Knowledge Base Question Answering on MetaQA 1hop
100Hits@1T5-small+prolog
Evaluation results
| Method | Links | |
|---|---|---|
| T5-small+prologtraining_samples=5002023.03 | 100 | |
| T5-small+prologtraining_samples=10002023.03 | 100 | |
| T5-small+prologtraining_samples=2502023.03 | 98.67 | |
| G-RetrieverTraining Paradigm=KG-driven RAG with training, Backbone=Llama 3 70B, Oracle Entities=true2024.12 | 98.5 | |
| SimGRAGTraining Paradigm=KG-driven RAG without training, Backbone=Llama 3 70B2024.12 | 98 | |
| GFC2023.03 | 97.7 | |
| EmbedKGQA2021.01 | 97.5 | |
| EmbedKGQA2023.03 | 97.5 | |
| TransferNet2023.03 | 97.5 | |
| EmbedKGQATraining Paradigm=Supervised task-specific2024.12 | 97.5 | |
| UniKGQATraining Paradigm=Supervised task-specific2024.12 | 97.5 | |
| TransfernetTraining Paradigm=Supervised task-specific2024.12 | 97.5 | |
| NSM+pTeacher Network=Parallel reasoning2021.01 | 97.3 | |
| NSM+hTeacher Network=Hybrid reasoning2021.01 | 97.2 | |
| NSMTeacher Network=None2021.01 | 97.1 | |
| NSM2023.03 | 97.1 | |
| NSMTraining Paradigm=Supervised task-specific2024.12 | 97.1 | |
| GraftNet2021.01 | 97 | |
| PullNet2021.01 | 97 | |
| SRN2021.01 | 97 | |
| GraftNet2023.03 | 97 | |
| PullNet2023.03 | 97 | |
| KV-Mem2021.01 | 96.2 | |
| KELPTraining Paradigm=KG-driven RAG with training, Backbone=Llama 3 70B2024.12 | 94.7 | |
| KG-GPTTraining Paradigm=KG-driven RAG without training, Backbone=Llama 3 70B, Oracle Entities=true2024.12 | 93.6 | |
| KAPINGTraining Paradigm=KG-driven RAG without training, Backbone=Llama 3 70B2024.12 | 90.8 | |
| ChatGPTTraining Paradigm=Pre-trained LLM, Prompting=12-shots2024.12 | 60 | |
| Llama 3 70BTraining Paradigm=Pre-trained LLM, Prompting=12-shots2024.12 | 56.7 |