Three tasks in the release carry an empty gold_context. coverage_precision returns cov = 1.0 when gold_size == 0, so every system scores perfect coverage on those tasks regardless of what it retrieves.
Worth either removing them or documenting them as excluded from aggregates.
Three tasks in the release carry an empty
gold_context.coverage_precisionreturnscov = 1.0whengold_size == 0, so every system scores perfect coverage on those tasks regardless of what it retrieves.Worth either removing them or documenting them as excluded from aggregates.