Reinforcement Learning-Based Reasoning Optimization for Autonomous LLM Agents. JAAIR 2026, 5 (1).