SyRHM: Symbolic Reasoning with Retrieval for Zero-shot Harmful Meme Detection

ACL ARR 2026 January Submission4456 Authors

Keywords: Symbolic Reasoning, Neuro-Symbolic AI, Multimodal Representation Learning, Harmful Content Detection

Abstract:

Detecting harmful memes is critical for maintaining safe online communities. However, harmful intent is often implicit, arising from visual–textual incongruity and cultural stereotypes, which challenges existing multimodal detectors. We propose SyRHM, a framework that decomposes hateful meme detection into meaning-grounded retrieval and symbolic multi-stage reasoning. SyRHM retrieves semantically related memes by parsing multimodal content into textual elements and description, providing grounded context beyond surface-level similarity. Building on the retrieved context, SyRHM uses a translator stage to convert multimodal inputs into symbolic intermediate representations, and then performs multi-stage reasoning via planner and solver stages, enabling expressive and interpretable analysis of harmful intent. Experiments on FHM, HarM, and MultiOff show that SyRHM consistently outperforms multimodal and reasoning-based baselines, while providing reasoning traces for harmful content.
The code is available at: https://anonymous.4open.science/r/SyRHM-CC46/README.md
Disclaimer: This paper contains offensive content that may be disturbing to some readers.

Paper Type: Long

Research Area: NLP Applications

Research Area Keywords: Hate speech detection, multimodal applications, NLP for social good

Contribution Types: Model analysis & interpretability

Languages Studied: English

Submission Number: 4456