<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>HardFlow: Hard-Constrained Sampling for Flow-Matching Models via Trajectory Optimization · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/2511.08425/</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/2511.08425/feed.xml" rel="self" type="application/rss+xml" /><item><title>HardFlow: Hard-Constrained Sampling for Flow-Matching Models via Trajectory Optimization — mentioned by verdverm</title><link>https://news.ycombinator.com/item?id=49699757</link><guid isPermaLink="false">urn:hackerpapers:mention:2511.08425:49699757</guid><pubDate>Mon, 14 Sep 2026 16:37:41 GMT</pubDate><description>&lt;p&gt;The actual paper: https://arxiv.org/abs/2511.08425v3 &amp;gt; Our key insight is to leverage numerical optimal control to steer the sampling trajectory so that constraints are satisfied precisely at the terminal time. Doesn&amp;#x27;t seem so &amp;quot;fool proof&amp;quot; to me as where the inevitable media spin will take it. Then, how do you know &amp;quot;where&amp;quot; to steer weights? &amp;quot;Safe&amp;quot; has no agreed upon definition&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2511.08425/"&gt;Paper and discussion history&lt;/a&gt;. A mention is not an endorsement.&lt;/p&gt;</description><category>cs.LG</category></item></channel></rss>
