Reproduce “SE-ADD: Self-Evolving Audio Deepfake Detection with Mistake-Driven Supervision” 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
Rong Wan, Wei Xie, Jiaxi Li, Wenwu Wang, Lu Yin, Yiliao Song, Xilu Wang
SE-ADD adapts audio language models to changing deepfake attacks by using each model’s cue-free mistakes to select additional training tasks built from its generated forensic cues. It retains ordinary authentic/spoof labels and updates a LoRA adapter across attack mixtures. On held-out attacks, EER fell from 36.72% to 7.52% for Qwen2-Audio and 19.93% to 3.97% for MOSS-Audio, but this benchmark does not establish universal robustness.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.