ResearchBenchmarksGrade-school-level math on GSM8KGrade-school-level math on GSM8KFollow4.73Avg LengthPRISM4.71484.81744.925.0226Feb 2, 2026Evaluation resultsMethodMethodLinksAvg LengthTPSPRISMModel=LLaMA-2, Temp=0Model=LLaMA-2, Temp=02026.024.73279.14PRISMModel=LLaMA-3, Temp=0Model=LLaMA-3, Temp=02026.025.11237.41