System-level Audio Aesthetics Prediction on AMC 2025 (eval)
0.463PQ MSEProposed (VERSA)
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Proposed (VERSA)components=VERSA-based predictor2025.12 | 0.463 | 0.893 | 0.164 | 0.946 | 1.596 | 0.911 | 1.099 | 0.908 | |
| B02 (Baseline)2025.12 | 0.632 | 0.866 | 0.226 | 0.934 | 1.142 | 0.841 | 1.478 | 0.81 | |
| T12 (AESCA)components=Ensemble of KAN #1–#4 & VERSA2025.12 | 0.682 | 0.916 | 0.401 | 0.938 | 1.853 | 0.946 | 1.237 | 0.924 | |
| Proposed (KAN #1)components=Single KAN-based predictor2025.12 | 0.774 | 0.911 | 0.407 | 0.936 | 2.037 | 0.939 | 1.331 | 0.935 | |
| Proposed (KAN #1–#4)components=Ensemble of four KAN-based predictors2025.12 | 0.783 | 0.915 | 0.418 | 0.938 | 2.031 | 0.941 | 1.351 | 0.939 |