Meta’s Early Experience Trains AI Without Rewards—and Outsmarts Imitation Learning

Most people think training AI agents needs rewards or expert demos. They're overthinking it. Meta...