Datasets:
| # papers | |
| Public BioScope papers, following BigBio. | |
| Native XML text, nested cue/scope offsets; clinical masked words excluded. | |
| Every sentence is kept; sentences without a cue answer both cue fields with null. | |
| Train-only: the collection has 9 documents, too few to split representatively. | |