AI and LLMs · Middle · Question 24 of 75
How do you make an LLM judge trustworthy?
- 1Ask it to give every answer a score out of 100 with no criteria
- 2Use the same model that generated the answers, unchanged
- 3Calibrate it against human labels and use a clear rubric
Show the answer
Correct answer: Calibrate it against human labels and use a clear rubric
Explanation
Measure agreement with human ratings on a sample. Specific rubrics ("mentions refund period, no invented fees") beat vague "rate quality".