Knowledge Graph Question Answering on MetaQA 2-hop (test)
100Hits@1TransferNet
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TransferNet2023.03 | 100 | — | |
| GFC2023.03 | 100 | — | |
| T5-small+prologtraining_samples=10002023.03 | 100 | — | |
| TransfernetTraining Paradigm=Supervised task-specific2024.12 | 100 | — | |
| PullNet2023.03 | 99.9 | — | |
| NSM2023.03 | 99.9 | — | |
| NSMTraining Paradigm=Supervised task-specific2024.12 | 99.9 | — | |
| T5-small+prologtraining_samples=5002023.03 | 99.33 | — | |
| UniKGQATraining Paradigm=Supervised task-specific2024.12 | 99 | — | |
| EmbedKGQA2023.03 | 98.8 | — | |
| EmbedKGQATraining Paradigm=Supervised task-specific2024.12 | 98.8 | — | |
| SimGRAGTraining Paradigm=KG-driven RAG without training, Backbone=Llama 3 70B2024.12 | 98.4 | — | |
| T5-small+prologtraining_samples=2502023.03 | 97.77 | — | |
| KELPTraining Paradigm=KG-driven RAG with training, Backbone=Llama 3 70B2024.12 | 96 | — | |
| GraftNet2023.03 | 94.8 | — | |
| KG-GPTTraining Paradigm=KG-driven RAG without training, Backbone=Llama 3 70B, Oracle Entities=true2024.12 | 93.6 | — | |
| G-RetrieverTraining Paradigm=KG-driven RAG with training, Backbone=Llama 3 70B, Oracle Entities=true2024.12 | 87.6 | — | |
| RoE-3BBackbone=Llama-3.2-3B-Instruct, Train Dataset=WebQSP2025.10 | 74.2 | 57.12 | |
| KAPINGTraining Paradigm=KG-driven RAG without training, Backbone=Llama 3 70B2024.12 | 71.2 | — | |
| GNN-RAG-3BBackbone=Llama-3.2-3B-Instruct, Train Dataset=WebQSP2025.10 | 41.7 | 13.94 | |
| Llama 3 70BTraining Paradigm=Pre-trained LLM, Prompting=12-shots2024.12 | 25.2 | — | |
| ChatGPTTraining Paradigm=Pre-trained LLM, Prompting=12-shots2024.12 | 23 | — | |
| G-Retrieval-3BBackbone=Llama-3.2-3B-Instruct, Train Dataset=WebQSP2025.10 | 12.7 | 5.58 | |
| RoG-3BBackbone=Llama-3.2-3B-Instruct, Train Dataset=WebQSP2025.10 | 10.8 | 6 |