Scaling Reinforcement Learning Beyond Math: Researchers from NVIDIA AI and CMU Propose Nemotron-CrossThink for Multi-Domain Reasoning with Verifiable Reward Modeling