Preprint Reports Near-Perfect Speech-Deepfake Source Tracing on a Closed Test
A preprint reports near-perfect speech-deepfake source classification on a closed 17-generator benchmark, with feature-importance rankings that broadly agree with SH
Developing Light