This research evaluates optimal strategies for scaling social media annotation to detect stance in HPV vaccine-related content using large language models.

• Systematically compares in-context learning and fine-tuning approaches across multiple LLMs (GPT-4, Mistral, Llama3)
• Examines prompt engineering variations including template design and shot sampling methods
• Identifies most efficient techniques for accurate HPV vaccine stance detection at scale
• Demonstrates how LLMs can help address medical misinformation through improved content moderation

This research provides critical insights for medical organizations seeking to monitor and address vaccine skepticism and misinformation at scale, enabling more targeted public health interventions using AI-powered content analysis.

Optimizing Social Media Annotation of HPV Vaccine Skepticism and Misinformation Using Large Language Models: An Experimental Evaluation of In-Context Learning and Fine-Tuning Stance Detection Across Multiple Models