CARE-RL: Capability-Aware Reinforcement Learning for Mitigating Cross-Domain Conflicts | ArxivCSExplorer