TrustRL-Eval: An Interpretable Reinforcement Learning Framework for Predicting and Optimizing Large Language Model Response Quality in Multi-Step Reasoning Tasks. JDIAS. 2026;1(1). Accessed July 18, 2026. https://www.jdataai.org/index.php/home/article/view/44