Tag
This paper proposes VaseMuseum, a lightweight multimodal agent framework that combines 3D digitization with vision-language models to create an interactive digital museum for ancient Greek pottery, addressing challenges of evidence grounding and hallucination through source-level and response-level reliability control.