internlm/Condor-SFT-20K

Condor是一个用于生成监督微调（SFT）高质量数据的数据集，由InternLM开发。它包含两个阶段：数据合成阶段和数据精炼阶段。数据合成阶段利用世界知识树来生成数据，而数据精炼阶段则采用自我反思精炼策略来优化响应。Condor旨在帮助大型语言模型（LLMs）提高对话能力，特别是在只有少量高质量数据的情况下。数据集适用于文本生成任务，并支持英文和中文。

Condor is a dataset for generating high-quality Supervised Fine-Tuning (SFT) data, developed by InternLM. It consists of two stages: data synthesis and data refinement. The data synthesis stage uses the World Knowledge Tree to generate data, while the data refinement stage employs a Self-Reflection Refinement strategy to optimize responses. Condor aims to enhance the conversational capabilities of Large Language Models (LLMs), especially when only a small amount of high-quality data is available. The dataset is suitable for text generation tasks and supports both English and Chinese.