ResearchBenchmarksMathematical Reasoning on GSM8K (first 500 samples)Follow62.8AccuracyStandard-0.6415.8332.348.77Nov 11, 2025Evaluation ResultsMethodMethodLinksAccuracyStandardNumber of Parameters=0.6BNumber of Parameters=0.6B2025.1162.8PonderNumber of Parameters=1.4BNumber of Parameters=1.4B2025.111.8