Question Answering on MetaQA 3-hop
100Hits@1TransferNet
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| TransferNetMethod category=Full graph/Retrieval-based2022.10 | 100 | — | — | — | — | |
| TransferNet2023.03 | 100 | — | — | — | — | |
| GFC2023.03 | 100 | — | — | — | — | |
| T5-small+prologtraining_samples=2502023.03 | 100 | — | — | — | — | |
| T5-small+prologtraining_samples=5002023.03 | 100 | — | — | — | — | |
| T5-small+prologtraining_samples=10002023.03 | 100 | — | — | — | — | |
| TransfernetTraining Paradigm=Supervised task-specific2024.12 | 100 | — | — | — | — | |
| TransferNetType=Embedding2026.02 | 100 | — | — | — | — | |
| SQALERMethod category=Full graph/Retrieval-based2022.10 | 99.9 | — | — | — | — | |
| SQALER+GNNMethod category=Full graph/Retrieval-based2022.10 | 99.9 | — | — | — | — | |
| EmQLMethod category=Full graph/Retrieval-based2022.10 | 99.1 | — | — | — | — | |
| UniKGQATraining Paradigm=Supervised task-specific2024.12 | 99.1 | — | — | — | — | |
| NSMTeacher Network=None2021.01 | 98.9 | — | — | — | — | |
| NSM+pTeacher Network=Parallel reasoning2021.01 | 98.9 | — | — | — | — | |
| NSM+hTeacher Network=Hybrid reasoning2021.01 | 98.9 | — | — | — | — | |
| NSMMethod category=Subgraph-based with PRN algorithm2022.10 | 98.9 | — | — | — | — | |
| NSM-distillMethod category=Subgraph-based with PRN algorithm2022.10 | 98.9 | — | — | — | — | |
| REAREVMethod category=Subgraph-based with PRN algorithm2022.10 | 98.9 | — | — | — | — | |
| NSM2023.03 | 98.9 | — | — | — | — | |
| NSMTraining Paradigm=Supervised task-specific2024.12 | 98.9 | — | — | — | — | |
| SimGRAGTraining Paradigm=KG-driven RAG without training, Backbone=Llama 3 70B2024.12 | 97.8 | — | — | — | — | |
| EMBRAGType=LLMs+KGs, Backbone=LLaMA2-Chat-7B2026.02 | 96.3 | — | — | — | — | |
| EmbedKGQA2021.01 | 94.8 | — | — | — | — | |
| EmbedKGQAMethod category=Full graph/Retrieval-based2022.10 | 94.8 | — | — | — | — | |
| EmbedKGQA2023.03 | 94.8 | — | — | — | — | |
| EmbedKGQATraining Paradigm=Supervised task-specific2024.12 | 94.8 | — | — | — | — | |
| EmbedKGQAType=Embedding2026.02 | 94.8 | — | — | — | — | |
| BiNetType=Embedding2026.02 | 94.8 | — | — | — | — | |
| PullNet2021.01 | 91.4 | — | — | — | — | |
| PullNetMethod category=Full graph/Retrieval-based2022.10 | 91.4 | — | — | — | — | |
| PullNet2023.03 | 91.4 | — | — | — | — | |
| PullNetType=Retrieval2026.02 | 91.4 | — | — | — | — | |
| KG-GPTTraining Paradigm=KG-driven RAG without training, Backbone=Llama 3 70B, Oracle Entities=true2024.12 | 88.2 | — | — | — | — | |
| GraftNet2021.01 | 77.7 | — | — | — | — | |
| GraftNetMethod category=Subgraph-based with PRN algorithm2022.10 | 77.7 | — | — | — | — | |
| GraftNet2023.03 | 77.7 | — | — | — | — | |
| GraftNetType=Retrieval2026.02 | 77.7 | — | — | — | — | |
| SRN2021.01 | 75.2 | — | — | — | — | |
| G-RetrieverTraining Paradigm=KG-driven RAG with training, Backbone=Llama 3 70B, Oracle Entities=true2024.12 | 54.9 | — | — | — | — | |
| KV-Mem2021.01 | 48.9 | — | — | — | — | |
| KV-MemMethod category=Full graph/Retrieval-based2022.10 | 48.9 | — | — | — | — | |
| KV-MemType=Embedding2026.02 | 48.9 | — | — | — | — | |
| KAPINGTraining Paradigm=KG-driven RAG without training, Backbone=Llama 3 70B2024.12 | 43 | — | — | — | — | |
| Llama 3 70BTraining Paradigm=Pre-trained LLM, Prompting=12-shots2024.12 | 42.3 | — | — | — | — | |
| ChatGPTType=LLMs2026.02 | 41.6 | — | — | — | — | |
| ChatGPT+CoTType=LLMs2026.02 | 40.1 | — | — | — | — | |
| ChatGPTTraining Paradigm=Pre-trained LLM, Prompting=12-shots2024.12 | 38.7 | — | — | — | — | |
| GraftNetreimplemented=true2019.04 | — | 77.7 | 40.2 | 59.2 | 66.4 | |
| KV-Memreimplemented=true2019.04 | — | 48.9 | 19.5 | 37.6 | 35.2 | |
| PullNetreimplemented=false2019.04 | — | 91.4 | 78.2 | 59.7 | 85.2 | |
| VRNreimplemented=false2019.04 | — | 62.5 | — | — | — |