Few-shot Relation Extraction on FewRel 1.0
90.3Accuracy (5-way 1-shot)REGRAB
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| REGRAB2022.11 | 90.3 | 94.25 | 84.09 | 89.93 | 89.64 | |
| HyperProto BERT+PretrainBackbone=BERT, Pre-training=true2022.11 | 89.9 | 96.23 | 83.55 | 92.9 | 90.65 | |
| MTB2022.11 | 89.8 | 93.59 | 83.37 | 88.64 | 88.85 | |
| HyperProto BERTBackbone=BERT2022.11 | 84.34 | 93.42 | 77.24 | 88.71 | 85.93 | |
| MLMAN2022.11 | 82.98 | 92.66 | 73.59 | 87.29 | 84.13 | |
| ProtoBERTBackbone=BERT2022.11 | 80.68 | 89.6 | 71.48 | 82.89 | 81.16 | |
| Proto-HATT2022.11 | 76.3 | 90.12 | 64.13 | 83.05 | 78.4 | |
| GNN BERTBackbone=BERT2022.11 | 75.66 | 89.06 | 70.08 | 76.93 | 77.93 | |
| ProtoCNNBackbone=CNN2022.11 | 69.2 | 84.79 | 56.44 | 75.55 | 71.5 | |
| SNAIL2022.11 | 67.29 | 79.4 | 53.28 | 68.33 | 67.08 | |
| GNN CNNBackbone=CNN2022.11 | 66.23 | 81.28 | 46.27 | 64.02 | 64.45 | |
| HyperProto CNNBackbone=CNN2022.11 | 66.05 | 87.31 | 56.74 | 77.87 | 71.99 | |
| Meta Net2022.11 | 64.46 | 80.57 | 53.96 | 69.23 | 67.06 |