Emotional Text-to-Speech on Mandarin emotional speech dataset
3.78E-MOSRRPO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RRPOsupervised_fine_tuning=true, differentiable_reward_optimization=true, robust_reward_policy_optimization=true2025.12 | 3.78 | 3.81 | |
| CosyVoice2 + SFT + DiffROsupervised_fine_tuning=true, differentiable_reward_optimization=true2025.12 | 3.65 | 3.61 | |
| CosyVoice2 + SFTsupervised_fine_tuning=true2025.12 | 3.52 | 3.72 | |
| CosyVoice2baseline=true2025.12 | 3.27 | 3.65 |