Emotional Applicability Classification on EAF human-consensus set (test)
92AccuracyTransformer
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TransformerSupervision Regime=Human-supervised, Backbone=RoBERTa-base2026.01 | 92 | 92 | 92 | |
| TransformerSupervision Regime=Autonomous-supervised (GPT labels), Backbone=RoBERTa-base2026.01 | 85 | 85 | 85 | |
| Logistic RegressionSupervision Regime=Human-supervised, Features=TF–IDF2026.01 | 84 | 84 | 84 | |
| Linear SVMSupervision Regime=Human-supervised, Features=TF–IDF2026.01 | 83 | 83 | 83 | |
| o1Mode=Zero-Shot2026.01 | 55 | 40 | 41 | |
| Always ApplicableType=Heuristic2026.01 | 52 | 34 | 36 | |
| Always Not ApplicableType=Heuristic2026.01 | 48 | 32 | 31 | |
| RandomType=Baseline2026.01 | 47 | 47 | 47 |