Towards LLM-based Fact Verification on News Claims with a Hierarchical Step-by-Step Prompting Method

Open paper

Benchmark	Model	Rank	Results
fake-news-detection-on-rawfc	HiSS	#2	F1: 53.9
fake-news-detection-on-rawfc	ReAct	#4	F1: 49.8
fake-news-detection-on-rawfc	Standard prompting with articles	#5	F1: 47.9
fake-news-detection-on-rawfc	CoT	#6	F1: 44.4