taln-ls2n/parareval
收藏Hugging Face2025-12-12 更新2025-12-20 收录
下载链接：
https://hf-mirror.com/datasets/taln-ls2n/parareval
下载链接
链接失效反馈官方服务：
资源简介：
ParaReval数据集是一个包含人类对生成修订版本成对评估的数据集。数据来源于ParaRev语料库的测试集，每个数据点（共516个）生成了6个修订版本，使用了6种不同的LLM模型。标注过程中，评估者需要回答多个问题，包括相关性、可接受性、偏好以及类别特定的评估（如学术风格改进、可读性和结构改进等）。数据集以JSONL格式提供，每个数据点包含原始段落、参考段落、模型生成的修订版本以及详细的评估标注。
The ParaReval dataset is a collection of human pairwise evaluations of generated revisions. The data is derived from the test set of the ParaRev corpus, with each datapoint (516 in total) generating 6 revised versions using 6 different LLMs. During annotation, evaluators answer multiple questions, including relatedness, acceptability, preference, and category-specific evaluations (e.g., improvement in academic style, readability, and structure). The dataset is provided in JSONL format, with each datapoint containing the original paragraph, reference paragraph, model-generated revisions, and detailed evaluation annotations.
提供机构：
taln-ls2n
