vikizhoume/amnesty_qa

Amnesty QA数据集是一个基于Amnesty国际人权报告构建的问答数据集，专门用于评估检索增强生成（RAG）系统。该数据集包含三种语言版本（英语、印地语和马拉雅拉姆语），其中英语版本为推荐使用版本。每个数据样本包含四个关键字段：用户提问（user_input）、参考答案（reference）、系统生成回答（response）和检索到的上下文（retrieved_contexts）。数据集仅提供评估集（eval split），包含20个精心设计的问答对。该数据集采用Apache 2.0许可协议。

The Amnesty QA Dataset is a grounded question-answering dataset created from reports collected from Amnesty International, designed for evaluating RAG (Retrieval-Augmented Generation) systems. It is available in three languages (English, Hindi, and Malayalam), with the English version being the recommended one. Each sample contains four key fields: user_input (the question), reference (ground truth answer), response (generated answer), and retrieved_contexts (relevant context passages). The dataset only provides an evaluation split (eval) containing 20 carefully curated question-answer pairs. The dataset is licensed under Apache 2.0.