ArtSeek: Deep artwork understanding via multimodal in-context reasoning and late interaction retrieval
ArtSeek is a multimodal framework for deep artwork understanding that combines large language models with retrieval-augmented generation. It uses a knowledge-grounded dataset called WikiFragments and novel query encoding to allow the system to analyze artworks without external links. The framework achieves high performance in retrieving facts, classifying art attributes, and reasoning about complex visual questions.