Tagengo

Tagengo数据集是由Lightblue KK创建的一个高质量多语言聊天数据集，包含超过70,000个提示-响应对，涵盖74种语言。该数据集由人类生成的提示和合成响应组成，用于训练和评估多语言大型语言模型。创建过程中，数据集通过严格的筛选和清洗，确保了数据的质量和多样性。Tagengo数据集主要应用于提升大型语言模型在多语言环境下的聊天能力，解决现有模型在非英语语言上的性能不足问题。

The Tagengo Dataset is a high-quality multilingual chat dataset developed by Lightblue KK. It contains over 70,000 prompt-response pairs spanning 74 languages. Composed of human-generated prompts and synthetic responses, this dataset is designed for training and evaluating multilingual large language models (LLMs). During its creation, the dataset underwent strict filtering and cleaning processes to guarantee its quality and diversity. The primary application of the Tagengo Dataset is to improve the chat capabilities of large language models in multilingual environments, addressing the performance deficiencies of existing models in non-English languages.