| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Sudoku solving | Sudoku-Extreme (test) | Accuracy100 | 31 | |
| Unconditional Sudoku Generation | Sudoku-Extreme unconditional (100K generated samples) | Validity99.05 | 5 | |
| Sudoku solving | Sudoku-Extreme (held-out) | Accuracy95.2 | 5 | |
| Sudoku Puzzle Solving | Sudoku-Extreme 17-clue puzzles (test) | Puzzle Accuracy97.3 | 5 | |
| Sudoku Solving | Sudoku-Extreme 423K (test) | Exact Match (EM)87.4 | 3 | |
| Symbolic Reasoning | Sudoku-Extreme (test) | Accuracy89.34 | 3 |