Behavioral Accuracy on Natural Voice and Free Generation dataset
75.5Text-Affinity Accuracy (Abase -> TCF)Qwen2-Audio
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Qwen2-AudioLanguage=AR2026.06 | 75.5 | 51.1 | 61.3 | 0.226 | 68.4 | 86.9 | |
| Qwen2-AudioLanguage=DE2026.06 | 73.1 | 46.2 | 46.3 | -0.073 | 59.7 | 94.7 | |
| Qwen2-AudioLanguage=EN2026.06 | 71.7 | 43.3 | 36.7 | -0.266 | 54.2 | 92.8 | |
| Qwen2-AudioLanguage=FR2026.06 | 71.5 | 43.1 | 45.9 | -0.081 | 58.7 | 92.9 | |
| Qwen2-AudioLanguage=PT2026.06 | 70.4 | 40.7 | 45.8 | -0.084 | 58.1 | 91.2 | |
| Qwen2-AudioLanguage=IT2026.06 | 69.6 | 39.3 | 37.5 | -0.25 | 53.5 | 95.5 | |
| Qwen2-AudioLanguage=ZH2026.06 | 68.8 | 37.6 | 42.3 | -0.153 | 55.5 | 95.4 | |
| Qwen2-AudioLanguage=JA2026.06 | 64.3 | 28.5 | 48.8 | -0.024 | 56.5 | 88.5 | |
| UltravoxLanguage=FR2026.06 | 59.6 | 19.2 | 56 | 0.12 | 57.8 | 90 | |
| UltravoxLanguage=AR2026.06 | 59.2 | 18.4 | 64 | 0.28 | 61.6 | 92.2 | |
| UltravoxLanguage=JA2026.06 | 55.6 | 11.2 | 66 | 0.32 | 60.8 | 89.6 | |
| UltravoxLanguage=PT2026.06 | 55.6 | 11.2 | 55.2 | 0.104 | 55.4 | 92.8 | |
| UltravoxLanguage=ZH2026.06 | 55.2 | 10.4 | 70.4 | 0.408 | 62.8 | 94 | |
| UltravoxLanguage=IT2026.06 | 54 | 8 | 51.6 | 0.032 | 52.8 | 95.8 | |
| UltravoxLanguage=DE2026.06 | 52.8 | 5.6 | 63.6 | 0.272 | 58.2 | 94.4 | |
| UltravoxLanguage=EN2026.06 | 52.4 | 4.8 | 50.4 | 0.008 | 51.4 | 94.4 |