https://hackerpaper.kaipfister.de/p/1705.03633/Collected HN comments and submissions linking this paper.enMon, 02 Nov 2026 11:16:22 GMThttps://news.ycombinator.com/item?id=41927778urn:hackerpapers:mention:1705.03633:41927778Wed, 23 Oct 2024 18:13:10 GMT<p>Tokenizers face an odd compute issue. Since they&#x27;re part of the pre-processing pipeline, you can&#x27;t quickly test them out for effectiveness. You have to restart a pretraining run to test downstream effectiveness. Separately, As much as an attention module can do universal nonlinear transformations....I wonder if it makes sense to add specifuc modules for some math primitives as well. I remember that the executor paper [1] (slightly precursor to the attention is allyou need paper) created self contained modules for operations like less than, count, sum and then explicitly orchestrated them in…</p><p><a href="https://hackerpaper.kaipfister.de/p/1705.03633/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.CVhttps://news.ycombinator.com/item?id=14317766urn:hackerpapers:submission:1705.03633:14317766Thu, 11 May 2017 17:23:11 GMT<p>Inferring and Executing Programs for Visual Reasoning</p><p></p><p>1 point; 0 discussion replies on HN. <a href="https://hackerpaper.kaipfister.de/p/1705.03633/">Paper and source history</a>. A submission is not an endorsement.</p>cs.CV