Mirage: Unveiling Hidden Artifacts in Synthetic Images with Large Vision-Language Models
This addresses the challenge of reliable AI image detection for security and verification, though it is incremental as it focuses on a specific artifact-based scenario.
The paper tackles the problem of AI-generated images evading standard detectors by introducing Mirage, a dataset of synthetic images with visible artifacts where current methods fail, and finds that Large Vision-Language Models are effective at detecting such images but decline in performance without artifacts.
Recent advances in image generation models have led to models that produce synthetic images that are increasingly difficult for standard AI detectors to identify, even though they often remain distinguishable by humans. To identify this discrepancy, we introduce \textbf{Mirage}, a curated dataset comprising a diverse range of AI-generated images exhibiting visible artifacts, where current state-of-the-art detection methods largely fail. Furthermore, we investigate whether Large Vision-Language Models (LVLMs), which are increasingly employed as substitutes for human judgment in various tasks, can be leveraged for explainable AI image detection. Our experiments on both Mirage and existing benchmark datasets demonstrate that while LVLMs are highly effective at detecting AI-generated images with visible artifacts, their performance declines when confronted with images lacking such cues.