Note ArtSeek article.

Note WikiFragments dataset. It contains all the fragments (paragraph + images (optional)) of the English Wikipedia. We save paragraph text, the related page, and the images that appear above the paragraph for a specific row.

Note A subset of WikiFragments, obtained by selecting only the rows related to the pages obtained by navigating recursively the subcategories of the Visual Arts category in Wikipedia. For each row, we report the rendered fragment, and ColQwen2 embeddings (both the embeddings obtained from the model and 9 pooled embeddings obtained with hierarchical clustering).