ResearchBenchmarksMathematical Reasoning on AQUA (val)Follow336Tokens at Best Step (K)aPSF270.72711.361,1521,592.64Apr 8, 2026Evaluation ResultsMethodMethodLinksTokens at Best Step (K)Best StepaPSF2026.043362ZERA2026.045293OPRO2026.045933APE2026.049486CriSPO2026.041,6447ProTeGi2026.041,8968GrIPS2026.041,96810