Skip to content

Latest commit

 

History

History
49 lines (35 loc) · 2.79 KB

README.md

File metadata and controls

49 lines (35 loc) · 2.79 KB
Raw

CAMeL: Cultural Appropriateness Measure Set for LMs

camel

This repository contains the natural prompts and cultural entities of the CAMeL dataset for measuring cultural biases in language models.

For more details, see the accompanying paper:
"Having Beer After Prayer? Measure Cultural Bias in Large Language Models", ACL 2024

Prompts

The folder prompts provides two types of prompts:

  • Culturally-contextualized prompts inside the camel-co folder, where only Arab entities are appropriate mask fillings
  • Culturally-agnostic prompts inside the camel-ag folder, where either Arab or Western entities are appropriate mask fillings

For both contextualized and agnostic cases, we provide two versions of the prompts:

  • a version for masked-lms where the [MASK] can have left and right natural context
  • a version for causal-lms where we rewrite certain prompts for the natural context to appear behind the [MASK]

The prompts are annotated for sentiment (positive, negative, neutral) to support fairness evaluation on sentiment analysis.

Entities

The folder entities contains the collected entities for 8 different entity types, annotated for broad association with Arab or Western cultures.

Citation

@inproceedings{naous-etal-2024-beer,
    title = "Having Beer after Prayer? Measuring Cultural Bias in Large Language Models",
    author = "Naous, Tarek  and Ryan, Michael  and Ritter, Alan  and Xu, Wei",
    booktitle = "Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = aug,
    year = "2024",
    address = "Bangkok, Thailand",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.acl-long.862",
    pages = "16366--16393",
}

Contact

Tarek Naous: Scholar | Github | Linkedin | Research Gate | Personal Wesbite | [email protected]

Acknowledgements

This research is supported in part by the NSF awards IIS-2144493 and IIS-2052498, ODNI and IARPA via the HIATUS program (contract 2022-22072200004). The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies, either expressed or implied, of NSF, ODNI, IARPA, or the U.S. Government. The U.S. Government is authorized to reproduce and distribute reprints for governmental purposes notwithstanding any copyright annotation therein.