I really think we need an intermediate between conference papers and arxiv, to just evaluate how reproducible/sane the paper is without evaluating whether it is important or not.

Because at this stage I genuinely can't tell if that's a press release, a report in a paper form, or an actual paper.