ResearchBenchmarksMathematical Reasoning on AIME 25 (Acc, Len)Follow37.5AccuracyOriginal21.619225.742129.86533.9879Aug 13, 2025Evaluation ResultsMethodMethodLinksAccuracySolution LengthOriginalModel Scale=7BModel Scale=7B2025.0837.514,622LCPOModel Scale=7BModel Scale=7B2025.0836.878,187OriginalModel Scale=1.5BModel Scale=1.5B2025.0822.7115,985LCPOModel Scale=1.5BModel Scale=1.5B2025.0822.237,194