Text Classification on GoEmotions (ECE, Brier, AUROC, F1)
6.5ECESemantic Softmax
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Semantic SoftmaxModel=Phi-4-mini2026.05 | 6.5 | 58.8 | 75.6 | 25.3 | |
| Semantic SoftmaxModel=Qwen-3-1.7B2026.05 | 6.9 | 59.1 | 76.3 | 26.7 | |
| StandardModel=Phi-4-mini2026.05 | 42.1 | 79.5 | 74.4 | 23.6 | |
| StandardModel=Qwen-3-1.7B2026.05 | 57.4 | 84.2 | 71.2 | 22.9 |