# Video Question Answering on MUSIC-AVQA

Source: [Video Question Answering on MUSIC-AVQA](<https://www.sota2.com/research/sota/video-question-answering-on-music-avqa>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T05:29:07.260Z

Evaluation of model performance on the MUSIC-AVQA dataset for Video Question Answering tasks, reporting accuracy metrics.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Video Question Answering |
| Dataset (including any stated split) | MUSIC-AVQA |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 2 of 2 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Accuracy](<https://www.sota2.com/research/sota/video-question-answering-on-music-avqa.md?metric=Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| VAST(1.3B) | 80.7 | Not recorded | 68fa70c91e209131cea4d159 |
| VALOR-L | 78.9 | Not recorded | 68fa70c91e209131cea4d159 |

### VAST(1.3B)

Accuracy: 80.7. Result ID: 699141233a6c31484906ceaf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70c91e209131cea4d159.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70c91e209131cea4d159.md>)

[Implementation](<https://github.com/TXH-mercury/VAST>)

Recorded evaluation configuration: {"Sample":"442M","extra\_modalities":"audio or subtitle"}. This record does not assert independent reproduction.

### VALOR-L

Accuracy: 78.9. Result ID: 699141233a6c31484906ceae.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70c91e209131cea4d159.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa767b1e209131cea6d4d7.md>)

[Implementation](<https://casia-iva-group.github.io/projects/VALOR>)

Recorded evaluation configuration: {"Sample":"433.5M","extra\_modalities":"audio or subtitle"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
