id,label,actual,limit,comparator,pass,unit n,Minimum held-out examples,54,50,≥,true,count accuracy,Candidate Accuracy,0.7777777777777778,0.75,≥,true,percent macroF1,Candidate macro-F1,0.7617571707045391,0.72,≥,true,percent regression,Maximum allowed regression,0.037037037037037035,'-0.03,≥,true,delta typo,Typo prediction-flip rate,0.018518518518518517,0.05,≤,true,percent context,Context prediction-flip rate,0.09259259259259259,0.15,≤,true,percent sliceN,Minimum per-language examples,18,15,≥,true,count sliceAcc,Minimum language Accuracy,0.7222222222222222,0.65,≥,true,percent coverage,Paired evaluation / stress coverage,54,54,'=,true,count