“A quantitative measure comparing the agreement between human-labeled evaluations and LLM-as-a-judge evaluations.”