D-CORE: Incentivizing Task Decomposition in Large Reasoning Models for Complex Tool Use Paper • 2602.02160 • Published 3 days ago • 12 • 6