R2E-Gym/32B_predictions

该数据集包含了多个字符串类型的特征字段，如Unnamed: 0, 地面真实奖励(gt_rewards), 预测奖励(predicted_rewards)等。数据集被分割为predictions部分，共有500个示例。数据集可用于分析预测奖励与实际奖励之间的关系，以及其它相关指标如步数(step_counts), 点对点比率(p2p_rates), 回归率(regression_rates)和平均肯定概率(avg_yes_prob)。

The dataset includes several string-typed feature fields such as Unnamed: 0, ground-truth rewards (gt_rewards), predicted rewards (predicted_rewards), etc. The dataset is split into a predictions section with a total of 500 examples. The dataset can be used to analyze the relationship between predicted and actual rewards, as well as other related metrics such as step counts (step_counts), peer-to-peer rates (p2p_rates), regression rates (regression_rates), and average yes probability (avg_yes_prob).