Benchmark RadarFakeI2V-Bench
FakeI2V-Bench is a benchmark for evaluating image-level and video-level deepfake detectors in video detection scenarios. It comprises 97,548 videos generated by recent generation models and covers multiple content categories, with detectors scored by AUC on detection tasks.
- Released
- 2026-08-04
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
Existing deepfake video benchmarks lack evaluation of image-level detectors' transferability to video, and this benchmark provides a large-scale, standardized protocol to measure detector performance across both detector types, aiding in model selection and development for practical deepfake video detection.
Motivation
Recent advances in video generation models have significantly intensified the deepfake threat, yet the current deepfake video detection benchmarks remain underdeveloped.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.