Interesting thing is that this is by and large the same as Adversarial Feature Learning (I think…
Ferenc Huszar

I’m an author, and can confirm I agree. It’s definitely not exactly a fair metric (though for *some* reason I can’t help but like it…). I suspect that part of AFL’s number comes from the relatively many revisions we’ve posted (6 of them, and another coming soon), with the bump it gets on arxiv-sanity every time a revision is posted due to it appearing again in the “last week/month/…” lists. Not sure what the best way to normalize for that would be, but if you did, the top 3 posters (140+ saves and 5+ revisions) might not look as impressive compared with, say, the top oral (64 saves and just a single revision).

Show your support

Clapping shows how much you appreciated Jeff Donahue’s story.