For regression: observed vs predicted scatter plot and per-fold residual boxplots (side by side via patchwork if available, otherwise just the scatter). For classification: confusion matrix heatmap.
Usage
# S3 method for class 'cv.fdata'
plot(x, ...)