This dataset is a collection of question-answering (QA) samples focused on various inference tasks, covering multiple domains such as culture, technology, nature, history, and people. Each sample includes structured information such as category, reasoning type, concept descriptions, subject-object relationships, supporting evidence, questions, and answers. It is designed to facilitate the development and evaluation of natural language understanding… See the full description on the dataset page:
https://huggingface.co/datasets/GarnettLiang/Omnibench-RAG.
