Single-emotion Steering on ESD, RAVDESS, and CREMA-D (in-distribution)
69.8E-SIM ScoreCoCoEmo
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CoCoEmoBackbone=IndexTTS2, alpha=62026.02 | 69.8 | 0.433 | 92.1 | 5.24 | |
| CoCoEmoBackbone=IndexTTS2, alpha=32026.02 | 60.4 | 0.254 | 92 | 4.94 | |
| InstructionBackbone=CosyVoice22026.02 | 60 | 0.278 | 88.6 | 4.5 | |
| CoCoEmoBackbone=CosyVoice2, alpha=62026.02 | 58.9 | 0.366 | 90.6 | 5.33 | |
| Emo_vectorBackbone=IndexTTS22026.02 | 51.8 | 0.226 | 71.8 | 5.7 | |
| CoCoEmoBackbone=CosyVoice2, alpha=32026.02 | 51.5 | 0.134 | 90.4 | 4.83 | |
| No-steerBackbone=IndexTTS22026.02 | 46.9 | 0.041 | 91.8 | 4.87 | |
| No-steerBackbone=CosyVoice22026.02 | 44.4 | 0.023 | 89.9 | 4.31 |