Adaptive AI-Based Examination Robustness Dataset for Mutation-Based Evaluation

This dataset supports research on robustness and consistency in AI-assisted examination systems. It introduces a mutation-based evaluation framework where semantically equivalent and non-equivalent variations of student answers are used to test grading stability. The dataset includes:- Original student answers- Mutated answer variants (paraphrase, grammar, formatting, adversarial)- AI grading outputs and consistency metrics It is designed to evaluate semantic consistency, grading reliability, and robustness of AI-based scoring systems.