VaseMuseum is a lightweight, modular multimodal-agent framework for digital museums focused on ancient Greek pottery. Its VaseAgent handles both 2D images and 3D artifacts, combining 3D-aware reasoning with retrieval from authoritative web and museum sources. Source-level controls select diverse, verifiable evidence, while response-level checks compare generated claims against the evidence pool and encourage neutral answers when support is weak or conflicting. A training-free GRPO-style selection mechanism favors responses with valid references and calibrated confidence without modifying the underlying VLM. The abstract reports improved citation validity, fewer hallucinations, and more neutral responses than search-enabled VLM baselines.
No heat snapshots are available in the last 24 hours.