Study finds multimodal LLMs perpetuate gender bias in musical instrument associations

A new study published on arXiv investigates gender bias in multimodal large language models (LLMs) by examining their associations with musical instruments. Researchers developed the Symphony-Bias dataset, which includes text, vision, and audio modalities for 22 instruments, and tested ten different LLMs. The findings indicate that 92% of the models' associations align with existing social science research on gendered instruments, with the harp and drums showing particularly consistent biases across all modalities. The study also observed that gender bias is weakest in audio, stronger in vision, and most pronounced in text. AI

IMPACT Highlights how multimodal LLMs can reinforce societal gender stereotypes, particularly through text and vision modalities, impacting how AI systems perceive and represent gendered associations.

RANK_REASON The cluster contains a research paper detailing findings on bias in multimodal LLMs.

Read on Hugging Face Daily Papers →

• Audio
• drum kit
• female
• harmonica
• male
• multimodal LLMs
• musical instruments
• non-binary person
• plain text
• Symphony-Bias
• Symphony of Bias
• visual perception
• arXiv
• Hugging Face Daily Papers

• paper
• safety

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →