Classification on RAS-Eval (task-stratified)
0.975AUROCXGBoost
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| XGBoostFeatures=Pooled SBERT features (mean + max pooling), Class weighting=balanced, Hyperparameters=200 rounds, histogram splitting2026.05 | 0.975 | 0.958 | 0.927 | 0.071 | |
| Random ForestFeatures=Pooled SBERT features (mean + max pooling), Class weighting=balanced, Hyperparameters=200 trees2026.05 | 0.974 | 0.972 | 0.966 | 0.191 | |
| Logistic Reg.Features=Pooled SBERT features (mean + max pooling), Class weighting=balanced, Optimizer=L-BFGS, C=1.02026.05 | 0.917 | 0.92 | 0.865 | 0.128 | |
| Linear SVMFeatures=Pooled SBERT features (mean + max pooling), Class weighting=balanced, Optimizer=SGD, Loss=hinge loss, alpha=10−42026.05 | 0.891 | 0.858 | 0.777 | 0.18 |