ResearchBenchmarksMedical Reasoning on SuperGPQA MedFollow0.3831AccuracyRLVR0.1328760.1978380.26280.327762Feb 11, 2026Evaluation ResultsMethodMethodLinksAccuracyRLVRScale=8BScale=8B2026.020.3831RLCERScale=8BScale=8B2026.020.365SFTScale=8BScale=8B2026.020.3375RLCERScale=4BScale=4B2026.020.3163RLVRScale=4BScale=4B2026.020.285SFTScale=4BScale=4B2026.020.2825BaseScale=8BScale=8B2026.020.2469BaseScale=4BScale=4B2026.020.1425