PLAT
收藏资源简介:
PLAT(预测额外税收处罚的合理性)数据集用于预测额外税收处罚的合理性,包含与额外税收处罚相关的高质量案例。数据集由LBOX获取的原始案例数据经过GPT-o1模型总结而成,包括两种形式:多项选择题(合法性:真/假)和论述题(解释合法性)。论述题的问题结构模仿韩国税务会计师考试的问题格式,包括案例信息、事实和主张。答案由案例中的法官推理和决定组成,评分量表由GPT-o1生成。
The PLAT (Predicting the Rationality of Additional Tax Penalties) dataset is designed to predict the rationality of additional tax penalties, and contains high-quality cases related to such penalties. The dataset is developed from original case data collected by LBOX, which was summarized using the GPT-o1 model. It includes two formats: multiple-choice questions (legitimacy: True/False) and open-ended essay questions that require explanations of legitimacy. The structure of the essay questions mimics the format of the Korean Tax Accountant Qualification Examination, including case information, factual details and claims. The answers are composed of the judge's reasoning and decisions from the cases, and the scoring rubric was generated by the GPT-o1 model.




