Entity Alignment on DBP15K ZH-EN
98H@1ChatEA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ChatEA2024.02 | 98 | — | 98.4 | |
| MEAformerPara.=14.2M, Iterative=true, Surface Forms=w/ SF2022.12 | 97.3 | 99.8 | 98.3 | |
| AgentEAMethod category=Large Language Models-based2026.04 | 97 | 98.1 | 97.4 | |
| BERT-INT2024.02 | 96.8 | — | 97.7 | |
| EA-Agent2026.04 | 96.67 | 98.2 | 97.1 | |
| MCLEA-NEvaluation Protocol=Unsupervised, Pseudo-alignment seed generation=Name-based (-N)2022.09 | 96 | 99.4 | 0.974 | |
| MCLEA-VEvaluation Protocol=Unsupervised, Pseudo-alignment seed generation=Visual-based (-V)2022.09 | 95.9 | 99.5 | 0.974 | |
| Seg-AlignBackbone=GPT-3.52026.04 | 95.3 | — | 95.3 | |
| Seg-AlignMethod category=Large Language Models-based2026.04 | 95.3 | — | — | |
| ChatEAMethod category=Large Language Models-based2026.04 | 95.1 | 96.5 | 97 | |
| MEAformerPara.=14.2M, Surface Forms (SF)=with, Non-iterative=true2022.12 | 94.8 | 99.3 | 96.5 | |
| MEAformerPara.=14.2M, Iterative=false, Surface Forms=w/ SF2022.12 | 94.8 | 99.3 | 96.5 | |
| MEAformerPara.=14.2M, Iterative=false, Unsupervised Setting=(S) Unsup.2022.12 | 91.7 | 98 | 94.1 | |
| MEAformerPara.=14.2M, Iterative=false, Unsupervised Setting=(V) Unsup.2022.12 | 90.9 | 97.4 | 93.3 | |
| SCMEAMethod category=Knowledge Representation Learning-based2026.04 | 90.8 | 97.9 | 93.5 | |
| ERMC+Iterative=false, Surface Forms=w/ SF2022.12 | 90.3 | 94.6 | 89.9 | |
| SEUEvaluation Protocol=Unsupervised2022.09 | 90 | 96.5 | 0.924 | |
| EASYEvaluation Protocol=Unsupervised2022.09 | 89.8 | 97.9 | 0.93 | |
| LLMEABackbone=ERNIE2026.04 | 89.8 | — | 89.8 | |
| LLMEAMethod category=Large Language Models-based2026.04 | 89.8 | 92.3 | — | |
| SDEAPara.=>110M, Surface Forms (SF)=with, Non-iterative=true2022.12 | 87 | 96.6 | 91 | |
| SDEA2026.04 | 87 | 96.6 | 91 | |
| PMFIteration strategy=Iterative2024.07 | 86.7 | — | 90.8 | |
| EasyEA*Method category=Large Language Models-based2026.04 | 86.3 | 97.7 | — | |
| RAEAEnsemble strategy=pre-weighted strategy2025.12 | 86.28 | 97.07 | 90.3 | |
| Dual-AMNMethod category=Knowledge Representation Learning-based2026.04 | 86.1 | 96.4 | 90.1 | |
| UMAEA*Iteration strategy=Iterative2024.07 | 85.6 | — | 90 | |
| CLEM†Iterative=false, Surface Forms=w/ SF2022.12 | 85.4 | 93.5 | 87.9 | |
| MEAformerPara.=13.7M, Iterative=true, Surface Forms=w/o SF2022.12 | 84.7 | 97 | 89.2 | |
| MEAformer*Iteration strategy=Iterative2024.07 | 84.7 | — | 89.2 | |
| TEA2026.04 | 84.5 | 98.2 | 88.7 | |
| RNMIterative=false, Surface Forms=w/ SF2022.12 | 84 | 91.9 | 87 | |
| PMFIteration strategy=Non-iterative2024.07 | 83.5 | — | 88.4 | |
| RAEAEnsemble strategy=average pooling2025.12 | 82.49 | 95.95 | 87.5 | |
| RREAEvaluation Protocol=Unsupervised2022.09 | 82.2 | 96.4 | — | |
| UMAEA*Iteration strategy=Non-iterative2024.07 | 81.8 | — | 87.7 | |
| BERT-INTPara.=>110M, Surface Forms (SF)=with, Non-iterative=true2022.12 | 81.4 | 83.7 | 82 | |
| BERT-INT2026.04 | 81.4 | 83.7 | 82 | |
| MCLEAIteration strategy=Iterative2024.07 | 81.1 | — | 86.5 | |
| Ours (sparse)L=102020.01 | 80.12 | 87.47 | — | |
| RREAsupervision=semi-supervised2020.08 | 80.1 | 94.8 | 85.7 | |
| MEAformer*Iteration strategy=Non-iterative2024.07 | 80 | — | 86 | |
| AttrGNN2025.12 | 79.6 | 92.93 | 84.5 | |
| RAEAEnsemble strategy=SVM2025.12 | 78 | 94.14 | 84.1 | |
| MRAEAEvaluation Protocol=Unsupervised2022.09 | 77.8 | 83.2 | — | |
