Reproduce “VeriSpoof: Beyond Real or Fake with One Audio LLM” 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
Yang Xiao, Ting Dang
VeriSpoof answers speech-forensics questions with one audio language model, scoring candidate answers and producing rationales checked against waveform measurements, labels, recording logs, or generator metadata. VeriSpoof-Plus reaches 2.14% pooled EER across 14 Speech-DF-Arena test sets. On 500 PartialEdit recordings, reasoning raises accuracy from 78.4% to 99.6%; tested edits do not establish universal faithfulness.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.