Speech Naturalness Assessment on CVSS-T
3.67UTMOSTri-Task Learning
Evaluation Results
| Method | Links | |
|---|---|---|
| Tri-Task LearningSystem=End-to-end, Optimization Stage=After DPO, Back-translation Evaluation Metric=BLEU2026.01 | 3.67 | |
| Tri-Task LearningSystem=End-to-end, Optimization Stage=After DPO, Back-translation Evaluation Metric=METEOR2026.01 | 3.66 | |
| VanillaSystem=End-to-end, Optimization Stage=Before DPO2026.01 | 3.64 | |
| VanillaSystem=End-to-end, Optimization Stage=After DPO, Back-translation Evaluation Metric=METEOR2026.01 | 3.64 | |
| Tri-Task LearningSystem=End-to-end, Optimization Stage=Before DPO2026.01 | 3.63 | |
| Chain of ModalitySystem=End-to-end, Optimization Stage=After DPO, Back-translation Evaluation Metric=METEOR2026.01 | 3.61 | |
| VanillaSystem=End-to-end, Optimization Stage=After DPO, Back-translation Evaluation Metric=BLEU2026.01 | 3.61 | |
| Chain of ModalitySystem=End-to-end, Optimization Stage=After DPO, Back-translation Evaluation Metric=BLEU2026.01 | 3.61 | |
| Chain of ModalitySystem=End-to-end, Optimization Stage=Before DPO2026.01 | 3.6 | |
| S2T+TTSSystem=Cascaded2026.01 | 3.08 |