Tag: evaluation

LLM-as-a-judge uses a language model to evaluate an agent result against a rubric, with evidence, uncertainty, and provenance kept explicit.

An agent eval is a repeatable test that combines an input, expected behavior, execution evidence, and a grader.