<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Reka — USASI news</title>
    <link>https://unitedstatesofamericasuperintelligence.com/companies/reka/</link>
    <atom:link href="https://unitedstatesofamericasuperintelligence.com/companies/reka/feed.xml" rel="self" type="application/rss+xml"/>
    <description>Dated, sourced news items in the USASI catalog that mention Reka. Independent project. Not a United States government website.</description>
    <language>en-us</language>
    <lastBuildDate>Tue, 13 Oct 2026 12:00:00 GMT</lastBuildDate>
    <item>
      <title>Reka releases the Reka Inverse Dynamics Model for labeling video with camera motion under Apache 2.0</title>
      <link>https://unitedstatesofamericasuperintelligence.com/news/reka-releases-inverse-dynamics-model/</link>
      <guid isPermaLink="true">https://unitedstatesofamericasuperintelligence.com/news/reka-releases-inverse-dynamics-model/</guid>
      <pubDate>Tue, 13 Oct 2026 12:00:00 GMT</pubDate>
      <category>research</category>
      <description>Reka published the Reka Inverse Dynamics Model (RIDM) in a research post dated October 2, 2026. The model card describes two compact models, a flow model and a pixel model, that predict camera motion from video and were trained on game renders; Reka describes the aim as labeling real-world footage with actions for interactive world models. The weights and inference code are on Hugging Face under the Apache License 2.0, while the flow model relies on separately obtained RAFT-small optical-flow weights from torchvision. The model card labels the release as preliminary. USASI has not tested these models.</description>
    </item>
  </channel>
</rss>
