Leveraging Advanced Prompting Strategies in Llama-8b for Enhanced Hyperpartisan News Detection

This paper explores advanced prompting strategies for hyperpartisan news detection using the Llama3-8b-Instruct model, an open-source LLM developed by Meta AI. We evaluate zero-shot, few-shot, and Chain-of-Thought (CoT) techniques on two datasets: SemEval-2019 Task 4 and a headline-specific corpus. Collaborating with a political science expert, we incorporate domain-specific knowledge and structured reasoning steps into our prompts, particularly for the CoT approach. Our findings reveal that some prompting strategies work better than others, specifically on LLaMA, depending on the dataset and the task. This unexpected result challenges assumptions about ICL efficacy on classification tasks. We discuss the implications of these findings for In-Context Learning (ICL) in political text analysis and suggest directions for future research in leveraging large language models for nuanced content classification tasks.