Question Answering on MetaQA 2-hop (Acc, F1) benchmark leaderboard · SOTA2 Research

Question Answering on MetaQA 2-hop (Acc, F1)

95.9AccuracyRELOOP

Evaluation Results

Method
RELOOP2025.10	95.9	91.1
RELOOP2025.10	93.2	89.7
AdaptiveRAG2025.10	88.2	90.1
Graph-constrained Reasoning (GcR)2025.10	86.7	88.1
Think on Graph (ToG)2025.10	83.2	84.8
Graph-S^32025.10	63.17	76.18
Graph-S^32025.10	53.73	65.6
Falcon-H1-7B-Instruct2025.10	48.3	—
Falcon3-10B-instruct2025.10	43	—
Qwen3-4B-Instruct-25072025.10	42.2	—
Graph-S^32025.10	40.17	55.49
Llama-3.1-8B-Instruct2025.10	37.8	—
KG-Agent2025.10	32.51	44.52
KG-Agent2025.10	30.19	43.95
KG-Agent2025.10	28.41	33.66
RAG/2hop2025.10	10.07	21.65
RAG/2hop2025.10	7.7	21.21
RAG/2hop2025.10	4.5	15.06
RAG/1hop2025.10	2.4	5.98
G-retriever2025.10	2.33	4.8
RAG/1hop2025.10	1.83	7.44
No graph2025.10	1.27	7.77
LightRAG2025.10	0.97	3.71
LightRAG2025.10	0.93	4.33
RAG/1hop2025.10	0.77	1.74
No graph2025.10	0.43	2.64
RAG/3hop2025.10	0.43	2.31
RAG/3hop2025.10	0.17	3.37
RAG/3hop2025.10	0.13	2.21
G-retriever2025.10	0.1	0.77
G-retriever2025.10	0.1	0.83
No graph2025.10	0.07	0.95
No retriever2025.10	0	0.49
ToG2025.10	0	0
LightRAG2025.10	0	0.19
No retriever2025.10	0	1.11
ToG2025.10	0	0.63
No retriever2025.10	0	0.34
ToG2025.10	0	0.12