Preprint
Aug 2026
Steering dense music retrieval with open-vocabulary concept discovery
A lightweight, training-free method that recovers a sparse set of audio features whose decoded representation reconstructs the target concept while remaining consistent with audio-space geometry, enabling more precise concept amplification and suppression with reduced drift on preservation metrics.
Julien Guinot, Alain Riou, Elio Quinton et al.
· 0 citations