CT-GENAI Question 28

Community Discussion

Written by other learners. Comments can disagree with the marked answer — check the official documentation when in doubt.

• Chose B0

  - LLM-as-judge uses a (typically more capable) model — like GPT-4o or Claude Opus — to score outputs of a system under test against a rubric. It scales evaluation beyond what humans can produce, but introduces judge bias and position bias that must be controlled (randomization, multiple judges, calibration). It complements but does not replace human evaluation.