ResearchBenchmarksMathematical Reasoning on AIME24, AIME25, AMC23, MATH-500, Minerva, Olymp. Nemontron-1.5B (average)Follow0.59Distinct-4DAPO0.5380.55150.5650.5785Jun 23, 2026Evaluation ResultsMethodMethodLinksDistinct-4Self BLEULogDetDAPO2026.060.5981.634.47ExTra2026.060.5880.64.98GRPO2026.060.5482.844.29