neural-bridge/rag-dataset-12000

Dataset Overview

Dataset Name

Retrieval-Augmented Generation (RAG) Dataset 12000

Dataset Description

• Purpose: To build optimized RAG models that enhance large language models (LLMs) by accessing external authoritative knowledge bases for response generation.
• Characteristics: Extends models to specific domains or internal organizational data without retraining, improving relevance, accuracy, and context specificity of outputs.

Dataset Structure

• Features:
  • context: string, a sequence of tokens.
  • question: string, a question related to the context.
  • answer: string, the answer to the question.
• Data Splits:
  • train: 9,600 samples.
  • test: 2,400 samples.

Language

• Language: English (en)

License

• License: Apache License 2.0

Data Source

• Source: Falcon RefinedWeb

Data Example

json { "context": "...", "question": "...", "answer": "..." }