Reinforcement Learning-Based Code Reasoning Optimization for Small Language Models. ijaies 2026, 1 (2).