ResearchBenchmarksMathematical Reasoning on GSM8K (PTR adv., ReAct adv., Avg. ∆EM)Follow3PTR Advancement AccuracyPTR2.852.92533.075Apr 5, 2026Evaluation ResultsMethodMethodLinksPTR Advancement AccuracyReAct Advancement AccuracyAverage Error Margin (∆EM)PTRAverage over=4 modelsAverage over=4 models2026.04310.129ReActAverage over=4 modelsAverage over=4 models2026.04310.129