DATASET: Kazakh QA pairs for retrieval-augmented generation with inline citation markers
What it is (scope & use)
A Kazakh-language QA dataset built from Wikipedia-style texts to train/evaluate RAG systems that must cite evidence. Each answer is written to include inline source pointers like doc:doc_0, so generators can surface provenance and evaluators can check grounding. This aligns with RAG’s emphasis on external evidence for knowledge-intensive QA and with… See the full description on the dataset page: https://huggingface.co/datasets/nur-dev/rag-qa-with-citations.
