This research establishes effective baseline approaches for Automated Fact-Checking (AFC) using Large Language Models to verify real-world claims quickly and accurately.

• Evaluates different labeling schemes (binary, three-class, five-class) for optimal fact classification
• Extends traditional verification with structured analysis, verdict classification, and explanation generation
• Demonstrates LLMs' effectiveness in fact-checking across multiple domains
• Creates benchmarks for future automated fact-checking systems

Why it matters: As misinformation proliferates online, automated fact-checking becomes critical for information security, helping organizations quickly verify claims and protect against false information at scale.

Towards Automated Fact-Checking of Real-World Claims: Exploring Task Formulation and Assessment with LLMs