jihuny/tldr-reference

TL;DR参考数据集是一个从OpenAI的summarize-from-feedback公共发布镜像而来的数据集，具体基于OpenAI的tldr_3_filtered数据集经过CleanRL预处理。该数据集包含来自Reddit的TL;DR（Too Long; Didnt Read）摘要参考数据，用于提示和参考训练以及评估任务。镜像版本保留了原始数据集的分割名称、列、标记化字段和行顺序，以便本地脚本可以使用较短的数据集IDjihuny/tldr-reference。数据集主要用于文本摘要任务，语言为英语。

The TL;DR Reference Dataset is a mirrored dataset from OpenAIs summarize-from-feedback public release, specifically based on OpenAIs tldr_3_filtered dataset with CleanRL preprocessing. It contains reference data from Reddit TL;DR (Too Long; Didnt Read) summaries for prompt and reference training and evaluation. The mirror retains the original split names, columns, tokenized fields, and row order to allow local scripts to use the shorter dataset ID jihuny/tldr-reference. The dataset is primarily used for text summarization tasks and is in English.