DAM
收藏资源简介:
该数据集是一个包含三个子集的集合,旨在用于大型语言模型(LLMs)的指令调优和评估。这些子集包括用于日语指令的Ichikara、用于数学推理的MetaMathQA和用于小学数学问题解决的Orca Math。每个子集都有特定的关注点和来源,数据集结合这些子集以支持LLMs在语言和数学任务中的训练和评估。数据集包含1,729个样本,格式为Alpaca指令模板,包括日语指令-响应对、带有解释的数学问答对以及小学数学问题和解决方案。该数据集适用于指令调优、评估LLMs的性能以及训练LLMs处理语言和数学问题。它不适用于需要非日语语言处理、超出小学水平的数学问题或未经适当偏见和限制评估的实际应用。
This dataset is a collection of three subsets intended for instruction tuning and evaluation of Large Language Models (LLMs). The three subsets are Ichikara for Japanese instruction tasks, MetaMathQA for mathematical reasoning, and Orca Math for primary school mathematics problem-solving. Each subset has its own specific focus and source. Combining these subsets, the dataset supports the training and evaluation of LLMs on both linguistic and mathematical tasks. The dataset contains 1,729 samples formatted in the Alpaca instruction template, including Japanese instruction-response pairs, math question-answer pairs with explanations, as well as primary school mathematics problems and their solutions. This dataset is applicable for instruction tuning, evaluating the performance of LLMs, and training LLMs to handle linguistic and mathematical problems. It is not suitable for use cases requiring non-Japanese language processing, mathematical problems beyond the primary school level, or real-world applications that have not been properly evaluated for bias and restrictions.




