ResearchBenchmarksMedical Calculation on MedCalc-Bench Cleaned & restructuredFollow62.7AccuracyGPT-4o52.809655.377357.94560.5127Feb 10, 2026Evaluation ResultsMethodMethodLinksAccuracyGPT-4oApproach=Step-wise eva...Approach=Step-wise eval diagnostic, Setting=Step-wise final-answer2026.0262.7MedRaCApproach=Agentic + ste...Approach=Agentic + step-wise eval, Setting=Retrieval + Python2026.0253.19