Harmony in Diversity: Multi-domain Contrastive Policy Optimization for Large Reasoning Models