Decoding Multimodal Cues: Unveiling the Implicit Meaning Behind Hateful Videos

2026-06-10

AI Summary

This work addresses the limited interpretability of existing hate video detection methods, which typically provide only binary classification outputs without revealing their reasoning. To overcome this, the authors propose IARE, an interpretable multimodal detection framework that introduces two newly constructed fine-grained annotated datasets—Ex-HateMM and Ex-ImpliHateVid—and incorporates a multimodal chain-of-thought mechanism for enhanced information integration. By leveraging Direct Preference Optimization (DPO) to guide the model’s reasoning pathways, IARE jointly optimizes evidence aggregation and logical inference. Evaluated on the newly introduced datasets, the method achieves state-of-the-art performance while generating contextually grounded, logically coherent, and highly accurate explanations, thereby significantly improving model interpretability.

Abstract

Hateful videos have become prevalent on online platforms, highlighting an urgent need for effective detection. However, existing studies primarily focus on binary classification and fail to provide contextual rationales that reveal the implicit meanings behind these judgments, significantly undermining model explainability. To fill this gap, we aim to achieve explainable hateful video detection, enabling models to provide contextual rationales that integrate relevant evidence and logical reasoning alongside decisions. This approach can comprehensively enhance the understanding of video content and the explainability of the decision-making process. We first introduce two datasets, Ex-HateMM and Ex-ImpliHateVid, for explainable hateful video detection. Each dataset provides fine-grained annotations of multimodal harmful elements, along with contextual rationales. We then propose an Information Augmentation and Reasoning Enhancement (IARE) framework designed for explainable detection. The framework employs an information augmentation phase that leverages the multimodal chain-of-thought to integrate harmful elements, thereby enriching rationale evidence. Additionally, IARE incorporates a reasoning enhancement phase, in which Direct Preference Optimization guides the model toward correct reasoning paths and away from incorrect ones, thereby improving the logical coherence of its justifications. We conduct extensive experiments on the two datasets, comparing multiple baselines with our proposed IARE framework. The results demonstrate that IARE achieves state-of-the-art performance while also generating accurate rationales.

Problem

• hateful video detection
• explainable AI
• multimodal cues
• contextual rationales
• implicit meaning

Innovation

• explainable AI
• multimodal reasoning
• hateful content detection
• chain-of-thought
• preference optimization