AI RESEARCH

Dimensionality-Aware Anomaly Detection in Learned Representations of Self-Supervised Speech Models

arXiv CS.LG

ArXi:2605.02715v1 Announce Type: cross Self-supervised speech models (S3Ms) achieve strong downstream performance, yet their learned representations remain poorly understood under natural and adversarial perturbations. Prior studies rely on representation similarity or global dimensionality, offering limited visibility into local geometric changes.