Bridging Neural Representations and Symbolic Semantics in Audio Classification.

arXiv:2610.40414v1 Announce Type: cross Abstract: Deep audio classifiers can achieve strong predictive performance, but their hidden representations remain difficult to interpret. Existing audio explainability methods often focus on input-level attribution, su