Idanwo ti o sopọ mọ itọsọna · Lile Ipele

Calibrating LLM Judges Against Humans Quiz

Calibrate LLM judges with shared rubrics, independent human labels, confusion matrices, suitable agreement statistics, and repeated held-out checks.

Jẹmọ itọsọna awọn ọnaCalibrating LLM Judges Against Humans
Ibeere 1 ti 8

Why is Cohen's kappa preferred over raw percent agreement when calibrating an LLM judge?