ConceptVectors

该数据集是一个基准数据集，包含了数百个常见概念及其在两个开源大型语言模型中的参数化知识轨迹。该数据集被用于分析不同遗忘方法对大型语言模型中概念向量影响的研究。具体任务是通过验证信息是否从模型参数中被移除，来评估遗忘性能。

This dataset is a benchmark dataset that includes hundreds of common concepts and their parameterized knowledge trajectories across two open-source large language models (LLMs). It is utilized in research analyzing the impact of different forgetting methods on concept vectors within large language models. The specific task is to evaluate forgetting performance by verifying whether information has been removed from the model parameters.