2-Way Classification on BEA Shared Task 2 Track 3 (Unseen Answers) 2026 (Evaluation)
0.674QWKMeta-prompt best variant per group
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Meta-prompt best variant per groupselection=best variant per group2026.05 | 0.674 | 86.3 | 14 | |
| Meta-prompt thinking mediumthinking_level=medium2026.05 | 0.654 | 85.3 | 25 | |
| Baseline prompting Gemini 3 Flashmodel=Gemini 3 Flash, mode=baseline prompting2026.05 | 0.598 | 83.5 | 37 | |
| Ensemble prompt tuning with Gemini baselinemethod=ensemble, base=Gemini baseline2026.05 | 0.537 | 80.9 | 42 | |
| SVM + TF-IDF (2-10)gramsclassifier=SVM, features=TF-IDF (2-10)grams2026.05 | 0.52 | 80.6 | 43 | |
| Prompt tuning with synthetic dataaugmentation=synthetic data2026.05 | 0.492 | 78.6 | 44 |