ResearchBenchmarksPlanning Task on Star graph task (Starhard) synthetic (test)Follow99.3Sequence-level Exact Match AccuracyDILM-S17.86839.00960.1581.291Jun 8, 2026Evaluation ResultsMethodMethodLinksSequence-level Exact Match AccuracyDILM-S2026.0699.3ILM2026.0699.1DILM-M2026.0695.1ARM2026.0623MDM2026.0621