ChatTS
收藏资源简介:
ChatTS数据集是由清华大学和字节跳动联合创建的,旨在通过合成数据增强时间序列与大型语言模型(LLMs)的对齐。该数据集包含多元时间序列数据和详细的属性描述,通过属性选择器和基于属性的时间序列生成器创建。数据集的创建过程包括生成多样化的文本描述和时间序列数据,以支持复杂的问答和推理任务。ChatTS数据集主要应用于时间序列分析领域,旨在解决时间序列理解和推理的问题,特别是在电力、医疗、交通、天气和金融等多个实际应用场景中。
The ChatTS dataset was jointly created by Tsinghua University and ByteDance, aiming to enhance the alignment between time series and large language models (LLMs) through synthetic data. This dataset contains multivariate time series data and detailed attribute descriptions, which are constructed via an attribute selector and an attribute-based time series generator. The dataset creation process includes generating diverse textual descriptions and time series data to support complex question answering and reasoning tasks. Primarily applied in the field of time series analysis, the ChatTS dataset is designed to address challenges in time series understanding and reasoning, especially across multiple real-world application scenarios such as power systems, healthcare, transportation, weather, and finance.




