Datasets:
Tasks:
Text Generation
Modalities:
Text
Formats:
json
Languages:
Chinese
Size:
10K - 100K
Tags:
finetuning
License:
File size: 1,061 Bytes
7d47856 23ce955 7d47856 23ce955 2f33353 23ce955 079ef5c 23ce955 d21b0a3 23ce955 2f33353 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 |
---
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). |