Multimodal Large Language Models to Support Real-World Fact-Checking

TL;DR

This paper systematically evaluates multimodal large language models' ability to support real-world fact-checking, revealing GPT-4V's superior performance and highlighting biases in open-source models, with implications for trustworthy misinformation detection.

Contribution

It introduces a novel framework for assessing MLLMs' fact-checking capabilities using intrinsic knowledge and reasoning, the first such evaluation in this context.

Findings

GPT-4V outperforms others in detecting misleading claims

Open-source models show strong biases and prompt sensitivity

Insights for developing secure, trustworthy multimodal models

Abstract

Multimodal large language models (MLLMs) carry the potential to support humans in processing vast amounts of information. While MLLMs are already being used as a fact-checking tool, their abilities and limitations in this regard are understudied. Here is aim to bridge this gap. In particular, we propose a framework for systematically assessing the capacity of current multimodal models to facilitate real-world fact-checking. Our methodology is evidence-free, leveraging only these models' intrinsic knowledge and reasoning capabilities. By designing prompts that extract models' predictions, explanations, and confidence levels, we delve into research questions concerning model accuracy, robustness, and reasons for failure. We empirically find that (1) GPT-4V exhibits superior performance in identifying malicious and misleading multimodal claims, with the ability to explain the unreasonable aspects and underlying motives, and (2) existing open-source models exhibit strong biases and are highly sensitive to the prompt. Our study offers insights into combating false multimodal information and building secure, trustworthy multimodal models. To the best of our knowledge, we are the first to evaluate MLLMs for real-world fact-checking.