A herbarium’s leaf-ID helper may mis-label at most 4% of a held-out tray of rare sheets, and recall on those rare sheets must stay at or above an agreed floor. What kind of acceptance criteria are those numbers?

At most 4% mis-labels on a held-out tray, and recall on rare sheets at or above an agreed floor, are acceptance criteria for AI functional correctness. Those are thresholds on wrong and missed calls. They are not a homework to recompute precision from a raw matrix, and not a retarget time or a hazard freeze.