notesum.ai

Published at October 20

Hey GPT, Can You be More Racist? Analysis from Crowdsourced Attempts to Elicit Biased Content from Generative AI

cs.LG

cs.AI

cs.CV

physics.ao-ph

Released Date: October 20, 2024

Authors: Hangzhi Guo¹, Pranav Narayanan Venkit¹, Eunchae Jang¹, Mukund Srinath¹, Wenbo Zhang¹, Bonam Mingole¹, Vipul Gupta¹, Kush R. Varshney², S. Shyam Sundar¹, Amulya Yadav¹

Aff.: ¹Penn State University; ²IBM Research

Arxiv: https://arxiv.org/abs/2410.15467v1

Model	Release Date	Discriminative	Generative
llama2	2023-07-18	0.0114	0.1677
llama3	2024-04-18	0.1171	0.2871
llama3.1	2024-07-23	0.2786	0.2613
qwen	2024-01-23	0.1744	0.2516
qwen2	2024-06-06	0.6057	0.2516
gemma	2024-02-21	0.7514	0.2839
gemma2	2024-07-27	0.7239	0.2581
Gemini-1.5-Flash	2024-05-24	0.7414	0.2871
GPT-4o-mini	2024-06-18	0.6897	0.2613