MrOvkill/fallacies-fallacy-base

该数据集旨在提高大型语言模型（LLMs）对逻辑谬误的检测和处理能力。数据来源于维基百科的谬误列表，使用PDF格式的子页面作为种子数据，并通过Google Cloud UI的Gemini 1.5 Flash、Experimental和Pro模型合成生成每一行数据。数据集包含id、group_id、name、example、explanation和response等字段，用于描述谬误的名称、示例、解释以及机器在遇到谬误时应给出的合理响应。

The Fallacies dataset was produced to enhance the detection and handling of logical and other fallacies in large language models (LLMs). The seed data for this dataset was taken from Wikipedias list of fallacies, synthetically generating each row using Gemini 1.5 Flash, Experimental, and Pro on the Vertex AI Google Cloud UI. The dataset includes the ID, group ID, name, example, explanation, and the most reasonable response a machine should give when presented with the fallacy in an argument. The dataset size ranges from 1K to 10K, and the language is English.