Domains — temporal-holdout benchmarks across disciplines

Each domain is a real ORKG research problem. We hold out a later "discovery" paper and give the compiler only strictly-earlier prior literature (the temporal wall). Grouped by discipline so you can see whether grounding, falsifiability, and composite stay stable across fields or whether some expose weaknesses in the schema.