A preprint reports that a calibrated pairwise auditor scored above a baseline, while concern-focused training showed lower false-positive calibration.