BytedTsinghua-SIA/hotpotqa

该数据集是为了训练和评估MemAgent模型而提供的关键数据集，MemAgent是一个为长文本上下文设计的LLM框架。数据集主要来源于HotpotQA数据集，并包括了为长文本多跳问答特别设计的合成数据。它包含了训练与开发数据集、主要任务评估数据以及分布外任务数据。

This dataset is essential for training and evaluating the MemAgent model, a framework designed for long-context LLMs. It is primarily derived from the HotpotQA dataset and includes synthetic data crafted for long-context multi-hop question-answering. The dataset contains training and development splits, main task evaluation data, and out-of-distribution task data.