Pro-Cap: Leveraging a Frozen Vision-Language Model for Hateful Meme Detection

TL;DR

Pro-Cap introduces a novel zero-shot approach for hateful meme detection by prompting a frozen vision-language model with hate-related questions, generating informative captions that improve detection accuracy across benchmarks.

Contribution

The paper presents a probing-based captioning method that leverages frozen PVLMs for hateful meme detection without fine-tuning, enhancing efficiency and effectiveness.

Findings

Pro-Cap achieves strong performance on three benchmarks.

The method effectively captures hateful content information.

It demonstrates good generalization across datasets.

Abstract

Hateful meme detection is a challenging multimodal task that requires comprehension of both vision and language, as well as cross-modal interactions. Recent studies have tried to fine-tune pre-trained vision-language models (PVLMs) for this task. However, with increasing model sizes, it becomes important to leverage powerful PVLMs more efficiently, rather than simply fine-tuning them. Recently, researchers have attempted to convert meme images into textual captions and prompt language models for predictions. This approach has shown good performance but suffers from non-informative image captions. Considering the two factors mentioned above, we propose a probing-based captioning approach to leverage PVLMs in a zero-shot visual question answering (VQA) manner. Specifically, we prompt a frozen PVLM by asking hateful content-related questions and use the answers as image captions (which we call Pro-Cap), so that the captions contain information critical for hateful content detection. The good performance of models with Pro-Cap on three benchmarks validates the effectiveness and generalization of the proposed method.