Fig. 2: Boxplots comparing the AUC scores distribution on the choice of using mean output pooling, summary-token pooling, or maximum pooling.
From: Benchmarking DNA foundation models for genomic and genetic tasks

We calculate AUC scores from all 52 binary sequence classification datasets included in this study. Boxplots show the median (center line), interquartile range (box = 25th–75th percentiles), and whiskers correspond to 1.5 × interquartile range. minima and maxima correspond to the whisker ends.