Benchmark RadarDailyBench
DailyBench is a unified benchmark for evaluating AI-generated image detectors on modern full-image synthesis and object-level manipulation. It comprises FakeBench, with images from recent open-source and commercial generative models, and ManipulationBench, with subtle edits to real images. Detectors are scored by balanced accuracy.
- Released
- 2026-07-27
- Readiness
- Inspectable
- Primary field
- Robotics & Autonomous Systems
Why it matters
Existing detection benchmarks lag behind current generative models, causing a gap between evaluation and real-world scenarios. DailyBench provides a realistic testbed to assess generalization to modern synthesis and manipulation, offering practical value for developing robust detectors.
Motivation
Recent advances in generative models have shifted AI-generated image detection from identifying easily distinguishable, fully synthetic images to identifying highly realistic content generated by both modern generation and manipulation pipelines.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.