List your company

Puzzle Solving on Sokoban

43.6Success RateUniform
16.76823.73430.737.666May 14, 2026

Evaluation results

MethodLinks
UniformAlgorithm=PPO, Model Scale=0.5B2026.0543.6
ActFocusAlgorithm=PPO, Model Scale=0.5B2026.0540.4
ActFocusAlgorithm=PPO, Model Scale=3B2026.0537.5
UniformAlgorithm=PPO, Model Scale=3B2026.0528.7
ActFocusAlgorithm=GRPO, Model Scale=0.5B2026.0523.4
ActFocusAlgorithm=GRPO, Model Scale=3B2026.0523.2
UniformAlgorithm=GRPO, Model Scale=0.5B2026.0520.7
UniformAlgorithm=GRPO, Model Scale=3B2026.0517.8