Symbolic Planning on 4x4 Sudoku
26.6Accuracy (Ngen=128)DISPO
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| DISPObase_model=LLaDA-8B-Instruct, algorithm=DISPO2026.02 | 26.6 | 23.3 | 20 | 23.3 | |
| DISPObase_model=SFT w/ s1k, algorithm=DISPO2026.02 | 26.4 | 25 | 11.7 | 21 | |
| diffu-GRPObase_model=SFT w/ s1k, algorithm=diffu-GRPO2026.02 | 22.1 | 16.7 | 9.5 | 16.1 | |
| diffu-GRPObase_model=LLaDA-8B-Instruct, algorithm=diffu-GRPO2026.02 | 18.4 | 12.9 | 11 | 14.1 | |
| SFT w/ s1kbase_model=SFT w/ s1k2026.02 | 16.5 | 8.5 | 4.6 | 9.9 | |
| LLaDA-8B-Instructbase_model=LLaDA-8B-Instruct2026.02 | 11.7 | 6.7 | 5.5 | 8 |