ResearchBenchmarksMassive Multitask Language Understanding on MMLU (CP Metrics)Follow2.39CPr-valueLLaMA 3B1.9222.04352.1652.2865May 22, 2026Evaluation ResultsMethodMethodLinksCPr-valueCPavgCPLLaMA 3BParameters=3BParameters=3B2026.052.392.552.56Mistral2026.052.32.422.47Phi 3.52026.052.172.332.24Qwen 7BParameters=7BParameters=7B2026.051.942.082.07