False Premise Detection on KG-FPQ benchmark leaderboard · SOTA2 Research

False Premise Detection on KG-FPQ

94.44TP RateG-retriever

Evaluation Results

Method	Links
G-retriever2025.04		94.44	99.05	0.95	5.56	97.12	95.24
G-retriever2025.04		88.89	86.67	13.33	11.11	87.89	88.57
G-retriever2025.04		82.22	93.33	6.67	17.78	86.97	83.81
with RAG2025.04		60	86.67	13.33	40	73.97	82.86
Direct Claim2025.04		44.44	73.33	26.67	55.56	59.7	69.2
Direct Claim2025.04		44.44	73.33	26.67	55.56	59.7	69.2
Direct Claim2025.04		44.44	73.33	26.67	55.56	59.7	69.2
with RAG2025.04		37.78	86.67	13.33	62.22	53.97	79.69
with RAG2025.04		33.33	80	20	66.67	48.78	73.33
GraphRAG/ToG2025.04		8.89	93.33	6.67	91.11	16.16	81.27
GraphRAG/ToG2025.04		8.89	93.33	6.67	91.11	16.16	81.27
GraphRAG/ToG2025.04		8.89	93.33	6.67	91.11	16.16	81.27