Maze Solving on Maze Hard
71.8RSRGPT-4 + CoT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-4 + CoTParams=≈1T2026.05 | 71.8 | 6.1 | |
| S-AI-RecursiveParams=<10M2026.05 | 68.4 | 11.2 | |
| TRMParams=7M2026.05 | 62.3 | 15.4 | |
| MiniTransformer+CoTParams=8M2026.05 | 58.3 | 15.3 | |
| Mamba-7MParams=≈7-8M2026.05 | 55.1 | 19.2 | |
| RWKV-7MParams=≈7-8M2026.05 | 52.6 | 19.8 | |
| S-AI-IoT (spatial)Params=2M2026.05 | 51.2 | — | |
| GPT-3.5 (5-shot)Params=175B2026.05 | 38.6 | — |