ResearchBenchmarksGrid-based maze navigation on MazeNamo 10x10 MediumFollow100Success RateLLM-Flax91.6893.849698.16Apr 29, 2026Evaluation ResultsMethodMethodLinksSuccess RateSuccess Rate DeltaTime (s)Path LengthLLM-FlaxRules=LLM-FlaxRules=LLM-Flax2026.04100—0.94916.6Pure FDConfig=Pure FDConfig=Pure FD2026.04100—1.415.7LLM-FlaxConfig=LLM-FlaxConfig=LLM-Flax2026.04100—0.94916.6ManualRules=ManualRules=Manual2026.049641.99719.8ManualConfig=ManualConfig=Manual2026.0496—1.99719.8PLOIConfig=PLOIConfig=PLOI2026.0492—0.8416.5