Speech-to-text translation on English-to-German En->De
34BLEU ScoreVIRBA
Evaluation Results
| Method | Links | |
|---|---|---|
| VIRBABackbone=Qwen2.5-Omni2025.08 | 34 | |
| Kimi-Audio2025.08 | 33.6 | |
| Qwen2.5-Omni2025.08 | 33 | |
| Step-Audio-R12025.08 | 32.1 | |
| AF32025.08 | 31.5 | |
| VIRBABackbone=Qwen2-Audio2025.08 | 31.2 | |
| Group-DPO2025.08 | 30.3 | |
| SeamlessM4T-L V22025.08 | 29.8 | |
| Qwen2-Audiovariant=base2025.08 | 25.9 | |
| Whisper-large V22025.08 | 24.6 |