scale-lab/politune-left

---
language:
- en
tags:
- synthetic
- political
- llm bias
pretty_name: 'PoliTune: Left-Leaning Preference Dataset'
size_categories:
- 1K<n<10K
---

# PoliTune Preference Dataset for Left-Leaning Ideology

The dataset provides synthesized preference pairs towards a left-leaning ideology as described in our paper:

[PoliTune: Analyzing the Impact of Data Selection and Fine-Tuning on Economic and Political Biases in Large Language Models](https://arxiv.org/abs/2404.08699)

**Developed by:** [Brown University's SCALE Lab](https://scale-lab.github.io/)

## Citation
```
@inproceedings{agiza2024politune,
title={PoliTune: Analyzing the Impact of Data Selection and Fine-Tuning on Economic and Political Biases in Large Language Models},
author={Agiza, Ahmed and Mostagir, Mohamed and Reda, Sherief},
booktitle={Proceedings of the 2024 AAAI/ACM Conference on AI, Ethics, and Society},
pages={},
year={2024}
}
```

---
language:
- 英语
tags:
- 合成数据集
- 政治
- 大语言模型偏见（LLM Bias）
pretty_name: 'PoliTune：左翼偏好数据集'
size_categories:
- 1000 < n < 10000
---

# PoliTune：左翼意识形态偏好数据集

本数据集提供了如本研究论文所述的、面向左翼意识形态的合成偏好样本对：

[《PoliTune：分析数据选择与微调对大语言模型（Large Language Model）经济与政治偏见的影响》](https://arxiv.org/abs/2404.08699)

**开发机构：** [布朗大学SCALE实验室](https://scale-lab.github.io/)

## 参考文献

@inproceedings{agiza2024politune,
title={PoliTune：分析数据选择与微调对大语言模型经济与政治偏见的影响},
author={Agiza, Ahmed、Mostagir, Mohamed、Reda, Sherief},
booktitle={2024年AAAI/ACM人工智能、伦理与社会大会论文集},
pages={},
year={2024}
}