AHA-Memes: A Fine-Grained Multimodal Benchmark for Understanding Hate in Arabic Memes

2026-07-29

AI Summary

This work addresses the lack of fine-grained, multi-label multimodal benchmarks for Arabic hate memes by introducing AHA-Memes, the first large-scale dataset comprising 5K human-annotated samples and approximately 66K silver-labeled instances. The annotation framework integrates cultural context and encompasses diverse attack strategies prevalent in Arabic online discourse. The study systematically evaluates text-only, image-only, and multimodal models—including unimodal architectures, late-fusion approaches, and both open- and closed-source vision-language models—under zero-shot, few-shot, and fine-tuned settings. Beyond releasing high-quality data and evaluation tools, this research establishes strong baselines and uncovers key challenges in detecting hate memes within Arabic cultural contexts, thereby advancing the understanding of multimodal hate content in underrepresented languages.

Abstract

Hateful memes are a growing form of multimodal online harm, where hostile intent is often conveyed through the joint interpretation of images, text, cultural references, and implicit targets. While hateful meme detection has advanced in high-resource languages, Arabic remains underexplored, with existing meme resources focusing mainly on propaganda or coarse harmful-content labels. We introduce AHA-Memes (Arabic HAteful Memes), which is, to our knowledge, the first large-scale Arabic hateful meme benchmark with fine-grained, multi-label annotations. The dataset includes 5K manually annotated memes using a taxonomy that captures hate types, i.e., attack strategies. We further provide ~66K silver-labeled memes to support future studies. We benchmark text-only, image-only, and late-fusion multimodal models, as well as few-shot in-context learning (ICL) and open- and closed-weight Vision-Language Models (VLMs) under zero-shot and fine-tuning settings. Our results establish strong baselines and highlight key challenges in culturally grounded Arabic hateful meme detection. We release the dataset, annotation guidelines, and evaluation scripts to support future research. WARNING: This paper contains examples that may be disturbing to readers.

Problem

• hateful memes
• Arabic
• multimodal
• fine-grained annotation
• online harm

Innovation

• fine-grained annotation
• Arabic hateful memes
• multimodal benchmark
• Vision-Language Models
• silver-labeled data