Chart every score across experiment runs, read summary statistics against your baseline, and compare candidates item by item filtered by score thresholds.