ResearchBenchmarksMathematical Reasoning on MATH (test) (Equation-level diversity)Follow0.196Equation-level DiversityUCPO0.137760.152880.1680.18312May 1, 2026Evaluation ResultsMethodMethodLinksEquation-level DiversityUCPO2026.050.196Pass@K2026.050.182GRPO2026.050.173FGRPO2026.050.171Ent-Adv2026.050.159KL-Cov2026.050.156Ent-Reg2026.050.14