Large language models built by Google, OpenAI, and academic research teams have matched or exceeded human-expert scores on ...