Search papers, labs, and topics across Lattice.
This study formalizes the prediction of taste from audio as a benchmark for music information retrieval, leveraging ten frozen audio encoders from the HEAR families and employing a multi-task regression head with gated late-fusion. The strongest models achieved a macro RMSE of 0.134, outperforming the previous state of the art and demonstrating that the model aligns more closely with group consensus than average human raters. Additionally, the research reveals that while the absolute error across encoders is statistically similar, gated late-fusion excels in rank correlation, indicating its potential for more nuanced retrieval tasks.
Predicting taste from audio embeddings not only surpasses human consensus but also redefines the benchmarks for music retrieval systems.
Crossmodal correspondences between sound and taste are well established in psychology and neuroscience, but largely absent from content-based multimedia retrieval. We formalise taste-from-audio prediction as a content-based music information retrieval benchmark over a perceptually validated multi-source corpus, comparing ten frozen audio encoders from the four HEAR families under a shared multi-task regression head, with gated late-fusion as a configurable variant. In order to assess the effectiveness of the models, we compute absolute error and rank correlation. The strongest systems predict the five tastes within a macro RMSE of 0.134; on held-out real music their error is less than half a single rater's deviation from the consensus (RMSE 0.13 vs. 0.28), so the model tracks the group consensus more closely than an average human rater, and well below the previous state of the art baseline (0.219). On absolute error the encoders are statistically flat, with a single VGGish matching the best fusion, but gated late-fusion's advantage is confined to rank correlation (macro Pearson r 0.724 vs. 0.666). Operationalised as a content-based retrieval index, the predicted taste space ranks a 309-item pool far more faithfully than a CLAP-text baseline, which sits at chance; ridge probes and an audio-bandstop knockout read the strongest representations against documented sound-taste correspondences.
