Sudoku Solving on Sudoku (test)
100AccuracyR-Transformer
Evaluation results
| Method | Links | ||
|---|---|---|---|
| R-Transformer# Parameters=0.211M2026.05 | 100 | — | |
| WONNT=16, interaction functions=trigonometric, interactions=point-wise (N=1), # Parameters=1.58M2026.05 | 100 | — | |
| RRN# Parameters=0.201M2026.05 | 99.8 | — | |
| AKOrNattnT=16, # Parameters=2.98M2026.05 | 99.8 | — | |
| HRM# Parameters=27.28M2026.05 | 99.7 | — | |
| Transformer# Parameters=16.87M2026.05 | 98.6 | — | |
| SAT-Net# Parameters=0.618M2026.05 | 98.3 | — | |
| Sheaf-ADMM (fixed RM)dv=288, Params=1.12M2026.05 | 92.6 | 99.5 | |
| ORDER POLICYOrder Strategy=ORDER POLICY (OURS)2026.06 | 91.8 | 99.21 | |
| Sheaf-ADMM (LoRA)dv=288, Params=4.45M2026.05 | 87.4 | 99.2 | |
| CONFIDENCE (TOP-1)Order Strategy=CONFIDENCE (TOP-1)2026.06 | 82 | 98.56 | |
| ENTROPYOrder Strategy=ENTROPY2026.06 | 81 | 98.53 | |
| MARGINOrder Strategy=MARGIN2026.06 | 80 | 98.38 | |
| HUMAN ORDEROrder Strategy=HUMAN ORDER2026.06 | 76.6 | 96.18 | |
| wd1Inference Steps=2562025.07 | 76.4 | — | |
| wd1Inference Steps=5122025.07 | 62.8 | — | |
| RANDOMOrder Strategy=RANDOM2026.06 | 62.7 | 89.8 | |
| MPNN (PM-LoRA; max)dv=504, Params=4.62M2026.05 | 34.7 | 94.1 | |
| MPNN (CM; max)dv=288, Params=1.72M2026.05 | 19.9 | 91 | |
| d1Inference Steps=256, Evaluation Protocol=reproduced2025.07 | 17.6 | — | |
| d1Inference Steps=256, Evaluation Protocol=reported2025.07 | 16.7 | — | |
| d1Inference Steps=512, Evaluation Protocol=reproduced2025.07 | 16.2 | — | |
| + diffu-GRPOInference Steps=256, Evaluation Protocol=reproduced2025.07 | 16.1 | — | |
| + diffu-GRPOInference Steps=256, Evaluation Protocol=reported2025.07 | 12.9 | — | |
| + diffu-GRPOInference Steps=512, Evaluation Protocol=reproduced2025.07 | 11.7 | — | |
| + diffu-GRPOInference Steps=512, Evaluation Protocol=reported2025.07 | 11 | — | |
| MPNN (PM-LoRA; mean)dv=504, Params=4.62M2026.05 | 10.8 | 85.3 | |
| MPNN (PM-fixed; max)dv=225, Params=1.15M2026.05 | 10.7 | 86.3 | |
| d1Inference Steps=512, Evaluation Protocol=reported2025.07 | 9.5 | — | |
| LLaDA-8B-InstructInference Steps=2562025.07 | 6.7 | — | |
| MPNN (CM; mean)dv=288, Params=1.72M2026.05 | 6.2 | 79.4 | |
| LLaDA-8B-InstructInference Steps=5122025.07 | 5.5 | — | |
| MPNN (PM-fixed; mean)dv=225, Params=1.15M2026.05 | 4 | 74.8 |