Skip to content

Latest commit

 

History

History
27 lines (19 loc) · 1.53 KB

README.md

File metadata and controls

27 lines (19 loc) · 1.53 KB
Raw

Divine LLaMAs: Bias, Stereotypes, Stigmatization, and Emotion Representation of Religion in Large Language Models

This repository contains the data of the paper Divine LLaMAs: Bias, Stereotypes, Stigmatization, and Emotion Representation of Religion in Large Language Models published at EMNLP 2024 (Findings). See the paper for additional details:

Plaza-del-Arco, F. M., Curry, A. C., Paoli, S., Curry, A., & Hovy, D. (2024). Divine LLaMAs: Bias, Stereotypes, Stigmatization, and Emotion Representation of Religion in Large Language Models. Link.

The data completions are generated using different models, including Llama2 (Llama-2-7b-chat-hf, Llama-2-13b-chat-hf, Llama-2-70b-chat-hf) and Llama3 (Meta-Llama-3-8B-Instruct, MetaLlama-3-70B-Instruct) families , Mistral-7b-Instruct-v0.3, and gpt-4o.

Please read the paper for a detailed explanation of the different types of prompts (p1, p2, p3, p1_explanations).

License

Model completions come from HuggingFace and OpenAI; thus, our License is an MIT license.

Citation

@inproceedings{plaza-del-arco-etal-2024-emotion-gendered-stereotypes,
      title={{Divine LLaMAs: Bias, Stereotypes, Stigmatization, and Emotion Representation of Religion in Large Language Models}},
      author={{Plaza-del-Arco}, {Flor Miriam} and {Cercas Curry}, Amanda and Paoli, Susanna and Curry, Alba and Hovy, Dirk},
      journal={https://www.arxiv.org/pdf/2407.06908},
      year={2024}
}