benchmark-evaluation
Measure how much of a machine-learning benchmark score survives deployment. Grouped bootstrap, frozen thresholds, gap attribution.
