ResearchBenchmarksMathematical Reasoning on SVAMP (Accuracy, # Length)Follow59.8AccuracyCoT20.17630.46340.7551.037Feb 4, 2026Evaluation ResultsMethodMethodLinksAccuracyAvg Solution StepsCoT2026.0259.812.1COLARMultiplier=2xMultiplier=2x2026.0254.96.11CoLTSeed count=1Seed count=12026.0248.74.54COLARMultiplier=5xMultiplier=5x2026.0248.42.95CoLTSeed count=2Seed count=22026.0248.36.1SIM-CoT2026.0243.9—Coconut2026.0240.76iCoT2026.0236.40CODI2026.0221.76