image forensics

"团队还提出了“先定位可疑区域，再进行细节审查”的两阶段检测范式，并构建了提供局部区域文本解释的数据集 FakeXplained。在面对可疑图像时，该方法不仅能准确判断其是否由 AI 生成，还能在图像上定位出存在伪造瑕疵或违背物理常识的区域，并同步生成详尽的解释。"

"Unlike prior single-paradigm approaches, which achieve <75% accuracy on out-of-distribution datasets, our method maintains 86.8% average accuracy across seven diverse test sets..."

"The research focuses on "Code-in-the-Loop Forensics" for image forgery detection."