https://hackerpaper.kaipfister.de/p/2001.09136/Collected HN comments and submissions linking this paper.enMon, 02 Nov 2026 11:16:22 GMThttps://news.ycombinator.com/item?id=39996773urn:hackerpapers:mention:2001.09136:39996773Wed, 10 Apr 2024 23:31:24 GMT<p>Top papers claim 99.5% accuracy. Since this is measured on a 10,000-sample evaluation set, there are 50 misclassified results. I&#x27;d love to see a table of them. Back in the day, papers claiming top performance on MNIST would just start to enumerate all of the misclassified test samples, e.g. https://www.researchgate.net/figure/All-of-the-misclassified... from 2002. Newer papers have whittled this down to five misclassified digits, e.g. see §4.4 from https://arxiv.org/pdf/2001.09136v6.pdf for this particular flourish, though this isn&#x27;t quite standard protocol (sample is counted as correct if…</p><p><a href="https://hackerpaper.kaipfister.de/p/2001.09136/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.CV