Paralinguistic-aware dialogue on ParaS2SBench
4.525Synthetic Age ScoreGPT-TTS
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| GPT-TTSCategory=Topline2025.11 | 4.525 | 4.691 | 4.812 | 4.791 | 4.705 | 4.71 | 4.824 | 4.766 | 4.725 | |
| Kimi-Audio GRPOCategory=Ours2025.11 | 4.496 | 4.49 | 4.239 | 4.538 | 4.441 | 4.394 | 3.927 | 4.161 | 4.382 | |
| Kimi-Audio SFTCategory=Ours2025.11 | 4.393 | 4.09 | 3.53 | 4.291 | 4.076 | 4.121 | 3.307 | 3.714 | 3.955 | |
| GeminiCategory=Closed Source2025.11 | 3.301 | 3.811 | 3.413 | 3.263 | 3.447 | 3.813 | 3.712 | 3.762 | 3.552 | |
| gpt-4o-audio-previewCategory=Closed Source2025.11 | 3.205 | 3.633 | 3.342 | 2.957 | 3.284 | 3.77 | 3.508 | 3.639 | 3.403 | |
| Qwen2.5 OmniCategory=Open Source2025.11 | 3.17 | 3.653 | 3.236 | 2.935 | 3.248 | 3.626 | 3.599 | 3.612 | 3.369 | |
| Kimi-AudioCategory=Open Source2025.11 | 3.141 | 2.673 | 3.091 | 2.665 | 2.892 | 1.365 | 1.166 | 1.265 | 2.35 | |
| LLaMa-Omni 2Category=Open Source2025.11 | 3.123 | 3.512 | 3.064 | 3.164 | 3.215 | 3.425 | 3.462 | 3.443 | 3.291 | |
| Whisper-GPT-TTSCategory=Baseline2025.11 | 3.05 | 3.121 | 2.916 | 3.005 | 3.022 | 3.562 | 3.412 | 3.487 | 3.176 | |
| GLM 4Category=Open Source2025.11 | 2.885 | 3.447 | 2.976 | 2.803 | 3.033 | 2.934 | 3.141 | 3.037 | 3.034 | |
| Freeze-OmniCategory=Open Source2025.11 | 2.819 | 2.316 | 2.884 | 2.701 | 2.68 | 2.835 | 3.061 | 2.948 | 2.769 |