Relation Extraction on TACRED v1.0 (5% train)
0.69Micro F1NLI-DeBERTa
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| NLI-DeBERTaMethod paradigm=Indirect Supervision, Backbone architecture=DeBERTa2022.05 | 0.69 | — | — | |
| NLI-DeBERTatraining_data_fraction=5%2021.09 | 0.69 | 0.641 | 0.748 | |
| SURE-BART-large-cnnMethod paradigm=Summarization-based, Backbone architecture=BART-large-cnn2022.05 | 0.653 | — | — | |
| SURE-PEGASUS-largeMethod paradigm=Summarization-based, Backbone architecture=PEGASUS-large2022.05 | 0.649 | — | — | |
| SURE-BART-large-xsumMethod paradigm=Summarization-based, Backbone architecture=BART-large-xsum2022.05 | 0.643 | — | — | |
| NLI-RoBERTaMethod paradigm=Indirect Supervision, Backbone architecture=RoBERTa2022.05 | 0.641 | — | — | |
| NLI-RoBERTatraining_data_fraction=5%2021.09 | 0.641 | 0.604 | 0.683 | |
| SURE-BART-largeMethod paradigm=Summarization-based, Backbone architecture=BART-large2022.05 | 0.638 | — | — | |
| IRE-RoBERTa-largeMethod paradigm=Classification-based, Backbone architecture=RoBERTa-large, Implementation=Re-implemented2022.05 | 0.636 | — | — | |
| KnowPromptMethod paradigm=Indirect Supervision, Implementation=Re-implemented2022.05 | 0.61 | — | — | |
| RECENTMethod paradigm=Classification-based, Implementation=Re-implemented2022.05 | 0.533 | — | — | |
| K-AdapterMethod paradigm=Classification-based, Implementation=Re-implemented2022.05 | 0.516 | — | — | |
| LUKEMethod paradigm=Classification-based, Implementation=Re-implemented2022.05 | 0.516 | — | — | |
| LUKEtraining_data_fraction=5%2021.09 | 0.516 | 0.571 | 0.47 | |
| K-Adaptertraining_data_fraction=5%2021.09 | 0.451 | 0.564 | 0.376 | |
| RoBERTaMethod paradigm=Classification-based, Implementation=Re-implemented2022.05 | 0.418 | — | — | |
| RoBERTatraining_data_fraction=5%2021.09 | 0.418 | 0.528 | 0.346 | |
| SpanBERTMethod paradigm=Classification-based, Implementation=Re-implemented2022.05 | 0.288 | — | — | |
| SpanBERTtraining_data_fraction=5%2021.09 | 0.288 | 0.363 | 0.239 |