ResearchBenchmarksAgentic Task Solving on EHRConFollow79.1Pass@3RetroAgent61.00465.70270.475.098May 8, 2026Evaluation ResultsMethodMethodLinksPass@3Pass@5RetroAgent2026.0579.185.2GiGPO2026.0578.183.3ReACT2026.0577.283.3A³variant=σ-Revealvariant=σ-Reveal2026.0577.279.6A³variant=Vanillavariant=Vanilla2026.057579.6HGPO2026.0571.175.9GSPO2026.0568.975.9rStar2026.0563.370.4LATS2026.0561.770.4