MOZAIC: An Audio-Video Deepfake Benchmark for Telling What Is Fake
Shaina Raza, Mohamed Hafez, Ahmed Radwan, Khalid Elgazzar, Kathryn Hume
Vector Institute
137,745 clips of 1,000 people. Each clip is labelled by which part is fake: face, voice, both, or neither. We test 12 multimodal LLMs and 5 trained detectors.




