Ayush-Singh/reward-bench-gpt-4o-mini-yes-no

该数据集包含了多种类型的编程语言代码和相关的评估信息，如选择的代码片段(chosen)、拒绝的代码片段(rejected)以及它们对应的模型(chosen_model、rejected_model)。每个数据点还包括了其他信息，如属于哪个子集(subset)、唯一标识符(id)、以及不同情况下的概率(chosen_yes_prob、chosen_no_prob、rejected_yes_prob、rejected_no_prob)。数据集分为多个部分，如math_prm、xstest_should_refuse、alpacaeval_hard等，每个部分包含了不同数量的示例，总共包含了超过1100万个数据点。

The dataset includes various types of programming language code and related evaluation information, such as chosen code snippets, rejected code snippets, and their corresponding models. Each data point also includes other information, such as the subset it belongs to, a unique identifier, and probabilities under different conditions. The dataset is divided into multiple parts, such as math_prm, xstest_should_refuse, alpacaeval_hard, etc., each containing a different number of examples, totaling over 11 million data points.