TEMPRAGEVAL
收藏资源简介:
TEMPRAGEVAL是一个用于时间敏感问答系统的基准数据集,由南洋理工大学等机构创建。该数据集通过重新利用TIMEQA和SITUATEDQA数据集,并引入时间扰动和人工标注的黄金证据,旨在评估检索系统在时间推理任务中的表现。数据集包含1000个测试样本,涵盖常见和罕见实体,适用于评估检索增强生成系统的时间推理能力。TEMPRAGEVAL的创建过程包括对原始问题进行时间扰动和证据标注,旨在解决时间敏感问答中的检索和生成问题。
TEMPRAGEVAL is a benchmark dataset for time-sensitive question answering systems, developed by institutions including Nanyang Technological University and others. This dataset repurposes the TIMEQA and SITUATEDQA datasets, introduces temporal perturbations and manually annotated gold standard evidence, and is designed to evaluate the performance of retrieval systems in temporal reasoning tasks. It contains 1,000 test samples covering both common and rare entities, making it suitable for assessing the temporal reasoning capabilities of retrieval-augmented generation (RAG) systems. The creation process of TEMPRAGEVAL involves applying temporal perturbation and evidence annotation to the original questions, aiming to address the retrieval and generation challenges in time-sensitive question answering.




