26hzhang/math_dapo_qwen3-4b_rollout_n_10

该数据集包含了多个特征，如提示信息（内容与角色）、数据来源、能力、奖励模型（包括地面真实和风格）、额外信息（包括索引、原始索引、原始奖励模型风格、解决方案、分割、主题）以及 rollout（包括最大令牌数、数量、通过率、rollout答案、rollout标签、rollout响应、温度、top_k、top_p）。数据集分为训练集，包含大约17万3千9百8十个示例，大小约为3GB。

The dataset includes multiple features such as prompt information (content and role), data source, ability, reward model (including ground truth and style), extra information (including index, raw index, raw rm style, solution, split, subject), and rollout (including max tokens, n, pass rate, rollout answers, rollout labels, rollout responses, temperature, top_k, top_p). The dataset is split into a training set, containing approximately 173,980 examples, with a size of about 3GB.