ResearchBenchmarksReasoning on MATH (test) (AUROC/p-value)Follow0.749AUROCTRACED0.746920.747460.7480.74854Mar 11, 2026Evaluation ResultsMethodMethodLinksAUROCp-valueTRACEDBase Model=DeepSeek-R1Base Model=DeepSeek-R12026.030.7490.045LR ProbeBase Model=DeepSeek-R1...Base Model=DeepSeek-R1, Method Status=Runner-up Baseline2026.030.747—