Paraphrase Detection on QQP (dev)
92.7AccuracyMegatron-3.9B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Megatron-3.9Btrained tokens ratio=1, fine-tuning=true2019.09 | 92.7 | — | |
| Megatron-1.3Btrained tokens ratio=1, fine-tuning=true2019.09 | 92.6 | — | |
| XLNettrained tokens ratio=2, fine-tuning=true2019.09 | 92.3 | — | |
| Megatron-336Mtrained tokens ratio=1, fine-tuning=true2019.09 | 92.3 | — | |
| RoBERTatrained tokens ratio=2, fine-tuning=true2019.09 | 92.2 | — | |
| ALBERTtrained tokens ratio=3, fine-tuning=true2019.09 | 92.2 | — | |
| EFLTraining Setting=Full training dataset, Backbone=RoBERTa-large2021.04 | — | 89.2 | |
| EFLTraining Setting=Few-shot with K=8, Backbone=RoBERTa-large2021.04 | — | 67.3 | |
| Fine-tuningTraining Setting=Full training dataset, Backbone=RoBERTa-large2021.04 | — | 89 | |
| Fine-tuningTraining Setting=Few-shot with K=8, Backbone=RoBERTa-large2021.04 | — | 58.8 | |
| LM-BFFTraining Setting=Few-shot with K=8, Backbone=RoBERTa-large2021.04 | — | 68.2 | |
| MajorityTraining Setting=Full training dataset, Backbone=RoBERTa-large2021.04 | — | 0 |