Doubly-Robust LLM-as-a-Judge: Externally Valid Estimation with Imperfect Personas

Open in new window