MultiturnRL/training-dataset

该数据集包含五个部分：1）ASearcher部分包含问题、答案、来源和增强答案等字段，有训练集分割；2）DeepMath部分包含问题、最终答案、难度、主题和相关解决方案等字段，有训练集分割；3）Default部分包含UUID、来源、提示、沙盒和额外信息等字段，有角色扮演、策略学习和测试集分割；4）Swegym部分包含实例ID、提示文本、补丁、测试补丁、创建时间、问题陈述、代码库、基础提交和版本等字段，有训练集分割；5）Taco部分包含问题、解决方案、起始代码、输入输出、难度、原始标签、名称、来源、标签、技能类型、URL、预期辅助空间、时间限制、日期、图片数量和内存限制等字段，有训练集分割。

The dataset consists of five parts: 1) The ASearcher section includes fields such as questions, answers, sources, and augmented answers, with a training set split; 2) The DeepMath section includes fields such as questions, final answers, difficulty, topics, and related solutions, with a training set split; 3) The Default section includes fields such as UUID, source, prompts, sandbox, and extra information, with splits for role-playing, strategy learning, and testing; 4) The Swegym section includes fields such as instance ID, hint text, patch, test patch, creation time, problem statement, repository, base commit, and version, with a training set split; 5) The Taco section includes fields such as questions, solutions, starter code, input/output, difficulty, raw tags, name, source, tags, skill types, URL, expected auxiliary space, time limit, date, picture number, and memory limit, with a training set split.