A standardized, holistic evaluation framework from Stanford University designed to measure the performance and safety of large language models.