Datasets:
Tasks:
Text Generation
Modalities:
Text
Formats:
json
Languages:
Chinese
Size:
10K - 100K
Tags:
finetuning
License:
| license: cc-by-4.0 | |
| language: | |
| - zh | |
| tags: | |
| - finetuning | |
| pretty_name: WebCPM_WK-Dataset | |
| task_categories: | |
| - text-generation | |
| # Dataset Card for WebCPM_WK | |
| ## Dataset Description | |
| - **Homepage:** https://huggingface.co/datasets/ZHR123/WebCPM_WK | |
| - **Repository:** | |
| - **Paper:** | |
| - **Leaderboard:** | |
| - **Point of Contact:** Huanran Zheng | |
| ### Dataset Summary | |
| 本数据集是由我们对[WebCPM](https://github.com/thunlp/WebCPM)的pipeline数据进行二次处理之后构建而成。 | |
| 主要包括过滤原始数据中的一些低质量数据,使用GPT4和ChatGPT扩充原始数据,以及使用随机替换、拼接的方式增强原始数据。 | |
| 该数据集主要的目的是通过指令微调的方式提高LLM的两个能力: | |
| 1. 给定问题和文档,抽取文档中与问题相关知识的能力。 | |
| 2. 给定参考材料和问题,根据参考材料回答问题的能力。 | |
| ### Licensing Information | |
| The dataset is available under the [Creative Commons NonCommercial (CC BY-NC 4.0)](https://creativecommons.org/licenses/by-nc/4.0/legalcode). |