Reproduce “Learning as Deepfakes Evolve: RF-Prompt for Continual Audio Deepfake Detection” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yuankun Xie, Xiaoxuan Guo, Xiaopeng Wang, Siqing Qin, Shaole Li, Kong Aik Lee
RF-Prompt addresses continual deepfake detection by retaining a shared real-speech prompt while adding mechanism-specific fake experts as generation methods arrive. It evaluates this setting with RAMI, which reuses real-speech domains and introduces four fake mechanisms. With frozen XLS-R 300M, RF-Prompt reaches 10.110% average EER and 10.370% pooled EER, below RAWM in single-seed experiments; performance on unseen generators remains open.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.