qwselfcorr/math_and_augmath_cor20k_20k_2e6_3ep_tmp10_turn2
收藏Hugging Face2025-02-01 更新2025-02-15 收录
下载链接：
https://hf-mirror.com/datasets/qwselfcorr/math_and_augmath_cor20k_20k_2e6_3ep_tmp10_turn2
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了索引、提示文本、答案序列、正确答案、首次奖励、代理标签和二次奖励等字段。它被划分为了训练集，其中包含20000个示例，总大小为73482986字节。数据集的目的是用于训练模型，可能涉及到根据提示生成答案，并评估答案的正确性以及奖励机制。
The dataset includes fields such as index, prompt text, answer sequence, correct answer, first reward, proxy label, and second reward. It is split into a training set, which contains 20,000 examples and has a total size of 73482986 bytes. The purpose of the dataset is to train models, possibly involving generating answers based on prompts and evaluating the correctness of the answers and reward mechanisms.
提供机构：
qwselfcorr
