Factual Probe on Google-RE
97.9P@1DEEPSTRUCT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DEEPSTRUCTmode=zero-shot2022.05 | 97.9 | — | — | — | — | |
| DEEPSTRUCTevaluation protocol=zero-shot2022.05 | 97.9 | — | — | — | — | |
| DEEPSTRUCTmode=multi-task2022.05 | 90.3 | — | — | — | — | |
| DEEPSTRUCTevaluation protocol=multi-task2022.05 | 90.3 | — | — | — | — | |
| Petroni et al., 20202022.05 | 78 | — | — | — | — | |
| LAMA-Oracle2022.05 | 74.3 | — | — | — | — | |
| LAMA2022.05 | 10.5 | — | — | — | — | |
| DEEPEXZero-Shot=true2021.09 | — | 67.8 | 91 | 64.1 | 74.3 | |
| LAMA-OracleOracle=true2021.09 | — | 70.6 | 98.1 | 65.1 | 78 | |
| LAMA-Original2021.09 | — | 16.1 | 1.4 | 14 | 10.5 | |
| ROBERTa-baseArchitecture=RoBERTa, Model Size=Base2023.05 | — | 11.52 | 5.51 | 2.68 | 6.57 | |
| ROBERTa-SE-baseArchitecture=RoBERTa, Model Size=Base, Training Strategy=Self-Evolution (SE)2023.05 | — | 15.16 | 6.35 | 9.61 | 10.37 |