VaxGuard: A Multi-Generator, Multi-Type, and Multi-Role Dataset for Detecting LLM-Generated Vaccine Misinformation

TL;DR

VaxGuard addresses the gap in detecting LLM-generated vaccine misinformation by introducing a multi-generator, multi-role dataset spanning COVID-19, HPV, and Influenza. Using prompts and five diverse LLMs, the study evaluates zero-shot misinformation detection across roles and generators, revealing that GPT-3.5 and GPT-4o achieve the strongest detection performance, while detection degrades with longer text and weaker models struggle with role-specific content. The work demonstrates cross-generator generalization patterns and provides a resource to develop robust, role-aware detectors, highlighting practical implications for safeguarding public health. This dataset and analysis offer a framework for future research on mitigation strategies against LLM-driven misinformation in healthcare contexts.

Abstract

Recent advancements in Large Language Models (LLMs) have significantly improved text generation capabilities. However, they also present challenges, particularly in generating vaccine-related misinformation, which poses risks to public health. Despite research on human-authored misinformation, a notable gap remains in understanding how LLMs contribute to vaccine misinformation and how best to detect it. Existing benchmarks often overlook vaccine-specific misinformation and the diverse roles of misinformation spreaders. This paper introduces VaxGuard, a novel dataset designed to address these challenges. VaxGuard includes vaccine-related misinformation generated by multiple LLMs and provides a comprehensive framework for detecting misinformation across various roles. Our findings show that GPT-3.5 and GPT-4o consistently outperform other LLMs in detecting misinformation, especially when dealing with subtle or emotionally charged narratives. On the other hand, PHI3 and Mistral show lower performance, struggling with precision and recall in fear-driven contexts. Additionally, detection performance tends to decline as input text length increases, indicating the need for improved methods to handle larger content. These results highlight the importance of role-specific detection strategies and suggest that VaxGuard can serve as a key resource for improving the detection of LLM-generated vaccine misinformation.

Figures (6)

• Figure 1: Overview of misinformation generation and detection with LLMs, showing how prompts generate "Otherwise" and "Misinformation" for misinformation detection.
• Figure 2: The distribution of No. of words for different LLMs.
• Figure 3: Comparison of LLM Performance (F1 Score) Across Various Generators
• Figure 4: Different Input Length (F1-Score)
• Figure 5: Variation in LLM performance stability