Sentiment Classification on IMDb (In-domain)
93.23AccuracyFull
Evaluation results
| Method | Links | |
|---|---|---|
| FullTraining set size=1707, Backbone=RoBERTa-base2022.03 | 93.23 | |
| Manual CADTraining set size=3,414 gold, Backbone=RoBERTa2022.03 | 92.7 | |
| Automatics CADTraining set size=3,414 (1,707 gold + 1,707 auto), Backbone=RoBERTa2022.03 | 91.82 | |
| Dynamic-MRTraining set size=800 (100 gold + 700 auto), Backbone=RoBERTa2022.03 | 91.06 | |
| DynamicTraining set size=800 (100 gold + 700 auto), Backbone=RoBERTa2022.03 | 90.84 | |
| Static + 350 autoTraining set size=400, Backbone=RoBERTa-base, Semi-factuals (n)=3502022.03 | 90.16 | |
| Static + 350 autoTraining set size=400 total (50 gold + 350 auto), Backbone=RoBERTa2022.03 | 90.16 | |
| Static + 1150 autoTraining set size=1200, Backbone=RoBERTa-base, Semi-factuals (n)=11502022.03 | 90.12 | |
| Static + 150 autoTraining set size=200, Backbone=RoBERTa-base, Semi-factuals (n)=1502022.03 | 90.08 | |
| Static + 750 autoTraining set size=800, Backbone=RoBERTa-base, Semi-factuals (n)=7502022.03 | 90.08 | |
| Static + 550 autoTraining set size=600, Backbone=RoBERTa-base, Semi-factuals (n)=5502022.03 | 90.04 | |
| Dynamic-FRTraining set size=800 (100 gold + 700 auto), Backbone=RoBERTa2022.03 | 89.85 | |
| Static + 950 autoTraining set size=1000, Backbone=RoBERTa-base, Semi-factuals (n)=9502022.03 | 89.83 | |
| RRTraining set size=400, Backbone=RoBERTa-base, Augmentation=Random Replacement2022.03 | 89.65 | |
| ALTraining set size=100 gold, Backbone=RoBERTa2022.03 | 88.64 | |
| StaticTraining set size=50, Backbone=RoBERTa-base2022.03 | 88.6 | |
| StaticTraining set size=50 gold, Backbone=RoBERTa2022.03 | 88.6 | |
| DPTraining set size=400, Backbone=RoBERTa-base, Augmentation=Duplication2022.03 | 86.7 |