Table 5.

Experimental results on the MetaQA dataset.

Model	MetaQA 1-hop		MetaQA 2-hop		MetaQA 3-hop
	Hits@1	F1	Hits@1	F1	Hits@1	F1
KV-MemNN	0.958	—	0.760	—	0.489	—
VRN	0.975	—	0.899	—	0.625	—
SGReader	0.967	0.960	0.807	0.798	0.610	0.580
GRAFT-Net	0.974	0.910	0.948	0.727	0.778	0.561
2HR-DR	0.988	0.973	0.937	0.814	—	—
GlobalGraph	0.990	0.976	0.955	0.830	0.814	0.624
Ours	0.972	0.985	0.968	0.937	0.963	0.910

The best results are indicated in bold values.