LLM as a Judge

An AI evaluation paradigm in which a large language model acts as an evaluator to assess the output quality of another model. Popularized through research by institutions including Stanford, the approach replaces hard-coded rules with natural language understanding capabilities to enable flexible, scalable evaluation.