Table 9.

Results of a music emotion recognition task on the DEAM dataset. ${}^{\ast }$ indicates that the model evaluates on different subsets of the dataset than our work and hence numbers are not directly comparable.

Model	$R_V^2$	$R_A^2$
MFCCs	.122	.327
CLMR26	.107	.384
M2BERT, no pre-training	.261	.515
M3BERTLarge	.266	.537
Hand-crafted Features70${}^{\ast }$	.278	.529
M3BERTSmall	.332	.521
M2BERT	.345	.562
VGGish	.395	.582

Highest values per metric are given in bold.