Multi-hop Question Answering on HotpotQA (EM, F1, Precision, Recall) benchmark leaderboard · SOTA2 Research

Multi-hop Question Answering on HotpotQA (EM, F1, Precision, Recall)

60.5EMGfm-Retriever

Evaluation Results

Method	Links
Gfm-Retriever2026.03		60.5	72.9	75.5	74.5
HippoRAG 22026.03		57.3	69.6	70.1	71.5
PropRAG2026.03		57.2	70.1	69.6	71
GFM-RAG2026.03		56	71.8	70.7	70.2
Gfm-Retriever2026.03		55.4	70.4	72.5	72.5
SubgraphRAG2026.03		52.6	65.8	66.2	67.5
GFM-RAG2026.03		51.6	66.9	67.7	68.2
FLARE2026.03		48.7	60.6	61.3	61.8
RAPTOR2026.03		48.2	59.2	59.9	60.3
HippoRAG2026.03		45.7	59.2	59.9	60.3
Adaptive-RAG2026.03		45.5	59.6	60.3	60.7
ColBERTv22026.03		45.5	58.4	59.1	59.5
ColBERTv22026.03		43.4	57.7	58.4	58.8
HippoRAG2026.03		41.8	55	55.7	56
GPT-4o-mini2026.03		40	53.2	55.4	72.2
LightRAG2026.03		36.8	48.3	48.9	49.2
GraphRAG2026.03		35.3	54.6	55.3	55.6
Contriever2026.03		34.9	51.2	51.8	52.2
GTR2026.03		33.8	51.9	52.5	52.9
G-Retriever2026.03		33.2	50.3	50.6	51.5