Redfern, C. Improving Reasoning Stability in Large Language Models via Iterative Self-Questioning and Semantic Calibration. JCSSA 2026, 6.