The paper introduces MemeMind, a large-scale dataset for detecting harmful memes that includes a detailed taxonomy and Chain-of-Thought reasoning annotations. It also proposes MemeGuard, a multimodal framework that uses a three-stage training strategy to improve visual understanding, reasoning, and discrimination of harmful content. Experiments show MemeGuard surpasses current state-of-the-art methods on MemeMind, advancing detection accuracy and interpretability.
By Hexiang Gu, Qifan Yu, Yuan Liu, Zikang Li, Saihui Hou, Jian Zhao, Zhaofeng He
arXiv:2609.13794v1 Announce Type: new
Abstract: Detecting harmful memes is critical for maintaining safe online communities. However, harmful intent is often implicit, arising from visual-textual inc...
By Hanling Wang, Chenlong Wei, Yingjuan Li, Di Wu, Yuchao Zhang, Xiaohui Zhu, Yao Zhu
arXiv:2610.19267v1 Announce Type: cross
Abstract: Online memes are often carriers of unsafe content such as disinformation and propaganda. Their rapid evolution in response to news events and strong...
By Bo Shao, Yihan Ma, Yiting Qu, Michael Backes, Yang Zhang
arXiv:2608. 16622v1 Announce Type: cross Abstract: Multimodal harmful meme detection is typically formulated as image--text harmfulness classification.
By Yujia Li, Yiqun Zhang, Zihan Cheng, Yijie Huang, Tenglong Ye, Zihan Wang, Xiaocui Yang, Shi Feng, Yifei Zhang, Daling Wang
arXiv:2606. 05316v1 Announce Type: new Abstract: Multimodal memes are dynamic and often require up to date background knowledge for interpretation.
By Shanhong Liu, Rui Cao, Pai Chet Ng, De Wen Soh
arXiv:2607. 15442v1 Announce Type: new Abstract: Internet memes intertwine visual cues, textual content, and cultural context, making them particularly challenging to interpret in scenarios where humor, sarcasm, and harmful intent coexist.
By Shanhong Liu, Pai Chet Ng, De Wen Soh, Malika Meghjani, Konstantinos N. Plataniotis
The paper introduces ProKDA, a progressive knowledge-to-decision alignment framework for explainable hateful meme detection. ProKDA separates explanation generation and label prediction into three sequential training stages—background knowledge learning, hatefulness detection learning, and hatefulness boundary alignment—reducing task interference. Experiments on three public benchmarks demonstrate that ProKDA achieves state‑of‑the‑art detection performance while providing accurate, evidence‑supported explanations for moderation decisions.
By Bo Xu, Chenyuan Wang, Xinyu Chen, Quanhao Zhu, Rui Lin, Liang Zhao, Hongfei Lin, Feng Xia
arXiv:2607. 03981v1 Announce Type: cross Abstract: Memes have become influential communication tools on social media, combining viral visuals with concise messaging to convey impactful ideas.
By Fatema Tuj Johora Faria, Mukaffi Bin Moin, Md. Mahfuzur Rahman, Pronay Debnath, Asif Iftekher Fahim, Faisal Muhammad Shah
MemeTAG introduces a dual‑objective framework for classifying harmful memes by combining keyword generation from a pretrained Vision‑Language Model with an Aggregated Tag Inference Network (ATIN) that condenses these keywords into a rich semantic embedding. The embedding is used as a target for an auxiliary reconstruction loss, encouraging deep alignment between visual and textual features. This approach, along with a three‑stage training strategy, achieves new state‑of‑the‑art results on the HarMeme, Hateful Memes Challenge, and PrideMM datasets.
By Akshit Sharma, Prashant W. Patil
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.
MemeLens is a unified multilingual, multitask Vision‑Language Model designed to improve meme understanding across a wide range of tasks such as hate, misogyny, propaganda, sentiment, and humour. The authors consolidated 38 public meme datasets, mapping their labels into a shared taxonomy of 20 tasks covering harm, targets, figurative intent, and affect, and conducted extensive experiments to show that multimodal training and a unified approach outperform fine‑tuning on individual datasets. All experimental resources, the model, and the datasets are released publicly for community use.
By Ali Ezzat Shahroor, Mohamed Bayan Kmainasi, Abul Hasnat, Dimitar Dimitrov, Giovanni Da San Martino, Preslav Nakov, Firoj Alam
Vis-Poison is a novel attack that poisons multimodal retrieval-augmented generation systems by inserting attacker-controlled images as visual evidence, without altering any textual metadata. The attack uses an automated multi-agent approach to create visually plausible poisoned images and has been tested on two multimodal RAG pipelines, four embedding models, and six generation models. In black-box settings, Vis-Poison achieves an end-to-end success rate between 40.16% and 65.40% against 30,000-entry knowledge bases, and remains effective against various multimodal large language models with an average success rate above 60%.
By Rujin Liang, Zhongpu Chen, Yuhao Lei, Xin Miao