Abductive Event Reasoning on Task 12 (test)
92.6Base AccuracyEnsemble Sonnet + GPT + Gemini
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Ensemble Sonnet + GPT + GeminiConfiguration Type=Ensemble2026.03 | 92.6 | 95.2 | |
| GPT-5.2Configuration Type=Individual Models2026.03 | 91.2 | 94.9 | |
| Gemini 3 FlashConfiguration Type=Individual Models2026.03 | 90.7 | 94.3 | |
| Claude Son. 4.5 ThinkingConfiguration Type=Individual Models2026.03 | 90.4 | 95.2 | |
| SC: Sonnet 3× (θ=0.50)Configuration Type=Self-Consistency, SC voting=majority voting, inclusion threshold (θ)=0.502026.03 | 90.2 | 94.8 | |
| SC: Gemini 5× (θ=0.50)Configuration Type=Self-Consistency, SC voting=majority voting, inclusion threshold (θ)=0.502026.03 | 90.2 | 94.3 |