LLM judge
an AI model used to assess another model's task or output
An LLM judge is a language model asked to rate something rather than do the task itself: how hard a request is, or whether an answer is good. Routing systems use one to decide when a cheaper model is enough and when to move up. It adds a small extra call and its own chance of error.
Think of it as
A shift supervisor who glances at each job to decide which worker should take it.
Ideas that use this term