List your company

Mathematical Reasoning on MATH-500 (Accuracy and Calibration Metrics)

77AccuracyVanilla SimpleTIR
76.89676.92376.9576.977Jan 12, 2026

Evaluation results

MethodLinks
Vanilla SimpleTIRModel=Qwen2.5-3B2026.01770.1510.1930.622
CAR (MSCR)Model=Qwen2.5-3B2026.0176.90.0570.1680.636