ResearchBenchmarksMathematical Reasoning on OmniMathMathematical Reasoning on OmniMathFollow30AccuracyRSD26.25627.22828.229.172May 7, 2026Evaluation resultsMethodMethodLinksAccuracyFLOPsA/C ScoreRSDExternal Models=trueExternal Models=true2026.05308.543.51OursExternal Models=falseExternal Models=false2026.0529.18.243.53STEERExternal Models=falseExternal Models=false2026.0528.39.163.097B OnlyExternal Models=N/AExternal Models=N/A2026.0528.213.562.08SpecReasonExternal Models=falseExternal Models=false2026.052812.82.181.5B OnlyExternal Models=N/AExternal Models=N/A2026.0526.42.729.7