[D] How to use RAG benchmarks in practice

I am working on a research projects which involves experimenting with RAGs. I want to run the models first to get an understanding of how the whole pipeline works. I found some datasets on HuggingFace (such as https://huggingface.co/datasets/explodinggradients/WikiEval).

My understanding of RAGs is that I should be given a datastore, and then I perform the task of question answering using the datastore. However, in these datasets, context is given along with the question, and I do not quite understand that. …

context datasets however machinelearning question question answering rag rags understanding