Data for the article "LLMs' Reading Comprehension Is Affected by Parametric Knowledge and Struggles with Hypothetical Statements"
The data in the data folder is available in two formats.
The data in the .json files is structured similarly o the data in SQaAD2.0 and can be used to evaluate the results with the SQuAD evaluation script.
The .txt files are structured with each line containing four tab-separated items: id, context, question, and possible answers. For answerable questions, possible answers are context spans separated by semicolons. For unanswerable questions, possible answers are represented by three dots (...).