# Music Understanding on X-ARES

Source: [Music Understanding on X-ARES](<https://www.sota2.com/research/sota/music-understanding-on-x-ares>)

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

Data updated: 2026-03-27T14:01:05.556Z

Evaluation of model performance specifically on music-related tasks within the X-ARES benchmark suite, covering components like FMA, GTZAN, MAESTRO, and NSynth.

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 | Music Understanding |
| Dataset (including any stated split) | X-ARES |
| Metric | FMA Score |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 19 of 19 |
| 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

- [FMA Score](<https://www.sota2.com/research/sota/music-understanding-on-x-ares.md?metric=FMA%20Score>): Higher is better; unit: PERCENT.

- [GTZAN Score](<https://www.sota2.com/research/sota/music-understanding-on-x-ares.md?metric=GTZAN%20Score>): Higher is better; unit: PERCENT.

- [MAESTRO Score](<https://www.sota2.com/research/sota/music-understanding-on-x-ares.md?metric=MAESTRO%20Score>): Higher is better; unit: PERCENT.

- [NSynth Score](<https://www.sota2.com/research/sota/music-understanding-on-x-ares.md?metric=NSynth%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | FMA Score | Result date | Report paper |
| --- | --- | --- | --- |
| DashengTokenizer | 66.29 | Not recorded | 69a5195d2d603e6de3e3c73d |
| DashengTokenizer | 64.34 | Not recorded | 69a5195d2d603e6de3e3c73d |
| SPEAR | 64.16 | Not recorded | 69a5195d2d603e6de3e3c73d |
| DashengTokenizer | 64 | Not recorded | 69a5195d2d603e6de3e3c73d |
| MSM-MAE | 62.74 | Not recorded | 69a5195d2d603e6de3e3c73d |
| OpenBeats | 62.4 | Not recorded | 69a5195d2d603e6de3e3c73d |
| SPEAR | 62.06 | Not recorded | 69a5195d2d603e6de3e3c73d |
| Whisper-Large-V3 | 58.9 | Not recorded | 69a5195d2d603e6de3e3c73d |
| SemantiCodec | 57.94 | Not recorded | 69a5195d2d603e6de3e3c73d |
| MingTok-Audio | 54.45 | Not recorded | 69a5195d2d603e6de3e3c73d |
| HuBERT | 42.81 | Not recorded | 69a5195d2d603e6de3e3c73d |
| WavLM | 36.11 | Not recorded | 69a5195d2d603e6de3e3c73d |
| DAC | 35.43 | Not recorded | 69a5195d2d603e6de3e3c73d |
| MingTok-Audio | 34.74 | Not recorded | 69a5195d2d603e6de3e3c73d |
| EzAudio | 30.74 | Not recorded | 69a5195d2d603e6de3e3c73d |
| UniFlow-Audio | 30.17 | Not recorded | 69a5195d2d603e6de3e3c73d |
| Encodec | 28.8 | Not recorded | 69a5195d2d603e6de3e3c73d |
| FlexiCodec | 0 | Not recorded | 69a5195d2d603e6de3e3c73d |
| WavTokenizer | 0 | Not recorded | 69a5195d2d603e6de3e3c73d |

### DashengTokenizer

FMA Score: 66.29. Result ID: 69c68d8a47f34ce924b0cbfb.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

Recorded evaluation configuration: {"Model Category":"Continuous Tokenizer"}. This record does not assert independent reproduction.

### DashengTokenizer

FMA Score: 64.34. Result ID: 69c68d8a47f34ce924b0cbf6.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

Recorded evaluation configuration: {"Model Category":"Encoder Model","Scale":"0.6B"}. This record does not assert independent reproduction.

### SPEAR

FMA Score: 64.16. Result ID: 69c68d8a47f34ce924b0cbf4.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/690442703ef7ccd74e5e4ac8.md>)

Recorded evaluation configuration: {"Model Category":"Encoder Model","Scale":"XLarge"}. This record does not assert independent reproduction.

### DashengTokenizer

FMA Score: 64. Result ID: 69c68d8a47f34ce924b0cbf5.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

Recorded evaluation configuration: {"Model Category":"Encoder Model","Scale":"Base"}. This record does not assert independent reproduction.

