Humanitys Last Exam: The Super-Benchmark AI Is Currently Failing

AI has outgrown its old tests. Scientists introduce "Humanitys Last Exam"-a challenge so difficult that the world's best AI models are scoring in the single digits.