Multi-task Classification Average on SST-2, BoolQ, Subj, Scicite, AGNews
72.4Average AccuracyCONDACC
Evaluation Results
| Method | Links | |
|---|---|---|
| CONDACCBackbone=GPTJ-6B2022.12 | 72.4 | |
| DATAMODELSBackbone=GPTJ-6B2022.12 | 72.4 | |
| CONDACCBackbone=OPT-13B2022.12 | 72 | |
| UN-CONDACCBackbone=GPTJ-6B2022.12 | 71.3 | |
| UN-CONDACCBackbone=OPT-13B2022.12 | 69.2 | |
| TOPPROMPTS-5Backbone=GPTJ-6B2022.12 | 69.1 | |
| DATAMODELSBackbone=OPT-13B2022.12 | 69.1 | |
| TOPPROMPTS-10Backbone=GPTJ-6B2022.12 | 67.9 | |
| TOPPROMPTS-10Backbone=OPT-13B2022.12 | 67.2 | |
| ONESHOTBackbone=GPTJ-6B2022.12 | 67 | |
| ALL + CALIBBackbone=OPT-13B2022.12 | 66.2 | |
| ONESHOTBackbone=OPT-13B2022.12 | 66.1 | |
| TOPPROMPTS-5Backbone=OPT-13B2022.12 | 66.1 | |
| ALL + CALIBBackbone=GPTJ-6B2022.12 | 65.5 | |
| UN-TOPPROMPTS-5Backbone=OPT-13B2022.12 | 65.5 | |
| UN-ONESHOTBackbone=GPTJ-6B2022.12 | 65.4 | |
| UN-TOPPROMPTS-5Backbone=GPTJ-6B2022.12 | 65.4 | |
| ALLBackbone=GPTJ-6B2022.12 | 65.2 | |
| RANDOMBackbone=GPTJ-6B2022.12 | 64.7 | |
| UN-ONESHOTBackbone=OPT-13B2022.12 | 64.3 | |
| ALLBackbone=OPT-13B2022.12 | 63.8 | |
| RANDOMBackbone=OPT-13B2022.12 | 62.6 | |
| UN-ALLBackbone=GPTJ-6B2022.12 | 61.8 | |
| UN-ALLBackbone=OPT-13B2022.12 | 58.3 |