SignalTech/AI
VeriSpoof: Beyond Real or Fake with One Audio LLM
High-priority alert based on source significance and keyword relevance.
Context
What happened: Researchers released VeriSpoof, an open audio language model that unifies deepfake detection, generator attribution, recording-condition analysis, synthesized-span localisation, and source verification with shared weights and checkable rationales.
Key numbers: VeriSpoof-Plus reaches 2.14% pooled EER across 14 Speech-DF-Arena
Why it matters: The authors report first place among open-weight systems on Speech-DF-Arena and SpoofRadar v1.0, and state that editing the stated synthesized span flips the model's decision, tying each verdict to evidence checkable against the audio.
What changed
Researchers released VeriSpoof, an open audio language model that unifies deepfake detection, generator attribution, recording-condition analysis, synthesized-span localisation, and source verification with shared weights and checkable rationales.
Why it matters
The authors report first place among open-weight systems on Speech-DF-Arena and SpoofRadar v1.0, and state that editing the stated synthesized span flips the model's decision, tying each verdict to evidence checkable against the audio.