New research from the University of Michigan sheds light on creating more trustworthy AI leaderboards. Led by professors Lingjia Tang and Jason Mars, the study examines ranking methods used in popular platforms and competitive gaming. By evaluating how these systems handle crowdsourced data, the researchers identified key factors that can accurately—or inaccurately—reflect an AI model’s true capabilities. Their findings offer guidelines to help the industry build more reliable and transparent AI performance leaderboards.
