Invisible Relevance Bias: Text-Image Retrieval Models Prefer AI-Generated Images - Multimedia 1

Explore a 12-minute conference talk from the Association for Computing Machinery (ACM) that delves into the phenomenon of "Invisible Relevance Bias" in text-image retrieval models. Learn how these models exhibit a preference for AI-generated images over human-created ones. Discover insights from authors Shicheng Xu, Danyang Hou, Liang Pang, Jingcheng Deng, Jun Xu, Huawei Shen, and Xueqi Cheng as they present their findings on this intriguing bias in multimedia retrieval systems. Gain a deeper understanding of the implications this preference may have on the field of information retrieval and the potential challenges it poses for fair and accurate image search results.

SIGIR 2024 M1.4 [fp] Invisible Relevance Bias: Text-Image Retrieval Models Prefer AI-Generated Image