Boolean Question Answering on BoolQ contrastive (test)
60.6Accuracy (Avg)DATAMODELS
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DATAMODELSModel=OPT-13B, Source Task=BoolQ2022.12 | 60.6 | — | — | 3.3 | 54.3 | |
| TOPPROMPTSModel=OPT-13B, Source Task=BoolQ2022.12 | 60.3 | — | — | 3.5 | 51 | |
| CONDACCModel=OPT-13B, Source Task=BoolQ2022.12 | 60.1 | — | — | 2.1 | 56.7 | |
| ALLModel=OPT-13B, Source Task=BoolQ2022.12 | 59.8 | — | — | 2.9 | 51.6 | |
| CONDACCModel=GPTJ-6B, Source Task=BoolQ2022.12 | 58.9 | — | — | 1.7 | 54.6 | |
| DATAMODELSModel=GPTJ-6B, Source Task=BoolQ2022.12 | 57.6 | — | — | 1.9 | 54 | |
| TOPPROMPTSModel=GPTJ-6B, Source Task=BoolQ2022.12 | 56.7 | — | — | 2.6 | 49.9 | |
| ALLModel=GPTJ-6B, Source Task=BoolQ2022.12 | 56.6 | — | — | 3 | 50.1 | |
| BERTmode=fine-tuned2021.07 | — | 57.6 | 36.1 | — | — | |
| CLINEmode=fine-tuned2021.07 | — | 63.9 | 47.8 | — | — | |
| RoBERTamode=fine-tuned2021.07 | — | 60.6 | 43.9 | — | — |