Zero-shot Natural Language Understanding on T0 (test)
65.5AccuracyT0-3B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| T0-3BPEFT=true2022.12 | 65.5 | 1 | |
| HINTInput format=Joint2022.12 | 65.4 | 1.1 | |
| HypterInput format=Joint2022.12 | 64.6 | 1 | |
| T0-3BModel replication status=replication2022.12 | 64.4 | 1 | |
| HINTInput format=Split2022.12 | 60.3 | 0.8 | |
| T0-3BPrompting status=No Prompt2022.12 | 57.5 | 0.8 | |
| HypterInput format=Split2022.12 | 56.2 | 0.8 | |
| T0-3BModel replication status=original2022.12 | 54.9 | 1 |