<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
  <title>MATRIX · Iris-3B</title>
  <link>https://derekwang2002.github.io/ai-intelligence/en/projects/iris-3b/</link>
  <description>Material changes in AI models, agents, engineering and infrastructure, with primary sources and recommendations.</description>
  <language>en</language>
  <lastBuildDate>Mon, 02 Nov 2026 16:56:00 GMT</lastBuildDate>
  <atom:link href="https://derekwang2002.github.io/ai-intelligence/en/feeds/projects/iris-3b.xml" rel="self" type="application/rss+xml"/>
  <item>
    <title>Iris-3B explores pixel-space diffusion as a general vision learner</title>
    <link>https://derekwang2002.github.io/ai-intelligence/en/events/ev-20261005-10/</link>
    <guid isPermaLink="false">ev-20261005-10-published</guid>
    <pubDate>Sun, 11 Oct 2026 16:02:40 GMT</pubDate>
    <category>New</category>
    <category>TRIAL</category>
    <description>&lt;p&gt;Sperid Labs released Apache-2.0 weights and code for Iris-3B, a 3B pixel-space diffusion model that generates images without a VAE and is fine-tuned for monocular depth and restoration. The model card reports zero-shot depth and 4x restoration results, but evidence remains author-reported and inference requires an NVIDIA GPU plus a separate 4B text encoder for generation.&lt;/p&gt;&lt;p&gt;Recommendation: TRIAL&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://huggingface.co/speridlabs/iris-3b&quot;&gt;Iris-3B&lt;/a&gt; · &lt;a href=&quot;https://arxiv.org/abs/2610.09450&quot;&gt;Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-Tuning&lt;/a&gt;&lt;/p&gt;</description>
  </item>
</channel>
</rss>
