Concept-1K

Concept-1K数据集由华南理工大学计算机科学与工程学院创建，包含1023个来自六个领域（经济、文化、科技、环境、教育和健康医疗）的最新概念。数据集通过使用GPT4生成每个概念的20个三元组，每个三元组转化为一对训练和测试实例，总计16653对。该数据集旨在支持实例增量学习（IIL）场景，帮助研究PLMs在增量学习中的遗忘问题，并推动更有效的增量学习技术的发展。

The Concept-1K Dataset was created by the School of Computer Science and Engineering, South China University of Technology. It contains 1,023 up-to-date concepts across six domains: economy, culture, technology, environment, education, and healthcare. The dataset generates 20 triples for each concept via GPT-4, and each triple is converted into one training instance and one test instance, resulting in a total of 16,653 instance pairs. This dataset is designed to support the Instance Incremental Learning (IIL) scenario, assist researchers in exploring the forgetting problem of Pre-trained Language Models (PLMs) during incremental learning, and promote the advancement of more effective incremental learning techniques.