# Audio Generation Quality on MusicCaps MusicGen 32kHz (val)

Source: [Audio Generation Quality on MusicCaps MusicGen 32kHz (val)](<https://www.sota2.com/research/sota/audio-generation-quality-on-musiccaps-musicgen-32khz-val>)

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

Data updated: 2026-05-26T13:25:45.736Z

Evaluation of Audio Generation Quality for MusicGen models (32kHz sampling rate) on the validation split of the MusicCaps dataset, focusing on Fréchet Audio Distance 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 | Audio Generation Quality |
| Dataset (including any stated split) | MusicCaps MusicGen 32kHz (val) |
| Metric | FAD (VGGish) |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 4 of 4 |
| 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

- [FAD (VGGish)](<https://www.sota2.com/research/sota/audio-generation-quality-on-musiccaps-musicgen-32khz-val.md?metric=FAD%20(VGGish)>): Lower is better; unit: SCALAR.

- [FAD (CLAP)](<https://www.sota2.com/research/sota/audio-generation-quality-on-musiccaps-musicgen-32khz-val.md?metric=FAD%20(CLAP)>): Lower is better; unit: SCALAR.

## Recorded results

| Method | FAD (VGGish) | Result date | Report paper |
| --- | --- | --- | --- |
| Unwatermarked Reference | 0.247 | Not recorded | 6a15297adbd7150506f7e460 |
| Base | 0.33 | Not recorded | 6a15297adbd7150506f7e460 |
| WMAR | 1.193 | Not recorded | 6a15297adbd7150506f7e460 |
| Simply Robust, Gradient-Free Watermark | 1.256 | Not recorded | 6a15297adbd7150506f7e460 |

### Unwatermarked Reference

FAD (VGGish): 0.247. Result ID: 6a159f45fbc373592c0ea5ea.

[Reporting paper](<https://www.sota2.com/research/paper/6a15297adbd7150506f7e460.md>)

Recorded evaluation configuration: {"h-gram":"None"}. This record does not assert independent reproduction.

### Base

FAD (VGGish): 0.33. Result ID: 6a159f45fbc373592c0ea5eb.

[Reporting paper](<https://www.sota2.com/research/paper/6a15297adbd7150506f7e460.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73131e209131cea5a4b7.md>)

Recorded evaluation configuration: {"h-gram":"0"}. This record does not assert independent reproduction.

### WMAR

FAD (VGGish): 1.193. Result ID: 6a159f45fbc373592c0ea5ec.

[Reporting paper](<https://www.sota2.com/research/paper/6a15297adbd7150506f7e460.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79901e209131cea7e45e.md>)

Recorded evaluation configuration: {"h-gram":"0"}. This record does not assert independent reproduction.

### Simply Robust, Gradient-Free Watermark

FAD (VGGish): 1.256. Result ID: 6a159f45fbc373592c0ea5ed.

[Reporting paper](<https://www.sota2.com/research/paper/6a15297adbd7150506f7e460.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a15297adbd7150506f7e460.md>)

Recorded evaluation configuration: {"h-gram":"0"}. This record does not assert independent reproduction.

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