LEAD: Length-Efficient Adaptive and Dynamic Reasoning for Large Language Models Paper • 2605.09806 • Published 6 days ago • 5