Speech deepfake detectors achieve low EER on labeled test sets but suffer high half total error rates at transferred thresholds on unlabeled data, and common score calibrations leave EER essentially unchanged.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2026 1verdicts
ACCEPT 1representative citing papers
citing papers explorer
-
When EER Hides Deployment Failure: Auditing Threshold Transfer and Unlabeled Score Calibration for Speech Deepfake Detectors
Speech deepfake detectors achieve low EER on labeled test sets but suffer high half total error rates at transferred thresholds on unlabeled data, and common score calibrations leave EER essentially unchanged.