Reasoning on Maze Hard
93.7pass@1 AccuracySR²
Evaluation Results
| Method | Links | |
|---|---|---|
| SR²Params=3.4M2025.10 | 93.7 | |
| Attractor Model# Parameters=27M2026.05 | 93.1 | |
| TRM# Parameters=7M2026.05 | 85.3 | |
| HRMParams=27.3M2025.10 | 74.5 | |
| HRM# Parameters=27M2026.05 | 74.5 | |
| Reflective ModelParams=27.3M2025.10 | 70.8 | |
| Recurrent DepthParams=3.4M2025.10 | 48.4 | |
| Attractor Model# Parameters=7M2026.05 | 46.7 | |
| Block Universal TransformerParams=3.4M2025.10 | 30.4 | |
| TransformerParams=27.3M2025.10 | 0 | |
| Deepseek R1# Parameters=671B2026.05 | 0 | |
| Claude 3.7# Parameters=?2026.05 | 0 | |
| O3-mini-high# Parameters=?2026.05 | 0 | |
| Transformer# Parameters=27M2026.05 | 0 | |
| TRM# Parameters=27M2026.05 | 0 |