ResearchBenchmarksMedical Reasoning on PubMedQA (test)Follow58.3AccuracyMixed51.85253.52655.256.874Jun 3, 2026Evaluation ResultsMethodMethodLinksAccuracyMixedTraining Strategy=MixedTraining Strategy=Mixed2026.0658.3BaseTraining Strategy=BaseTraining Strategy=Base2026.0654.2VanillaTraining Strategy=VanillaTraining Strategy=Vanilla2026.0652.1Sub-onlyTraining Strategy=Sub-...Training Strategy=Sub-only2026.0652.1