mlsanityData to 14 Oct 2026

STEMTOX: From Collaborative Tags to Fine-Grained Toxic Meme Detection via Entropy-Guided Multi-Task Learning

arXiv 2508.04166 ↗First posted 6 Aug 2025cs.CV cs.CLcited by 0

On the map: LLM safety and security → Safety alignment, refusal and guardrails

Abstract (arXiv)

Memes, as a widely used mode of online communication, often serve as vehicles for spreading harmful content. However, limitations in data accessibility and the high costs of dataset curation hinder the development of robust meme moderation systems. To address this challenge, in this work, we introduce a first-of-its-kind dataset - TOXICTAGS consisting of 6,300 real-world meme-based posts annotated in two stages: (i) binary classification into toxic and normal, and (ii) fine-grained labelling of toxic memes as hateful, dangerous, or offensive. A key feature of this dataset is that it includes collaborative tags associated with the original posts, enhancing the context of each meme. In addition, we propose a novel entropy-guided multi-tasking framework -- STEMTOX -- that leverages these collaborative tags alongside visual and textual inputs within a robust classification framework. Experimental results show that incorporating these tags substantially enhances the performance of state-of-the-art VLMs in toxicity detection tasks. Our contributions offer a novel and scalable foundation for improved content moderation in multimodal online environments. Warning: Contains potentially toxic contents.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Safety alignment, refusal and guardrails, one of 12 directions in LLM safety and security. Builds on 27 of its 45 references that are in the map. First posted on arXiv 6 Aug 2025.

Builds on

27 of 45 references.

Closest ancestors · 1

Same direction, newest first

  1. 2024Exploring the Limits of Zero Shot Vision Language Models for Hate Meme Detection: The Vulnerabilities and their Interpretations19 Feb

Other references · 26

  1. 2023Improved Baselines with Visual Instruction Tuningcited by 4,418
  2. 2016A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networkscited by 2,385
  3. 2017Automated Hate Speech Detection and the Problem of Offensive Languagecited by 736
  4. 2020The Hateful Memes Challenge: Detecting Hate Speech in Multimodal Memescited by 490
  5. 2020HateXplain: A Benchmark Dataset for Explainable Hate Speech Detectioncited by 407
  6. 2023Recognize Anything: A Strong Image Tagging Modelcited by 309
  7. 2021MultiVerS: Improving scientific claim verification with weak supervision and full-document contextcited by 143
  8. 2021MOMENTA: A Multimodal Framework for Detecting Harmful Memes and Their Targetscited by 86
  9. 2023Tag2Text: Guiding Vision-Language Model via Image Taggingcited by 86
  10. 2021Detecting Harmful Memes and Their Targetscited by 78
  11. 2023Prompting for Multimodal Hateful Meme Classificationcited by 69
  12. 2023Open-Set Image Tagging with Multi-Grained Text Supervisioncited by 67
14 more
  1. 2023Probing LLMs for hate speech detection: strengths and vulnerabilitiescited by 52
  2. 2024GOAT-Bench: Safety Insights to Large Multimodal Models through Meme-Based Social Abusecited by 52
  3. 2023Pro-Cap: Leveraging a Frozen Vision-Language Model for Hateful Meme Detectioncited by 47
  4. 2025Can't See the Forest for the Trees: Benchmarking Multimodal Safety Awareness for Multimodal LLMscited by 23
  5. 2025Multimodal Language Models See Better When They Look Shallowercited by 23
  6. 2023BanglaAbuseMeme: A Dataset for Bengali Abusive Meme Classificationcited by 18
  7. 2024Modularized Networks for Few-shot Hateful Meme Detectioncited by 15
  8. 2024Vision-Language Models Create Cross-Modal Task Representationscited by 15
  9. 2023Exploring Large Language Models for Multi-Modal Out-of-Distribution Detectioncited by 13
  10. 2023Multi-modal Hate Speech Detection using Machine Learningcited by 8
  11. 2024Bridging Modalities: Enhancing Cross-Modality Hate Speech Detection with Few-Shot In-Context Learningcited by 7
  12. 2023PromptMTopic: Unsupervised Multimodal Topic Modeling of Memes using Large Language Modelscited by 5
  13. 2025Toxicity Begets Toxicity: Unraveling Conversational Chains in Political Podcastscited by 4
  14. 2025MMBERT: Scaled Mixture-of-Experts Multimodal BERT for Robust Chinese Hate Speech Detection under Cloaking Perturbationscited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025From Recognition to Reasoning: Advancing Multimodal Harmful Meme Detection via Chain-of-Thought Alignment
  2. 2024Just KIDDIN: Knowledge Infusion and Distillation for Detection of INdecent Memes
  3. 2026HarmTrace: Anchor-Calibrated Decoupled Optimization for Fine-Grained Target Identification in Harmful Memes
  4. 2023Beneath the Surface: Unveiling Harmful Memes with Multimodal Reasoning Distilled from Large Language Models
  5. 2025SMARTER: A Data-efficient Framework to Improve Toxicity Detection with Explanation via Self-augmenting Large Language Models
  6. 2025Beyond One-Size-Fits-All: Personalized Harmful Content Detection with In-Context Learning
  7. 2025LLM-based Semantic Augmentation for Harmful Content Detection
  8. 2025From Shallow Humor to Metaphor: Towards Label-Free Harmful Meme Detection via LMM Agent Self-Improvement