| Qwen2-Audiovariant=base2025.08 | 45.2 | |
| Whisper-large V22025.08 | 40.8 |