Explore high-quality datasets for your AI and machine learning projects.
KnowEdit is a benchmark dataset for knowledge editing of large language models, covering six tasks: ZsRE, Wiki_recent, Wiki_counterfact, WikiBio, ConvSent, and Sanitation. It is built by reorganizing and extending existing datasets (e.g., WikiBio, ZsRE, WikiData_Counterfact) to provide comprehensive evaluation for knowledge editing in LLMs. The dataset structure includes knowledge insertion, modification, and deletion tasks, each with corresponding training and test data. The README also provides example code for loading and using the dataset with the EasyEdit tool, and details differences in loading methods and data structures across the constituent datasets.
