RLAIF/dpo_answer_only_with_gold_labels_kl_estimation
收藏Hugging Face2025-08-07 更新2025-08-09 收录
下载链接：
https://hf-mirror.com/datasets/RLAIF/dpo_answer_only_with_gold_labels_kl_estimation
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含以下字段：步数(step)、问题(question)、参考(ref)、当前(current)、参考与当前的差异(ref_current)、当前与参考的差异(current_ref)、KL散度(kl_divergence)、金标准参考与当前的差异(gold_ref_current)、金标准当前与参考的差异(gold_current_ref)、估计的KL散度(kl_estimated)。数据集分为训练集(train)，共有43692个样本，大小为209,586,914字节。数据集的下载大小为98,962,229字节。
The dataset includes the following fields: step, question, ref, current, ref_current, current_ref, kl_divergence, gold_ref_current, gold_current_ref, kl_estimated. The dataset is split into a training set (train) with a total of 43,692 samples, and a size of 209,586,914 bytes. The download size of the dataset is 98,962,229 bytes.
提供机构：
RLAIF
