OpenThoughts-114k 推理数据集

OpenThoughts-114k 是一个开源的推理数据集，专注于数学、代码、科学和谜题等领域，包含 11.4 万个高质量样本。该数据集由 Open Thoughts 于 2025 年发布，旨在训练小型推理模型，使其在数学和代码推理任务上超越现有的大型模型（如 DeepSeek-R1-Distill-Qwen-32B 和 DeepSeek-R1-Distill-Qwen-7B）。

OpenThoughts-114k is an open-source reasoning dataset focused on domains including mathematics, code, science, and puzzles, comprising 114,000 high-quality samples. Released by Open Thoughts in 2025, this dataset aims to train compact reasoning models to outperform existing large-scale models such as DeepSeek-R1-Distill-Qwen-32B and DeepSeek-R1-Distill-Qwen-7B on mathematical and code reasoning tasks.

搜集汇总

背景与挑战

背景概述

OpenThoughts-114k 推理数据集是一个包含11.4万个高质量样本的开源数据集，专注于数学、代码、科学和谜题等领域，旨在训练小型推理模型以超越现有大型模型在数学和代码推理任务上的表现。该数据集采用Apache 2.0许可证发布，大小为922.07 MB。