Speech-to-Speech Translation on CVSS-T en→zh
32.5ASR-BLEUMoVE
Evaluation Results
| Method | Links | |
|---|---|---|
| MoVE2026.04 | 32.5 | |
| Kimi + LoRATraining data source=Ours, Training duration (hours)=50h2026.04 | 32 | |
| Kimi + LoRATraining data source=Ours, Training duration (hours)=100h2026.04 | 31.2 | |
| Kimi + LoRATraining data source=SynStard, Training duration (hours)=100h2026.04 | 29.9 | |
| gpt-4o-audio-preview2026.04 | 26.3 | |
| SeamlessM4T-Large-v22026.04 | 25.8 | |
| Kimi-Audio-7B-Instruct2026.04 | 25 | |
| SeamlessExpressive2026.04 | 23.8 | |
| Kimi + LoRATraining data source=SeamlessAlign, Training duration (hours)=67h2026.04 | 15.7 | |
| CascadedOne-shot acoustic prompt=true2026.04 | 9.7 |