hkust-nlp/dart-math-pool-gsm8k-query-info
收藏Hugging Face2024-07-19 更新2024-06-12 收录
下载链接：
https://hf-mirror.com/datasets/hkust-nlp/dart-math-pool-gsm8k-query-info
下载链接
链接失效反馈官方服务：
资源简介：
DART-Math数据集是用于数学推理的最先进且数据高效的开源指令调优数据集。它通过难度感知拒绝采样（DARS）方法构建，特别强调对困难查询的偏向性。数据集包含约585k到591k个数学问答对样本，分别在MATH和GSM8K训练集上应用DARS-Prop2Diff和DARS-Uniform策略构建。DART-Math-Hard在多个具有挑战性的数学推理基准测试中达到了最先进的性能。
The dataset, named DART-Math-Pool-GSM8K Query Information, contains synthesis information of queries from the GSM8K training set, such as the numbers of raw/correct samples of each synthesis job. Usually used with dart-math-pool-gsm8k. The dataset features include query_id, n_steps, domain, n_samples_list, n_corrects_list, tot_n_samples, tot_n_corrects, pass_rate, and pass_rate_bernoulli_std. The dataset is split into a training set with 7473 examples. It is part of the DART-Math collection, which is a state-of-the-art, data-efficient, open-source instruction tuning dataset for mathematical reasoning. The dataset is used for text-generation tasks, specifically in the domains of mathematics and statistics.
提供机构：
hkust-nlp
原始信息汇总
数据集概述
数据集特征
- query_id: 字符串类型
- n_steps: 浮点数类型
- domain: 字符串类型
- n_samples_list: 整数序列类型
- n_corrects_list: 整数序列类型
- tot_n_samples: 整数类型
- tot_n_corrects: 整数类型
- pass_rate: 浮点数类型
- pass_rate_bernoulli_std: 浮点数类型
数据集划分
- 训练集 (train):
- 数据大小: 24,832,604 字节
- 示例数量: 7,473
数据集大小
- 下载大小: 402,245 字节
- 数据集总大小: 24,832,604 字节
