Logic Puzzle on Sudoku (Success rate)
99.6Success RateActFocus
Evaluation Results
| Method | Links | |
|---|---|---|
| ActFocusAlgorithm=PPO, Model Scale=1.5B2026.05 | 99.6 | |
| UniformAlgorithm=PPO, Model Scale=1.5B2026.05 | 84.2 | |
| ActFocusAlgorithm=PPO, Model Scale=0.5B2026.05 | 82 | |
| ActFocusAlgorithm=GRPO, Model Scale=1.5B2026.05 | 80.9 | |
| UniformAlgorithm=GRPO, Model Scale=1.5B2026.05 | 70.9 | |
| UniformAlgorithm=PPO, Model Scale=0.5B2026.05 | 16.2 | |
| UniformAlgorithm=GRPO, Model Scale=0.5B2026.05 | 0 | |
| ActFocusAlgorithm=GRPO, Model Scale=0.5B2026.05 | 0 |