A Multi-Domain Benchmark for Detecting AI-Generated Text-Rich Images from GPT-Image-2

arXiv·medium signal

Introduces a multi-domain benchmark for detecting AI-generated text-rich images produced by GPT-Image-2, where such images often carry privacy-sensitive, transactional, or decision-relevant information. As image generation crosses into convincing document/screenshot synthesis, detection benchmarks like this matter for fraud and provenance defenses. By Yijin Wang, Shuyi Wang, Wenhan Zhang et al. (cs.CV/cs.AI).