Dataset · RUC-NLPIR
36 RAG benchmarks in the FlashRAG suite
RUC-NLPIR/FlashRAG_datasets
Preprocessed datasets for RAG research: 36 benchmarks and FlashRAG resources.
What's inside
Part of the FlashRAG toolkit for reproducing and developing RAG research. It includes 36 preprocessed RAG benchmark datasets and 16 modern RAG algorithms. Text data in English, size ranging from 1 to 10 million; tasks include question-answering and summarization.
What you can do with it
- Reproduce SOTA RAG pipelines
- Train or evaluate a retriever/generator for RAG
- Compare RAG algorithms across 36 benchmarks
- Tasks
- Question answering, Summarization
- Languages
- English
- Size
- 1M–10M rows
- Downloads this month
- 13.5K
- Likes
- 96
Models trained on it
- mradermacher/eubiota-planner-8b-i1-GGUFReinforcement learning 2.2K
- mradermacher/ReSearch-Qwen-7B-Instruct-i1-GGUF 1.2K
- mradermacher/ReSearch-Qwen-7B-i1-GGUF 1.2K
- mradermacher/ReSearch-Qwen-32B-i1-GGUF 962
- Manojb/Qwen-7B-toolcalling-ReSearch-gguf-Q8_0-codexText generation 734
- mradermacher/eubiota-planner-8b-GGUFReinforcement learning 380
- mradermacher/ReSearch-Qwen-32B-Instruct-i1-GGUF 354
- mradermacher/A2Search-3B-Instruct-i1-GGUFReinforcement learning 338
How to load it
from datasets import load_dataset
ds = load_dataset("RUC-NLPIR/FlashRAG_datasets")