actual local dataset validation; no retrieval, generation or quality benchmark
Questions: 8
Sources: 4
Splits: {"development": 4, "held_out": 4}
Wrong documentation version: TargetError
Stale target hash: TargetError
Answerable question without a passage: TargetError
Edited held-out question: SplitBoundaryError
Lock from another dataset: SplitBoundaryError
Python 3.13 facts documented only; no 3.13 execution. Semantic relevance is a reviewed label, not a validated model answer.
