Natural Language Inference on QNLI few-shot zero-shot
71.6AccuracyRETROPROMPT
Evaluation results
| Method | Links | |
|---|---|---|
| RETROPROMPTshots=162022.05 | 71.6 | |
| LM-BFFshots=16, demonstration_type=discrete2022.05 | 69.2 | |
| LM-BFFshots=16, demonstration_type=manual2022.05 | 64.6 | |
| RETROPROMPTshots=42022.05 | 62.2 | |
| LM-BFFshots=4, demonstration_type=discrete2022.05 | 61.7 | |
| KPTshots=16, extra_knowledge=true2022.05 | 61.5 | |
| LM-BFFshots=4, demonstration_type=manual2022.05 | 61.1 | |
| Fine-Tuningshots=162022.05 | 60.2 | |
| KPTshots=4, extra_knowledge=true2022.05 | 59.2 | |
| RETROPROMPTshots=02022.05 | 56.2 | |
| KPTshots=0, extra_knowledge=true2022.05 | 55.3 | |
| Fine-Tuningshots=42022.05 | 54.2 | |
| LM-BFFshots=0, demonstration_type=discrete2022.05 | 53.5 | |
| LM-BFFshots=0, demonstration_type=manual2022.05 | 50.5 | |
| Fine-Tuningshots=02022.05 | 49.5 | |
| LOTClassshots=0, training_on_unlabeled=true2022.05 | 36.5 |