darshannere/Random-15k_AceReason-1.1-SFT
收藏Hugging Face2025-10-11 更新2025-10-25 收录
下载链接：
https://hf-mirror.com/datasets/darshannere/Random-15k_AceReason-1.1-SFT
下载链接
链接失效反馈官方服务：
资源简介：
AceReason 11 100K是一个为了微调和评估小型到中型大规模语言模型推理能力而设计的数据集。它包含了数学、逻辑、编码和科学推理等多种类型的推理任务。数据集由15000个随机示例组成，总共包含100000个推理提示和完成示例，每个示例格式为JSON对象，包括指令、输入和输出字段。该数据集可用于指令微调或推理评估。
AceReason 11 100K is a dataset designed for fine-tuning and evaluating the reasoning capabilities of small- to mid-scale large language models. It includes diverse reasoning tasks such as mathematics, logic, coding, and scientific reasoning. The dataset consists of 15,000 random samples out of a total of 100,000 reasoning prompts and completions, with each example formatted as a JSON object containing fields for the instruction, input, and output. It can be used for instruction-tuning or reasoning evaluation.
提供机构：
darshannere
