A model-graded judge returns scores clustered around 6 out of 10 for almost every answer. What change most directly addresses this?
Choose an answer to check it. Afterward, select any option to review its explanation.
A wider scale does not change the judge's tendency toward the middle.
Without structured reasoning first, judges drift toward a safe middle score; calibration checks agreement with humans.
More cases do not fix a judge that cannot discriminate.
Exact match cannot grade open-ended quality.
id q4f58ea1e7b