Oolong-real is a dataset from the paper Oolong: Evaluating Long Context Reasoning and Aggregation Capabilities. See the paper for more details on the dataset construction.
To run the standard evaluation setting you will need the dnd config.
input: context_window_text + "\n" + question (these are separated because the context window text can be cached for reuse across multiple input queries)
output: answer
