MALLS

该数据集包含了34,000个高质量的、多样化的句子级别的自然语言与一阶逻辑配对，这些配对是从GPT-4中收集而来的。在创建此数据集时，我们采用了一个提示管道，通过调整提示以确保自然语言与一阶逻辑配对的多样性和有效性。该数据集的规模为34,000对，其任务是进行自然语言到一阶逻辑的翻译。

This dataset contains 34,000 high-quality, diverse sentence-level pairs of natural language and first-order logic, which were collected from GPT-4. When constructing this dataset, we employed a prompting pipeline by refining prompts to ensure the diversity and validity of the natural language-first-order logic pairs. With a total of 34,000 pairs, this dataset is designed for the task of translating natural language into first-order logic.