<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
  <title>MATRIX · OpenFrontiers</title>
  <link>https://derekwang2002.github.io/ai-intelligence/en/projects/openfrontiers/</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/openfrontiers.xml" rel="self" type="application/rss+xml"/>
  <item>
    <title>OpenFrontiers tests whether agents rediscover or retrieve mathematics</title>
    <link>https://derekwang2002.github.io/ai-intelligence/en/events/ev-20261027-10/</link>
    <guid isPermaLink="false">ev-20261027-10-published</guid>
    <pubDate>Sat, 31 Oct 2026 15:56:40 GMT</pubDate>
    <category>New</category>
    <category>WATCH</category>
    <description>&lt;p&gt;OpenFrontiers publishes a 12,151-problem corpus and a 200-problem benchmark for separating retrieval of known mathematical solutions from closed-book rediscovery. The authors also provide a higher-quality 50-problem subset and Lean 4 statements. On 40 resolved HQ-50 tasks, the best closed-book agent rediscovered at most seven, while live-web success mostly retrieved published solutions.&lt;/p&gt;&lt;p&gt;Recommendation: WATCH&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://arxiv.org/abs/2610.24690&quot;&gt;OpenFrontiers paper&lt;/a&gt; · &lt;a href=&quot;https://huggingface.co/datasets/openfrontiers/OpenFrontiers&quot;&gt;OpenFrontiers dataset&lt;/a&gt;&lt;/p&gt;</description>
  </item>
</channel>
</rss>
