<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>Expert Routing on Cognaptus</title>
    <link>https://cognaptus.com/tags/expert-routing/</link>
    <description>Recent content in Expert Routing on Cognaptus</description>
    <generator>Hugo -- 0.145.0</generator>
    <language>en-us</language>
    <lastBuildDate>Sat, 12 Sep 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://cognaptus.com/tags/expert-routing/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Sparse Per Token, Dense Per Batch: XShare Reprices MoE Routing at Inference Time</title>
      <link>https://cognaptus.com/blog/2026-09-12-sparse-per-token-dense-per-batch-xshare-reprices-moe-routing-at-inference-time/</link>
      <pubDate>Sat, 12 Sep 2026 00:00:00 +0000</pubDate>
      <guid>https://cognaptus.com/blog/2026-09-12-sparse-per-token-dense-per-batch-xshare-reprices-moe-routing-at-inference-time/</guid>
      <description>XShare shows that MoE serving efficiency depends on the expert footprint of the whole batch, turning expert budgets into a controllable inference-time throughput and quality decision.</description>
    </item>
  </channel>
</rss>
