Dataset Card for GSM8K-PR

Dataset Summary

The GSM8K-PR dataset is a modified version of the well-known GSM8K dataset, designed to enhance the evaluation of large language models (LLMs) in long-context settings. This dataset retains the original structure of GSM8K, but with one important modification: all personal pronouns (such as “he,” “she,” “it,” “they,” etc.) have been replaced with the specific person, object, or thing they refer to. The goal of this modification is to make the dataset more suitable for tasks that involve long-form text processing and reasoning.

Potential Errors

Although we have put considerable effort into curating and modifying the GSM8K-PR dataset, there may still be some errors that have not yet been identified. If you come across any issues, inconsistencies, or errors within the dataset, please do not hesitate to contact us.