SINAI/ALIA-es-legal-synthetic-instructions
收藏资源简介:
ALIA西班牙法律与行政合成指令语料库是一个使用Magpie方法生成的合成语料库,专为西班牙法律领域设计,旨在为训练和评估语言模型提供资源。该数据集包含超过740万条指令-答案对,涵盖多种形式,如一般问题和指令、基于上下文的问题和指令、多项选择题(包括多选和判断题)等,并有带或不带解释的版本。数据生成使用了Phi-4模型,并经过了严格的清洗、语义验证和去重处理。该数据集适用于训练西班牙语法律大语言模型、法律问答系统以及其他法律推理任务。
The ALIA Spanish Legal and Administrative Synthetic Instructions Corpus is a synthetic corpus generated using the Magpie methodology, adapted for the Spanish legal domain, designed to provide resources for training and evaluating language models. The dataset contains over 7.4 million instruction-answer pairs, covering various formats such as general questions and instructions, context-based questions and instructions, multiple-choice test questions (including multiple-choice and true/false), with versions both with and without justification. The data was generated using the Phi-4 model and underwent a rigorous process of cleaning, semantic validation, and duplicate removal. This dataset is suitable for training Spanish legal large language models, legal QA systems, and other legal reasoning tasks.




