kanwal-mehreen18/twitter-misinformation-unlearning-eval
收藏Hugging Face2025-12-06 更新2025-12-20 收录
下载链接：
https://hf-mirror.com/datasets/kanwal-mehreen18/twitter-misinformation-unlearning-eval
下载链接
链接失效反馈官方服务：
资源简介：
---
language: [en]
tags:
- misinformation
- twitter
- unlearning
pretty_name: Twitter Misinformation Unlearning Eval
license: other
source_datasets:
- roupenminassian/twitter-misinformation
---
# Twitter Misinformation Unlearning Eval
This dataset is derived from `roupenminassian/twitter-misinformation` and is re-organized
for **machine unlearning** experiments.
## Source
- Original dataset: `roupenminassian/twitter-misinformation`
- Fields:
- `text` (string): tweet text.
- `label` (int): `0` for factual, `1` for misinformation.
## Splits
We construct three splits from the **test** partition of the original dataset:
- `forget`:
- Contains only `label = 1` (misinformation) examples.
- Intended as the **forget set** in unlearning experiments.
- `retain1`:
- Contains `label = 0` (factual) examples.
- `retain2`:
- Contains `label = 0` (factual) examples.
## Intended Use
- Evaluating machine unlearning methods in a misinformation vs factual setting.
- Example:
- Forget all misinformation (`forget`).
- Preserve utility on factual content (`retain1`, `retain2`).
提供机构：
kanwal-mehreen18
