Audio-to-Text Retrieval on MSRVTT
36.4AccuracyBrokenBind
Evaluation results
| Method | Links | |
|---|---|---|
| BrokenBindBackbone=VCLIP+CLAP, Evaluation Protocol=BrokenBind, Source Dataset=AVE2026.02 | 36.4 | |
| VCLIP+CLAP+FTBackbone=VCLIP+CLAP, Evaluation Protocol=Fine-tuning, Source Dataset=AVE2026.02 | 17.6 | |
| VCLIP+CLAPBackbone=VCLIP+CLAP, Evaluation Protocol=Zero-shot, Source Dataset=AVE2026.02 | 15.8 |