Audio-Visual Question Answering on MusicAVQA Video-centric (source domain)
57.5Average ScoreBase VLM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Base VLM2025.10 | 57.5 | — | |
| VISCOPArchitecture=Universal Encoder (UE)2025.10 | 56.8 | -0.7 | |
| UE + LLM (LoRA)Architecture=Universal Encoder (UE), Tuning=LoRA2025.10 | 52.2 | -5.3 |