Puzzle Solving on Sokoban
43.6Success RateUniform
Evaluation Results
| Method | Links | |
|---|---|---|
| UniformAlgorithm=PPO, Model Scale=0.5B2026.05 | 43.6 | |
| ActFocusAlgorithm=PPO, Model Scale=0.5B2026.05 | 40.4 | |
| ActFocusAlgorithm=PPO, Model Scale=3B2026.05 | 37.5 | |
| UniformAlgorithm=PPO, Model Scale=3B2026.05 | 28.7 | |
| ActFocusAlgorithm=GRPO, Model Scale=0.5B2026.05 | 23.4 | |
| ActFocusAlgorithm=GRPO, Model Scale=3B2026.05 | 23.2 | |
| UniformAlgorithm=GRPO, Model Scale=0.5B2026.05 | 20.7 | |
| UniformAlgorithm=GRPO, Model Scale=3B2026.05 | 17.8 |