The method is useful when you want to check quality at scale. On the other hand, it's blunt without calibration against human raters and without strict scoring criteria. A common mistake is to assume the judging model is automatically objective.
LLM-as-judge
A testing method (eval) where one language model (LLM) judges the answers from another.
Want to know what it means for you?
We go through your systems and show where AI helps and how it is governed.