Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Single-Agent Spatial Puzzles on Sokoban (In-domain)
Loading...
77.3
Success Rate
MAGE
31.852
43.651
55.45
67.249
Mar 4, 2026
Success Rate
Updated 1mo ago
Evaluation Results
Method
Method
Links
Success Rate
MAGE
Category=Meta-RL
2026.03
77.3
GiGPO
Category=RL
2026.03
71.9
LAMER
Category=Meta-RL
2026.03
68.8
GRPO
Category=RL
2026.03
60.2
Reflexion
Category=In-Context Le...
2026.03
43.8
ReAct
Category=In-Context Le...
2026.03
38.3
A-MEM
Category=Memory-based
2026.03
36.7
Memento
Category=Memory-based
2026.03
33.6
Feedback
Search any
task
Search any
task
