ResearchBenchmarksLanguage Understanding on MMLU (Accuracy Quartiles and Token Costs)Follow80.38Mean Accuracy (±STD)ACTIVE79.610479.810280.0180.2098May 7, 2026Evaluation ResultsMethodMethodLinksMean Accuracy (±STD)Accuracy (Q1)Accuracy (Worst-25%)Token Cost (Select)Token Cost (Train)Token Cost (Test)ACTIVESetup=ACTIVESetup=ACTIVE2026.0580.3879.0878.561.150.161RANDOMSetup=RANDOMSetup=RANDOM2026.0579.6978.477.26—0.160.99NO TRAINSetup=NO TRAINSetup=NO TRAIN2026.0579.6478.4377.23——1.23