R1-Zero's "Aha Moment" in Visual Reasoning on a 2B Non-SFT Model

Zhou, Hengguang, Li, Xirui, Wang, Ruochen · arXiv · 2025