WithinUsAI/kimi_2.6_Thinking_Distilled_5k

Kimi 2.6 Thinking — Distilled Reasoning Dataset是一个包含5,000个高质量、无重复推理示例的数据集，旨在模拟Kimi 2.6 Thinking模型的思维链推理轨迹。该数据集覆盖14个领域，包括数学（如代数、微积分、几何）、编码问题、科学推理、逻辑谜题、实际推理和伦理推理等，每个示例都提供结构化推理步骤（包括逐步逻辑和验证）以及最终答案。数据集采用JSONL格式，每个示例包含instruction（问题或提示）、reasoning（提炼的思维/推理轨迹）、response（最终简洁答案）和domain（领域标识符）等字段，适用于微调、蒸馏或评估任务。Kimi 2.6 Thinking — Distilled Reasoning Dataset is a collection of 5,000 high-quality, no-duplicate reasoning examples designed to mirror the chain-of-thought reasoning traces of Kimi 2.6 Thinking. It spans 14 domains, including mathematics (e.g., algebra, calculus, geometry), coding problems, science reasoning, logic puzzles, practical reasoning, and ethics reasoning. Each example features structured reasoning with step-by-step logic and verification, along with a final concise answer. The dataset is formatted in JSONL, with fields such as instruction, reasoning, response, and domain, making it suitable for fine-tuning, distillation, or evaluation tasks.