Benchmark RadarVendorBench-100
Evaluates deepfake image detectors across three paradigms—commercial APIs, vision LLMs, and open-source detectors—on a fixed 100-image adversarial corpus. Uses a unified output schema and scores primarily by Matthews correlation coefficient with ROC-AUC.
- Released
- 2026-07-07
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
Provides a common ground for comparing disparate detector types, addressing the lack of unified evaluation. Identifies metric correlation and calibration issues that matter for real-world deployment decisions.
Motivation
Deepfake image detection is served by three fundamentally different paradigms - commercial APIs, zero-shot vision-language models (LLMs), and open-source detectors - that are rarely evaluated under a common protocol, making direct comparison difficult.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.