Combinatorial Reasoning on 3-SAT
90.9AccuracyPolicy
Evaluation Results
| Method | Links | |
|---|---|---|
| PolicyTraining protocol=Joint training2026.05 | 90.9 | |
| High conf.Training protocol=Joint training2026.05 | 89.8 | |
| BaselineTraining protocol=Joint training2026.05 | 88.8 | |
| MarginTraining protocol=Joint training2026.05 | 85.2 | |
| Oracle policyTraining protocol=Policy-only training2026.05 | 82.3 | |
| PolicyTraining protocol=Policy-only training2026.05 | 76.1 | |
| MarginTraining protocol=Policy-only training2026.05 | 76 | |
| High conf.Training protocol=Policy-only training2026.05 | 75.9 |