Abstract: The key challenge in AIGC forgery detection is achieving generalizable and interpretable results. Although Vision-Language Models (VLMs) naturally excel in these aspects, their performance ...