List your company

Text-Audio Alignment on Baton-Pair two-sound-event

71Pair AccuracyAQAScore
13.38428.34243.358.258Jan 21, 2026

Evaluation results

MethodLinks
AQAScoreBackbone=Qwen2.5-Omni-7B2026.0171
AQAScoreBackbone=Qwen2.5-Omni-3B2026.0170.3
CLAPScoreVariant=630k-best2026.0169
AQAScoreBackbone=AF32026.0166.3
CLAPScoreTraining Data=Music+Speech2026.0166.2
AQAScoreBackbone=AF3-Think2026.0165.8
CLAPScoreTraining Data=Music+AS2026.0165.8
CLAPScoreTraining Data=Music+Speech+AS2026.0164.7
CLAPScoreTraining Data=630k-AS2026.0164.6
AQAScoreBackbone=AF3-Chat2026.0163.1
PromptingBackbone=Qwen2.5-Omni-7B2026.0155.8
PromptingBackbone=AF3-Chat2026.0154.6
PromptingBackbone=AF32026.0150.6
PromptingBackbone=Qwen2.5-Omni-3B2026.0150.3
PromptingBackbone=AF3-Think2026.0149.8
CascadeBackbone=Qwen2.5-Omni-7B2026.0146.1
CascadeBackbone=AF3-Chat2026.0145.5
CascadeBackbone=AF32026.0143.3
CascadeBackbone=Qwen2.5-Omni-3B2026.0137.3
CascadeBackbone=AF3-Think2026.0115.6