This dataset is part of TokSuite, a comprehensive benchmark designed to measure how different tokenization strategies affect language model performance and robustness. This specific subset contains Farsi (Persian) language multiple-choice text completion questions with various real-world perturbations that test tokenizer robustness.
