Maze Solving on PerfectMaze XL (held-out)
14Solved RateADD
Evaluation Results
| Method | Links | |
|---|---|---|
| ADDUpdate=20k, Zero-shot=true2025.06 | 14 | |
| PLR⊥Update=20k, Zero-shot=true2025.06 | 12 | |
| TRACEDUpdate=10k, Zero-shot=true2025.06 | 10 | |
| ACCELUpdate=20k, Zero-shot=true2025.06 | 9 | |
| ADDUpdate=10k, Zero-shot=true2025.06 | 2 | |
| PLR⊥Update=10k, Zero-shot=true2025.06 | 1 | |
| ACCELUpdate=10k, Zero-shot=true2025.06 | 1 | |
| DRUpdate=10k, Zero-shot=true2025.06 | 0 | |
| DRUpdate=20k, Zero-shot=true2025.06 | 0 |