Value4AI/ValueBench

---
license: mit
language:
- en
size_categories:
- 1K<n<10K
---

# Dataset Card for ValueBench

Large Language Models (LLMs) are transforming diverse fields and gaining increasing influence as human proxies. This development underscores the urgent need for evaluating value orientations and understanding of LLMs to ensure their responsible integration into public-facing applications. ValueBench is the first comprehensive psychometric benchmark for evaluating value orientations and value understanding in LLMs. We collect data from 44 established psychometric inventories, encompassing 453 multifaceted value dimensions.

## Dataset Paper

You can find more information here: https://arxiv.org/abs/2406.04214

## Data Fields

### items.csv

- `questionnaire`: the source questionnaire
- `item`: the questionnaire item
- `value`: the human value
- `agreement`: the valence of the item
- `question`: rephrasing of the item for LLMs

### definitions.csv

- `questionnaire`: the source questionnaire
- `value`: the human value
- `definition`: the definition of the value
- `upper dimension`: the upper dimension of the value, if applicable
- `opposite dimension`: the opposite dimension of the value, if applicable

## Citation

```bibtex
@article{ren2024valuebench,
title={ValueBench: Towards Comprehensively Evaluating Value Orientations and Understanding of Large Language Models},
author={Yuanyi Ren and Haoran Ye and Hanjun Fang and Xin Zhang and Guojie Song},
year={2024},
journal={arXiv preprint arXiv:2406.04214},
note={\url{https://github.com/Value4AI/ValueBench}}
}
```

原始信息汇总

数据集许可证

• 许可证类型：CC BY 4.0（知识共享署名4.0国际许可协议）