SVR-R1: Bootstrapping Multi-modal Reasoning with Self-verification in Reinforcement Learning Written on July 13, 2026