Intellecta

Intellecta数据集是由Bud Ecosystem Inc创建的一个综合性合成数据集，旨在提升当代语言模型的认知处理能力。该数据集包含11.53亿个tokens，其中69.5%为合成数据，30.5%为来自学术出版物的教科书数据。数据集通过Mixtral-8x7B-Instruct-v0.1模型生成，涵盖编程、数学、自然语言处理等多个领域，旨在增强语言模型的复杂思维和教育叙事生成能力。创建过程中，采用动态提示生成系统，确保数据多样性，避免模型过度拟合。Intellecta数据集的应用领域广泛，旨在解决语言模型在复杂认知任务中的局限性，推动AI领域的进步和创新。

Intellecta Dataset is a comprehensive synthetic dataset created by Bud Ecosystem Inc., designed to boost the cognitive processing capabilities of modern language models. It comprises 1.153 billion tokens, where 69.5% are synthetically generated data and 30.5% are textbook materials sourced from academic publications. Developed using the Mixtral-8x7B-Instruct-v0.1 model, the dataset spans multiple disciplines including programming, mathematics, natural language processing and others, with the core objective of enhancing language models' complex reasoning and educational narrative generation capacities. During its construction, a dynamic prompt generation system was utilized to ensure data diversity and prevent model overfitting. The Intellecta Dataset has broad application prospects, aiming to address the limitations of language models in complex cognitive tasks and drive progress and innovation within the AI domain.