Sentiment Classification on SST-5 (train)
52.4AccuracyOPTIMA with μ = 0.1
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| OPTIMA with μ = 0.1Augmentation strategy=OPTIMA, Prior mean (mu)=0.1, Learned dropout probability (p_learned)=0.0625, Backbone=DistilBERT2025.05 | 52.4 | 1.161 | 14.2 | |
| OPTIMA with μ = 0.3Augmentation strategy=OPTIMA, Prior mean (mu)=0.3, Learned dropout probability (p_learned)=0.3, Backbone=DistilBERT2025.05 | 52.4 | 1.086 | 4.6 | |
| Fixed pdrop = 0.04Augmentation strategy=Fixed token-dropout, Dropout probability (p_drop)=0.04, Backbone=DistilBERT2025.05 | 52.2 | 1.18 | 15.4 | |
| BO-Fixed pdrop = 0.3Augmentation strategy=Bayesian Optimization (BO-Fixed), Dropout probability (p_drop)=0.3, Backbone=DistilBERT2025.05 | 52.1 | 1.086 | 4.3 | |
| No AugAugmentation strategy=None, Backbone=DistilBERT2025.05 | 51.6 | 1.24 | 19 | |
| Fixed pdrop = 0.0625Augmentation strategy=Fixed token-dropout, Dropout probability (p_drop)=0.0625, Backbone=DistilBERT2025.05 | 51.6 | 1.162 | 14.3 |