| AttrGNN+Iterative=false, Surface Forms=w/ SF2022.12 | 77.7 | 92 | 82.9 | |
| MEAformerPara.=13.7M, Iterative=false, Surface Forms=w/o SF2022.12 | 77.1 | 95.1 | 83.5 | |
| MCLEAIteration strategy=Non-iterative2024.07 | 76.4 | — | 83.4 | |
| EVAIterative Learning (IL)=true2020.09 | 76.1 | 90.7 | 81.4 | |
| MRAEAsupervision=semi-supervised2020.08 | 75.7 | 93 | 82.7 | |
| EVAIteration strategy=Iterative2024.07 | 75 | — | 81 | |
| HyperKA (semi)dimension=75, learning_type=semi-supervised2020.10 | 74.3 | — | 80.5 | |
| TransEdge2025.12 | 73.51 | 91.92 | 80.1 | |
| TransEdgesupervision=semi-supervised, translation-based=true2020.08 | 73.5 | 91.9 | 80.1 | |
| TransEdgedimension=75, learning_type=semi-supervised2020.10 | 73.5 | — | 80.1 | |
| NMN2025.12 | 73.3 | 86.9 | 70.2 | |
| Dual-AMN2025.12 | 73.1 | 92.3 | 79.9 | |
| EVAIterative Learning (IL)=false2020.09 | 72 | 92.5 | 79.3 | |
| JEANSVariant=EDL2020.05 | 71.9 | 89.5 | 79.1 | |
| EVAIteration strategy=Non-iterative2024.07 | 71.9 | — | 79.6 | |
| RREAsupervision=basic2020.08 | 71.5 | 92.9 | 79.4 | |
| MCLEA*Para.=13.2M, Iterative=false, Surface Forms=w/o SF2022.12 | 71.5 | 92.3 | 78.8 | |
| JEANSVariant=SFM2020.05 | 71.3 | 88.5 | 77.3 | |
| RDGCN2026.04 | 70.8 | 84.6 | 74.6 | |
| RDGCNMethod category=Knowledge Representation Learning-based2026.04 | 70.8 | 84.6 | 74.6 | |
| RDGCN2020.01 | 70.75 | 84.55 | — | |
| RDGCN2025.12 | 70.75 | 84.55 | 74.9 | |
| MSNEAIteration strategy=Non-iterative2024.07 | 68.3 | — | 76.2 | |
| EVA*Para.=13.3M, Iterative=false, Surface Forms=w/o SF2022.12 | 68 | 91 | 76.2 | |
| GMNN2020.01 | 67.93 | 78.48 | — | |
| Ours (sparse)L=02020.01 | 67.59 | 87.47 | — | |
| TransEdgesupervision=basic, translation-based=true2020.08 | 65.9 | 90.3 | 74.8 | |
| NAEA2020.01 | 65.01 | 86.73 | — | |
| NAEAsupervision=semi-supervised2020.08 | 65 | 86.7 | 72 | |
| NAEAdimension=75, learning_type=semi-supervised2020.10 | 65 | — | 72 | |
| NAEAIterative Learning (IL)=true2020.09 | 65 | 86.7 | 72 | |
| NAEA2020.05 | 65 | 86.7 | 72 | |
| NAEAIterative=true, Surface Forms=w/o SF2022.12 | 65 | 86.7 | 72 | |
| MSNEAIteration strategy=Iterative2024.07 | 64.8 | — | 72.8 | |
| MMEAIterative Learning (IL)=true2020.09 | 64.7 | 85.8 | — | |
| MRAEAsupervision=basic2020.08 | 63.8 | 88.6 | 73.6 | |
| BOOTEA2020.01 | 62.94 | 84.75 | — | |
| BootEA2025.12 | 62.94 | 84.75 | 70.3 | |
| BootEAsupervision=semi-supervised, translation-based=true2020.08 | 62.9 | 84.7 | 70.3 | |
| BootEAdimension=75, learning_type=semi-supervised2020.10 | 62.9 | — | 70.3 | |
| BOOTEAIterative Learning (IL)=true2020.09 | 62.9 | 84.7 | 70.3 | |
| BootEA2020.05 | 62.9 | 84.7 | 70.3 | |
| BootEAIterative=true, Surface Forms=w/o SF2022.12 | 62.9 | 84.7 | 70.3 | |
| BootEAMethod category=Knowledge Representation Learning-based2026.04 | 62.9 | 84.8 | 70.3 | |
| MMEAIteration strategy=Non-iterative2024.07 | 60.9 | — | 68.5 | |
| RKDEA2025.12 | 60.3 | 87.2 | 70.3 | |
| MSNEA*Para.=14.1M, Iterative=false, Surface Forms=w/o SF2022.12 | 60.1 | 83 | 68.4 | |
| Ψθ₁ = MLPL=02020.01 | 58.53 | 78.04 | — | |
| HyperKADimensions=75, 75, 752020.10 | 57.2 | 86.5 | 67.8 | |
| AliNet2025.12 | 55.23 | 85.21 | 65.7 | |
| AliNetfull version=true2019.11 | 53.9 | 82.6 | 62.8 | |
| AliNetsupervision=basic2020.08 | 53.9 | 82.6 | 62.8 | |
| AliNetDimensions=500, 400, 3002020.10 | 53.9 | 82.6 | 62.8 | |
| ALINETIterative Learning (IL)=false2020.09 | 53.9 | 82.6 | 62.8 | |
| AliNet2020.05 | 53.9 | 82.6 | 62.8 |