ResearchBenchmarksGrid-based maze navigation on MazeNamo 15x15 MediumFollow100Success RatePure FD75.76882.05988.3594.641Apr 29, 2026Evaluation ResultsMethodMethodLinksSuccess RateSuccess Rate DeltaTime (s)Path LengthPure FDConfig=Pure FDConfig=Pure FD2026.04100—27.3331.8ManualRules=ManualRules=Manual2026.0496.708.7430.9LLM-FlaxRules=LLM-FlaxRules=LLM-Flax2026.0496.7—9.00731ManualConfig=ManualConfig=Manual2026.0496.7—8.7430.9LLM-FlaxConfig=LLM-FlaxConfig=LLM-Flax2026.0496.7—9.00731PLOIConfig=PLOIConfig=PLOI2026.0476.7—5.0930