Sharpness, a measure of loss sensitivity to weight perturbations, correlates with speech deepfake detection error on unseen data, and Sharpness-Aware Minimization reduces both sharpness and error in most settings.
Our findings show that sharpness increases under domain shifts, indicating model sensitivity to unseen conditions
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
From Sharpness to Better Generalization for Speech Deepfake Detection
Sharpness, a measure of loss sensitivity to weight perturbations, correlates with speech deepfake detection error on unseen data, and Sharpness-Aware Minimization reduces both sharpness and error in most settings.