### MSM-MAE

FMA Score: 62.74. Result ID: 69c68d8a47f34ce924b0cbf1.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6eb81e209131cea41c4f.md>)

Recorded evaluation configuration: {"Model Category":"Encoder Model"}. This record does not assert independent reproduction.

### OpenBeats

FMA Score: 62.4. Result ID: 69c68d8a47f34ce924b0cbee.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a1a1e209131cea812f1.md>)

Recorded evaluation configuration: {"Model Category":"Encoder Model"}. This record does not assert independent reproduction.

### SPEAR

FMA Score: 62.06. Result ID: 69c68d8a47f34ce924b0cbf3.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/690442703ef7ccd74e5e4ac8.md>)

Recorded evaluation configuration: {"Model Category":"Encoder Model","Scale":"Base"}. This record does not assert independent reproduction.

### Whisper-Large-V3

FMA Score: 58.9. Result ID: 69c68d8a47f34ce924b0cbef.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e2c1e209131cea3e51b.md>)

[Implementation](<https://github.com/openai/whisper>)

Recorded evaluation configuration: {"Model Category":"Encoder Model"}. This record does not assert independent reproduction.

### SemantiCodec

FMA Score: 57.94. Result ID: 69c68d8a47f34ce924b0cbeb.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75ec1e209131cea6aa24.md>)

Recorded evaluation configuration: {"Model Category":"Codec"}. This record does not assert independent reproduction.

### MingTok-Audio

FMA Score: 54.45. Result ID: 69c68d8a47f34ce924b0cbfa.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6912c3042828b7cc6937dba6.md>)

Recorded evaluation configuration: {"Model Category":"Continuous Tokenizer","Variant":"Unified"}. This record does not assert independent reproduction.

### HuBERT

FMA Score: 42.81. Result ID: 69c68d8a47f34ce924b0cbf0.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a2a1e209131cea26880.md>)

[Implementation](<https://github.com/pytorch/fairseq/tree/master/examples/hubert>)

Recorded evaluation configuration: {"Model Category":"Encoder Model"}. This record does not assert independent reproduction.

### WavLM

FMA Score: 36.11. Result ID: 69c68d8a47f34ce924b0cbf2.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dfa1e209131cea3d210.md>)

[Implementation](<https://aka.ms/wavlm>)

Recorded evaluation configuration: {"Model Category":"Encoder Model"}. This record does not assert independent reproduction.

### DAC

FMA Score: 35.43. Result ID: 69c68d8a47f34ce924b0cbe9.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71091e209131cea4ecde.md>)

[Implementation](<https://github.com/descriptinc/descript-audio-codec>)

Recorded evaluation configuration: {"Model Category":"Codec"}. This record does not assert independent reproduction.

### MingTok-Audio

FMA Score: 34.74. Result ID: 69c68d8a47f34ce924b0cbf9.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6912c3042828b7cc6937dba6.md>)

Recorded evaluation configuration: {"Model Category":"Continuous Tokenizer","Variant":"Acoustic"}. This record does not assert independent reproduction.

### EzAudio

FMA Score: 30.74. Result ID: 69c68d8a47f34ce924b0cbf7.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79821e209131cea7de21.md>)

Recorded evaluation configuration: {"Model Category":"Continuous Tokenizer"}. This record does not assert independent reproduction.

### UniFlow-Audio

FMA Score: 30.17. Result ID: 69c68d8a47f34ce924b0cbf8.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bb71e209131cea8a1bf.md>)

Recorded evaluation configuration: {"Model Category":"Continuous Tokenizer"}. This record does not assert independent reproduction.

### Encodec

FMA Score: 28.8. Result ID: 69c68d8a47f34ce924b0cbea.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6db91e209131cea3b969.md>)

Recorded evaluation configuration: {"Model Category":"Codec"}. This record does not assert independent reproduction.

### FlexiCodec

FMA Score: 0. Result ID: 69c68d8a47f34ce924b0cbec.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bd21e209131cea8ab31.md>)

Recorded evaluation configuration: {"Model Category":"Codec"}. This record does not assert independent reproduction.

### WavTokenizer

FMA Score: 0. Result ID: 69c68d8a47f34ce924b0cbed.

[Reporting paper](<https://www.sota2.com/research/paper/69a5195d2d603e6de3e3c73d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa771b1e209131cea7109d.md>)

Recorded evaluation configuration: {"Model Category":"Codec"}. This record does not assert independent reproduction.

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