<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
	<channel>
		<title>Computer-Vision on AIBussin — AI applications, systems and books</title>
		<link>https://aibussin.com/categories/computer-vision/</link>
		<description>Recent content in Computer-Vision on AIBussin — AI applications, systems and books</description>
		<generator>Hugo</generator>
		<language>en-US</language>
		
		
		
		
			<lastBuildDate>Sun, 24 May 2026 13:16:50 +0100</lastBuildDate>
		
			<atom:link href="https://aibussin.com/categories/computer-vision/index.xml" rel="self" type="application/rss+xml" />
			<item>
				<title>Thinking in Primitives: Why AI Reasoning Should Learn to Point</title>
				<link>https://aibussin.com/post/primitive/</link>
				<pubDate>Sun, 24 May 2026 13:16:50 +0100</pubDate>
				<guid>https://aibussin.com/post/primitive/</guid>
				<description>&lt;p&gt;&lt;strong&gt;From visual primitives to context-filtered reasoning, grounded verification, and AI movie repair&lt;/strong&gt;&lt;/p&gt;&#xA;&lt;h2 id=&#34;tldr&#34;&gt;TL;DR&lt;/h2&gt;&#xA;&lt;p&gt;This post argues that AI reasoning should not operate over everything it can see, read, or detect. It should operate over the &lt;strong&gt;right primitives for the current task&lt;/strong&gt;.&lt;/p&gt;&#xA;&lt;p&gt;The paper &lt;em&gt;Thinking with Visual Primitives&lt;/em&gt; shows that multimodal models reason better when they can point to visual entities using boxes and points. That solves a &lt;strong&gt;Reference Gap&lt;/strong&gt;: language is often too vague to anchor reasoning to the right part of an image.&lt;/p&gt;</description>
			</item>
	</channel>
</rss>
