When giving a talk, if something is past ~1990, there's only the journal publication year to give to the audience. When something has only an arxiv year, then that's clearly the appropriate year. However, sometimes things take a long time to get published (refereeing and/or backlogs) and end up making things become anachronistic. One can now have a result that was posted in 2014, then published in 2016, while another paper is posted and published in 2015 that generalizes the paper in 2014.
How do you give coherent and consistent choice of dates in a talk that are clear to the audience? Is there a guideline of which year to cite?
(I'm in mathematics if that is relevant.)