Music Generation on Subjective Evaluation Set
4Overlap (Po)Ground Truth
Evaluation Results
| Method | Links | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Ground Truth2024.05 | 4 | 4 | 4.47 | 3.6 | 4.1 | 3.8 | 3.87 | 3.87 | — | — | — | — | — | — | — | |
| QA-MDTtext prefix=high quality, quality token=p-MOS level 52024.05 | 3.27 | 3.77 | 3.69 | 4.19 | 3.54 | 3.94 | 4.23 | 4 | — | — | — | — | — | — | — | |
| MusicGenstatus=re-inferred, checkpoint=official2024.05 | 2.83 | 3.54 | 2.63 | 2.92 | 3.41 | 3 | 4.33 | 3.83 | — | — | — | — | — | — | — | |
| U-Nettext prefix=high quality2024.05 | 2.8 | 3.34 | 3.46 | 4.08 | 3.4 | 3.96 | 3.88 | 3.96 | — | — | — | — | — | — | — | |
| AudioLDM 2status=re-inferred, checkpoint=official2024.05 | 2.03 | 2.42 | 3.03 | 3.61 | 3.21 | 3.71 | 3.85 | 3.85 | — | — | — | — | — | — | — | |
| ACE-Step2026.01 | — | — | — | — | — | — | — | — | 67.42 | 67.16 | 69.07 | 71.4 | 62.04 | 61.6 | 67.94 | |
| DiffRhythm 22026.01 | — | — | — | — | — | — | — | — | 57.99 | 57.13 | 61.63 | 62.58 | 54.77 | 53.15 | 61.13 | |
| HeartMuLa2026.01 | — | — | — | — | — | — | — | — | 69.55 | 71.06 | 73.44 | 73.18 | 66.73 | 65.06 | 70.51 | |
| Suno-v4.52026.01 | — | — | — | — | — | — | — | — | 78.1 | 75.14 | 78.8 | 79.14 | 71.26 | 72.95 | 77.17 | |
| YuE2026.01 | — | — | — | — | — | — | — | — | 56.95 | 54.46 | 60.07 | 60.81 | 55.68 | 56.29 | 61.25 |