Learning from Mistakes: Enhancing Harmful Meme Detection via Misjudgment Risk Patterns

2025-10-10

AI Summary

Existing harmful meme detection methods struggle to identify rhetorically driven implicit harms—such as irony and metaphor—leading to high false-negative and false-positive rates. To address this, we propose PatMD, a novel framework that constructs a retrievable knowledge base of misclassification-risk patterns and leverages pattern retrieval coupled with dynamic reasoning to guide multimodal large language models (MLLMs) away from superficial content matching toward structured, pattern-aware risk identification. PatMD enables fine-grained misclassification prevention: on a benchmark of 6,626 memes, it improves average F1-score by 8.30% and accuracy by 7.71%, consistently outperforming state-of-the-art methods across five distinct harmful content detection tasks. Our core contribution is the first formalization of misclassification risk as explicit, retrievable, and reasoning-guidable knowledge—significantly enhancing MLLMs’ robustness in detecting covert harmful semantics.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Data Mining & Knowledge Management: Mining of Visual, Multimedia & Multimodal Data

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Internet memes have emerged as a popular multimodal medium, yet they are increasingly weaponized to convey harmful opinions through subtle rhetorical devices like irony and metaphor. Existing detection approaches, including MLLM-based techniques, struggle with these implicit expressions, leading to frequent misjudgments. This paper introduces PatMD, a novel approach that improves harmful meme detection by learning from and proactively mitigating these potential misjudgment risks. Our core idea is to move beyond superficial content-level matching and instead identify the underlying misjudgment risk patterns, proactively guiding the MLLMs to avoid known misjudgment pitfalls. We first construct a knowledge base where each meme is deconstructed into a misjudgment risk pattern explaining why it might be misjudged, either overlooking harmful undertones (false negative) or overinterpreting benign content (false positive). For a given target meme, PatMD retrieves relevant patterns and utilizes them to dynamically guide the MLLM's reasoning. Experiments on a benchmark of 6,626 memes across 5 harmful detection tasks show that PatMD outperforms state-of-the-art baselines, achieving an average of 8.30% improvement in F1-score and 7.71% improvement in accuracy, demonstrating strong generalizability and improved detection capability of harmful memes.

Problem

• Detecting harmful memes using implicit rhetorical devices
• Addressing misjudgment risks in multimodal meme detection
• Improving detection accuracy for ironic and metaphorical memes

Innovation

• Learning from misjudgment risk patterns
• Proactively guiding MLLMs to avoid pitfalls
• Dynamic reasoning using retrieved risk patterns