ResearchBenchmarksMathematical Reasoning on AIME 24-26 (Pass@1 accuracy)Follow20Pass@1 AccuracyPUM9.612.31517.7Jun 5, 2026Evaluation ResultsMethodMethodLinksPass@1 AccuracyPUMSetting=Process-OnlySetting=Process-Only2026.0620PUM+GRPOSetting=Outcome-AnchoredSetting=Outcome-Anchored2026.0617.78Vanilla GRPOSetting=Outcome-AnchoredSetting=Outcome-Anchored2026.0616.67PRM+GRPOSetting=Outcome-AnchoredSetting=Outcome-Anchored2026.0614.44PQMSetting=Process-OnlySetting=Process-Only2026.0613.33PQM+GRPOSetting=Outcome-AnchoredSetting=Outcome-Anchored2026.0610PRMSetting=Process-OnlySetting=Process-Only2026.0610