Sudoku Puzzle Solving on Sudoku Extreme
98.75AccuracyPTRM
Evaluation Results
| Method | Links | |
|---|---|---|
| PTRM# Params=5M, K (stochastic rollouts)=100, D (depth)=64, sigma (exploration parameter)=0.3, architecture=TRM-MLP2026.05 | 98.75 | |
| TRM# Params=5M, architecture=TRM-MLP2026.05 | 87.4 | |
| Standard TRM, our reproduction# Params=5M, architecture=TRM-MLP2026.05 | 87.28 | |
| HRM# Params=27M2026.05 | 55 |