I try to save all ArXiv papers via the HTML view, when the HTML view seems legible enough. This mostly works great, and is often a better experience in Reader than the PDF version, but the math comes out pretty terrible. In general, the Reader-parsed version is just kind of worse, since the HTML view is designed to be super-readable already.

I know that parsing is partly just to make highlights tractable, but I use the ArXiv HTML view enough that I figure I'd try to bring it to your attention. Ideally, there could be a special case for the math blocks, where as much formatting was preserved as possible?