RepLiQA is an evaluation dataset that contains Context-Question-Answer triplets, where contexts are non-factual but natural-looking documents about made up entities such as people or places that do not exist in reality. RepLiQA is human-created, and designed to test for the ability of Large Language Models (LLMs) to find and use contextual information in provided… See the full description on the dataset page:
https://huggingface.co/datasets/ServiceNow/repliqa.
