Evaluating Answer Leakage Robustness of LLM Tutors against Adversarial Student Attacks | ArxivCSExplorer