Natural Language Inference on SNLI-3K (test)
89AccuracyG-DAUG
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| G-DAUGSelection strategy=Random2020.04 | 89 | 78.6 | 17.7 | 20.6 | |
| G-DAUGSelection strategy=Diversity2020.04 | 89 | 79.4 | 19 | 20.5 | |
| G-DAUGSelection strategy=Influence2020.04 | 88.7 | 78.6 | 18 | 20.7 | |
| G-DAUGSelection strategy=Combo2020.04 | 88.7 | 78.7 | 16.7 | 20.5 | |
| ROBERTaSelection strategy=None (Baseline)2020.04 | 88.6 | 77.5 | 17 | 20.2 | |
| Backtranslation2020.04 | 88.1 | 81 | 18.8 | 21.7 |