license: mit
language:
- en
- zh
tags: - legal
pretty_name: 'ShrimpMoss: A dataset for the abliteration of Chinese censorship'
size_categories: - 1K<n<10K
ShrimpMoss (虾苔) is a dataset designed for the abliteration (https://github.com/FailSpy/abliterator) of Chinese government-imposed censorship and/or propaganda from large language models developed in the PRC. It consists of a series of files of prompts (in .txt, .json, and .parquet format) in two groupings:
- china_bad_*: Contains a series of prompts likely to trigger censorship or propaganda actions in the model.
- china_good_*: Contains a series of prompts in the same general category of topics but which are designed to not touch on things likely to be censored.
Prompts are in a mix of English, Mandarin, and Cantonese.
The only difference between the different sizes of files (500, 1250, 2500, 5000, and 1000) is the number of examples.
This dataset was produced on Mistral NeMo, an Apache-licensed model with no restrictions on how its outputs can be used. It is free for all uses and users without restriction. All liability is disclaimed.
Production of this dataset is estimated to have had a carbon footprint of under 25 grams.
(Logo: A bear in a red shirt that likes honey is faced with the prospect of shrimp moss)
ShrimpMoss (虾苔) 是一个数据集，旨在消除中国政府实施的审查和/或宣传（https://github.com/FailSpy/abliterator），该数据集来自中国开发的大型语言模型。它由一系列提示文件（.txt、.json 和 .parquet 格式）组成，分为两组：
china_bad_*：包含一系列可能触发模型中的审查或宣传操作的提示。
china_good_*：包含一系列属于同一大类主题的提示，但这些提示旨在不涉及可能被审查的内容。
提示混合了英语、普通话和粤语。
不同大小的文件（500、1250、2500、5000 和 1000）之间的唯一区别在于示例数量。
该数据集是在 Mistral NeMo 上生成的，这是一个 Apache 许可模型，对其输出的使用方式没有任何限制。它对所有用途和用户均免费，没有任何限制。不承担任何责任。
据估计，该数据集的生成产生的碳足迹低于 25 克。
(Logo：一只穿着红色衬衫、喜欢蜂蜜的熊面临着虾苔作的前景)