<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Improved Transformer for High-Resolution GANs · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/2106.07631/</link><description>Collected HN comments and submissions linking this paper.</description><language>en</language><lastBuildDate>Mon, 02 Nov 2026 11:16:22 GMT</lastBuildDate><atom:link href="https://hackerpaper.kaipfister.de/p/2106.07631/feed.xml" rel="self" type="application/rss+xml" /><item><title>Improved Transformer for High-Resolution GANs — mentioned by godelski</title><link>https://news.ycombinator.com/item?id=44850948</link><guid isPermaLink="false">urn:hackerpapers:mention:2106.07631:44850948</guid><pubDate>Sat, 09 Aug 2025 22:31:25 GMT</pubDate><description>&lt;p&gt;Sure. This was the paper[0]. Here&amp;#x27;s a few more you might find these interesting. Google&amp;#x27;s Transformer GAN[1] (not a transformer at all resolutions). Diffusion-GAN[2] is a hybrid architecture. Remember that technically the GAN process can use any underlying architecture. Arguably you could say some of the training steps in LLMs are GANs. And I think this one is also interesting in a similar respect[3]. Before PWC went down, StyleSAN[4] was the SOTA on FFHQ, but IIRC this doesn&amp;#x27;t change the architecture so it should probably work on all the other architectures too (comes with compute costs, but…&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2106.07631/"&gt;Paper and discussion history&lt;/a&gt;. A mention is not an endorsement.&lt;/p&gt;</description><category>cs.CV</category></item></channel></rss>
