Saved in:
Bibliographic Details
Main Author: Bas, Tetiana
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2411.13738
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915028960018432
author Bas, Tetiana
author_facet Bas, Tetiana
contents This study investigates gender bias in large language models (LLMs) by comparing their gender perception to that of human respondents, U.S. Bureau of Labor Statistics data, and a 50% no-bias benchmark. We created a new evaluation set using occupational data and role-specific sentences. Unlike common benchmarks included in LLM training data, our set is newly developed, preventing data leakage and test set contamination. Five LLMs were tested to predict the gender for each role using single-word answers. We used Kullback-Leibler (KL) divergence to compare model outputs with human perceptions, statistical data, and the 50% neutrality benchmark. All LLMs showed significant deviation from gender neutrality and aligned more with statistical data, still reflecting inherent biases.
format Preprint
id arxiv_https___arxiv_org_abs_2411_13738
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Assessing Gender Bias in LLMs: Comparing LLM Outputs with Human Perceptions and Official Statistics
Bas, Tetiana
Computation and Language
Machine Learning
This study investigates gender bias in large language models (LLMs) by comparing their gender perception to that of human respondents, U.S. Bureau of Labor Statistics data, and a 50% no-bias benchmark. We created a new evaluation set using occupational data and role-specific sentences. Unlike common benchmarks included in LLM training data, our set is newly developed, preventing data leakage and test set contamination. Five LLMs were tested to predict the gender for each role using single-word answers. We used Kullback-Leibler (KL) divergence to compare model outputs with human perceptions, statistical data, and the 50% neutrality benchmark. All LLMs showed significant deviation from gender neutrality and aligned more with statistical data, still reflecting inherent biases.
title Assessing Gender Bias in LLMs: Comparing LLM Outputs with Human Perceptions and Official Statistics
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2411.13738