single mention · data curation
Synthetic long-context data via permutation and concatenation of multimodal documents
Permutes and concatenates multimodal documents and subtasks to create long contexts requiring cross-context integration.
The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See synthetic data for the methods filed beside it.
Filed under data curation › synthetic data