Discriminative Tasks on BIG-bench (13 tasks, val)
58.7AccuracyUD+
Evaluation Results
| Method | Links | |
|---|---|---|
| UD+Scale=XXL, Zero-shot=true2022.11 | 58.7 | |
| UD+Model Scale=XXL, Zero-shot=true2022.11 | 56.7 | |
| UDScale=XXL, Zero-shot=true2022.11 | 55.5 | |
| UDModel Scale=XXL, Zero-shot=true2022.11 | 51.3 | |
| Flan-T5Model Scale=XXL, Zero-shot=true2022.11 | 49.4 | |
| UDScale=XL, Zero-shot=true2022.11 | 48.9 | |
| T0Scale=XXL, Zero-shot=true2022.11 | 47.4 | |
| UDModel Scale=XL, Zero-shot=true2022.11 | 47.3 | |
| T0Model Scale=XXL, Zero-shot=true2022.11 | 45 | |
| T0Scale=XL, Zero-shot=true2022.11 | 44.8 | |
| Flan-T5Model Scale=XL, Zero-shot=true2022.11 | 44.6 | |
| UDModel Scale=Large, Zero-shot=true2022.11 | 44.2 | |
| UDScale=Large, Zero-shot=true2022.11 | 43.5 | |
| T0Model Scale=XL, Zero-shot=true2022.11 | 40.4 | |
| T0Scale=Large, Zero-shot=true2022.11 | 39.6 | |
| Flan-T5Model Scale=Large, Zero-shot=true2022.11 | 39.5 | |
| T0Model Scale=Large, Zero-shot=true2022.11 | 38.9 |