# Just KIDDIN: Knowledge Infusion and Distillation for Detection of INdecent Memes

arXiv: https://arxiv.org/abs/2411.12174 · HTML: https://mlsanity.com/paper/2411.12174/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 35 of its 63 references that are in the map.
- First posted on arXiv 19 Nov 2024.

## Builds on (most cited first, 30 of 35)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [Graph Attention Networks](https://mlsanity.com/paper/1710.10903/) (2017)
- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [Parameter-Efficient Transfer Learning for NLP](https://mlsanity.com/paper/1902.00751/) (2019)
- [Modeling Relational Data with Graph Convolutional Networks](https://mlsanity.com/paper/1703.06103/) (2017)
- [ConceptNet 5.5: An Open Multilingual Graph of General Knowledge](https://mlsanity.com/paper/1612.03975/) (2016)
- [The Hateful Memes Challenge: Detecting Hate Speech in Multimodal Memes](https://mlsanity.com/paper/2005.04790/) (2020)
- [QA-GNN: Reasoning with Language Models and Knowledge Graphs for Question Answering](https://mlsanity.com/paper/2104.06378/) (2021)
- [MiniLMv2: Multi-Head Self-Attention Relation Distillation for Compressing Pretrained Transformers](https://mlsanity.com/paper/2012.15828/) (2020)
- [GreaseLM: Graph REASoning Enhanced Language Models for Question Answering](https://mlsanity.com/paper/2201.08860/) (2022)
- [Supervised Multimodal Bitransformers for Classifying Images and Text](https://mlsanity.com/paper/1909.02950/) (2019)
- [Toxicity Detection: Does Context Really Matter?](https://mlsanity.com/paper/2006.00998/) (2020)
- [Can Knowledge Graphs Reduce Hallucinations in LLMs? : A Survey](https://mlsanity.com/paper/2311.07914/) (2023)
- [MOMENTA: A Multimodal Framework for Detecting Harmful Memes and Their Targets](https://mlsanity.com/paper/2109.05184/) (2021)
- [Detecting Harmful Memes and Their Targets](https://mlsanity.com/paper/2110.00413/) (2021)
- [Enabling Multimodal Generation on CLIP via Vision-Language Knowledge Distillation](https://mlsanity.com/paper/2203.06386/) (2022)
- [Prompting for Multimodal Hateful Meme Classification](https://mlsanity.com/paper/2302.04156/) (2023)
- [Hate-CLIPper: Multimodal Hateful Meme Classification based on Cross-modal Interaction of CLIP Features](https://mlsanity.com/paper/2210.05916/) (2022)
- [Towards Language Models That Can See: Computer Vision Through the LENS of Natural Language](https://mlsanity.com/paper/2306.16410/) (2023)
- [Pro-Cap: Leveraging a Frozen Vision-Language Model for Hateful Meme Detection](https://mlsanity.com/paper/2308.08088/) (2023)
- [Defining and Detecting Toxicity on Social Media: Context and Knowledge are Key](https://mlsanity.com/paper/2104.10788/) (2021)
- [Knowledge Infused Learning (K-IL): Towards Deep Incorporation of Knowledge in Deep Learning](https://mlsanity.com/paper/1912.00512/) (2019)
- [Multimodal Hate Speech Detection from Bengali Memes and Texts](https://mlsanity.com/paper/2204.10196/) (2022)
- [Beneath the Surface: Unveiling Harmful Memes with Multimodal Reasoning Distilled from Large Language Models](https://mlsanity.com/paper/2312.05434/) (2023)
- [A Survey on Echo Chambers on Social Media: Description, Detection and Mitigation](https://mlsanity.com/paper/2112.05084/) (2021)
- [Sentence-Level or Token-Level? A Comprehensive Study on Knowledge Distillation](https://mlsanity.com/paper/2404.14827/) (2024)
- [A survey on knowledge-enhanced multimodal learning](https://mlsanity.com/paper/2211.12328/) (2022)
- [ToxVidLM: A Multimodal Framework for Toxicity Detection in Code-Mixed Videos](https://mlsanity.com/paper/2405.20628/) (2024)
- [GOLD: Generalized Knowledge Distillation via Out-of-Distribution-Guided Language Data Generation](https://mlsanity.com/paper/2403.19754/) (2024)

## Built on by (most cited first, 8 of 8)

- [Playing Devil's Advocate: Unmasking Toxicity and Vulnerabilities in Large Vision-Language Models](https://mlsanity.com/paper/2501.09039/) (2025)
- [Large Language Models for Cancer Communication: Evaluating Linguistic Quality, Safety, and Accessibility in Generative AI](https://mlsanity.com/paper/2505.10472/) (2025)
- [From Reddit to Generative AI: Evaluating Large Language Models for Anxiety Support Fine-tuned on Social Media Data](https://mlsanity.com/paper/2505.18464/) (2025)
- [PrismAgent: Illuminating Harm in Memes via a Zero-Shot Interpretable Multi-Agent Framework](https://mlsanity.com/paper/2605.02940/) (2026)
- [Computational Humor with Multimodal LLMs: Methods, Datasets, Evaluation, and Challenges](https://mlsanity.com/paper/2607.19011/) (2026)
- [Towards High-Level Semantic Intelligence](https://mlsanity.com/paper/2607.24082/) (2026)
- [Learn Before You Judge: Progressive Knowledge-to-Decision Alignment for Explainable Hateful Meme Detection](https://mlsanity.com/paper/2609.19778/) (2026)
- [MemeLogos: Knowledge-Driven Retrieval-Augmented Reasoning for Harmful Meme Detection](https://mlsanity.com/paper/2610.15463/) (2026)

## Abstract (arXiv)

Toxicity identification in online multimodal environments remains a challenging task due to the complexity of contextual connections across modalities (e.g., textual and visual). In this paper, we propose a novel framework that integrates Knowledge Distillation (KD) from Large Visual Language Models (LVLMs) and knowledge infusion to enhance the performance of toxicity detection in hateful memes. Our approach extracts sub-knowledge graphs from ConceptNet, a large-scale commonsense Knowledge Graph (KG) to be infused within a compact VLM framework. The relational context between toxic phrases in captions and memes, as well as visual concepts in memes enhance the model's reasoning capabilities. Experimental results from our study on two hate speech benchmark datasets demonstrate superior performance over the state-of-the-art baselines across AU-ROC, F1, and Recall with improvements of 1.1%, 7%, and 35%, respectively. Given the contextual complexity of the toxicity detection task, our approach showcases the significance of learning from both explicit (i.e. KG) as well as implicit (i.e. LVLMs) contextual cues incorporated through a hybrid neurosymbolic approach. This is crucial for real-world applications where accurate and scalable recognition of toxic content is critical for creating safer online environments.