TrustRL-Eval: An Interpretable Reinforcement Learning Framework for Predicting and Optimizing Large Language Model Response Quality in Multi-Step Reasoning Tasks. Journal of Data Intelligence and AI Systems, [S. l.], v. 1, n. 1, 2026. Disponível em: https://www.jdataai.org/index.php/home/article/view/44. Acesso em: 18 jul. 2026.