AI engineering glossary · Evals guardrails AI security
What is judge calibration?
Judge calibration: The process of checking that an automated grader's scores agree with trusted human judgements and adjusting the judge prompt, rubric or model until they do. Without it, a judge may look confident yet drift from what users actually value, so teams measure agreement on a human-labelled sample.
Judge calibration: an example
Humans and a judge model grade the same 100 answers; agreement is 72 percent, so the team adds examples to the judge prompt and reaches 88 percent.
Related terms
More Evals guardrails AI security terms
- Least privilege
- Llama Guard
- LLM-as-a-judge
- lm-evaluation-harness
- Many-shot jailbreaking
- Membership inference attack
- Misinformation risk
- MITRE ATLAS
Where you learn it
Judge calibration is part of the AI Engineering Program, which teaches it with notes, a hands-on lab and an exam. The first lesson is free.