Researchers are using NPR Sunday Puzzle questions to benchmark AI reasoning models, showcasing new methods to evaluate machine problem-solving skills against human cognition challenges.