Text-Audio Alignment on RELATE-Pair
77.6Pair AccuracyAQAScore
Evaluation Results
| Method | Links | |
|---|---|---|
| AQAScoreBackbone=Qwen2.5-Omni-7B2026.01 | 77.6 | |
| CLAPScoreTraining Data=Music+AS2026.01 | 76 | |
| CLAPScoreTraining Data=630k-AS2026.01 | 75.2 | |
| AQAScoreBackbone=Qwen2.5-Omni-3B2026.01 | 75 | |
| AQAScoreBackbone=AF32026.01 | 73.6 | |
| AQAScoreBackbone=AF3-Think2026.01 | 72.4 | |
| AQAScoreBackbone=AF3-Chat2026.01 | 72.1 | |
| CLAPScoreTraining Data=Music+Speech2026.01 | 71 | |
| CLAPScoreVariant=630k-best2026.01 | 71 | |
| CLAPScoreTraining Data=Music+Speech+AS2026.01 | 69.8 | |
| CascadeBackbone=Qwen2.5-Omni-3B2026.01 | 68.6 | |
| CascadeBackbone=Qwen2.5-Omni-7B2026.01 | 67.1 | |
| PromptingBackbone=Qwen2.5-Omni-3B2026.01 | 66.2 | |
| CascadeBackbone=AF32026.01 | 64.5 | |
| PromptingBackbone=AF3-Chat2026.01 | 63.3 | |
| PromptingBackbone=Qwen2.5-Omni-7B2026.01 | 61 | |
| CascadeBackbone=AF3-Chat2026.01 | 60 | |
| CascadeBackbone=AF3-Think2026.01 | 55.6 | |
| PromptingBackbone=AF32026.01 | 44.5 | |
| PromptingBackbone=AF3-Think2026.01 | 41.7 |