Emotion Classification on held-out 36-item set
36Routed CountBlack-box 1B
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Black-box 1BModel Scale=1B, System Type=Black-box2026.07 | 36 | 27.8 | 27.8 | — | |
| Black-box 3BModel Scale=3B, System Type=Black-box2026.07 | 36 | 30.6 | 30.6 | — | |
| Black-box 8BModel Scale=8B, System Type=Black-box2026.07 | 36 | 38.9 | 38.9 | — | |
| Black-box RoBERTaSystem Type=Black-box, Model=RoBERTa, Extra Information=trained on full corpus2026.07 | 36 | 11.1 | 11.1 | — | |
| AppraisalExtra Information=given oracle ratings2026.07 | 36 | 36.1 | 36.1 | — | |
| Concept-bottleneck LLMSystem Type=Concept-bottleneck2026.07 | 36 | 22.2 | 22.2 | — | |
| SenticNet 9 lexiconSystem Type=Lexicon2026.07 | 36 | 25 | 25 | — | |
| Fine-tuned 3BModel Scale=3B, System Type=Fine-tuned2026.07 | 28 | 32.1 | 25 | 100 | |
| Fine-tuned 8BModel Scale=8B, System Type=Fine-tuned2026.07 | 25 | 48 | 33.3 | 100 | |
| Gold explication (routed)System Type=Gold explication2026.07 | 24 | 45.8 | 30.6 | 100 | |
| Fine-tuned 1BModel Scale=1B, System Type=Fine-tuned2026.07 | 14 | 35.7 | 13.9 | 100 |