alon-albalak/qwen-235b-a22b-noveltybench-comprehensive-evaluation-judgev1

该数据集包含结构化对话数据及其评估指标，主要特征包括：对话ID、提示词、消息内容、完成内容等基础字段；奖励分数、裁判评分及其统计量（均值、标准差、最大最小值）；分区信息和效用评分等高级指标。数据集仅包含100个样本的训练集。

The dataset contains structured dialogue data and evaluation metrics, with main features including: basic fields like dialogue ID, prompts, message content, completions; evaluation metrics like reward scores, judge scores and their statistics (mean, std, min/max); advanced metrics like partition information and utility scores. The dataset only contains a training split with 100 samples.