Long-horizon Reasoning on Checkers Jumping (n=13 to n=16)
100Performance (n=13)Lookahead-Enhanced Atomic Decomposition
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Lookahead-Enhanced Atomic DecompositionModel=GPT-5.2, Voting=true, k=8, h=42026.03 | 100 | 100 | 87 | 100 | |
| Atomic DecompositionModel=GPT-5.2, Voting=true2026.03 | 20 | 100 | 0 | 96 | |
| Atomic DecompositionModel=GPT-5.2, Voting=false2026.03 | 0 | 36 | 0 | 4 |