Salesforce/APIGen-MT-5k

APIGen-MT-5k数据集是一个由APIGen-MT生成的5000个多轮对话的集合，用于模拟代理人与人类的交互。这个数据集是针对公共数据集中高质量多轮代理人交互数据的缺乏和高成本手动收集这类数据的问题而设计的。每个任务都经过三个层次的验证，确保其可靠性和正确性。数据集涵盖零售和航空领域，采用ShareGPT-like JSON格式。

The APIGen-MT-5k dataset is a collection of 5000 multi-turn dialogues generated by APIGen-MT, designed to simulate interactions between an AI agent and humans. This dataset addresses the lack of high-quality multi-turn agent interaction data in public datasets and the high cost of manually collecting such data. Each task in the dataset is verified through three hierarchical stages to ensure its reliability and correctness. The dataset covers the Retail and Airline domains and is formatted in a ShareGPT-like JSON structure.