agentlans/epic-thinking

该数据集是一个综合性的推理数据集，汇集了来自多个公开来源的推理相关数据，包括思维链、指令微调、监督微调和蒸馏数据。数据集支持英语和中文，包含两个配置：all（完整数据）和filtered（过滤后数据）。它旨在用于训练和提升语言模型的推理能力，覆盖数学、科学、代码、通用推理等多个领域。

This dataset is a comprehensive reasoning dataset that aggregates reasoning-related data from multiple public sources, including chain-of-thought, instruction-tuning, supervised fine-tuning, and distillation data. The dataset supports both English and Chinese, and includes two configurations: all (full data) and filtered (filtered data). It is designed for training and enhancing the reasoning capabilities of language models, covering various domains such as mathematics, science, code, and general reasoning.