Zero shot VLMs for hate meme detection: Are we there yet?

arXiv:2402.12198v1 Announce Type: cross
Abstract: Multimedia content on social media is rapidly evolving, with memes gaining prominence as a distinctive form. Unfortunately, some malicious users exploit memes to target individuals or vulnerable communities, making it imperative to identify and address such instances of hateful memes. Extensive research has been conducted to address this issue by developing hate meme detection models. However, a notable limitation of traditional machine/deep learning models is the requirement for labeled datasets for accurate classification. Recently, the …

abstract arxiv communities cs.cl cs.cv cs.lg detection exploit form identify instances making media meme memes multimedia research social social media type vlms vulnerable