Currently, this relies on moderators and users flagging duplicate articles, which is very labor intensive. The automated software can kill duplicates with identical URLs, but it would take some pretty sophisticated AI to reliably figure out that two articles with entirely different titles are talking about the same topic (and which of the articles is the primary or higher-quality reference).
A key aid would be a disincentive to posting duplicates, e.g. lose rather than gain karma points by doing so.
Many sites are trolling for extra hits by deliberately munging their URLs, either appending random crap to the URL itself (e.g. gamedevdaily.io) or appending fake fragment identifiers (medium, signalvnoise) to evade automatic dupe detection using simple URL duplication.
In the meantime, dupes are "winners" for both the poster (intercepting activity and, thereby, points) and the site (gaining more visibility, since your site appears 2X-4X more in HN than it would otherwise).