ResearchBenchmarksMathematical Reasoning on Average (GSM8K, MATH, MMLU-MATH, MATHQA) (test)Follow51.4AccuracyDR+DA31.53636.69341.8547.007Jan 29, 2026Evaluation ResultsMethodMethodLinksAccuracyDEDR+DADepth=32, 0-SHOT=trueDepth=32, 0-SHOT=true2026.0151.43.5DRDepth=32, 0-SHOT=trueDepth=32, 0-SHOT=true2026.0147.62.6DRDepth=16, 0-SHOT=trueDepth=16, 0-SHOT=true2026.0145.74DR+DADepth=16, 0-SHOT=trueDepth=16, 0-SHOT=true2026.0143.42.9LADepth=32, 0-SHOT=trueDepth=32, 0-SHOT=true2026.0139.51LADepth=16, 0-SHOT=trueDepth=16, 0-SHOT=true2026.0132.31