<?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>ASR on Jiaqi</title>
    <link>https://blog.jiaqiguo.xyz/tags/asr/</link>
    <description>Recent content in ASR on Jiaqi</description>
    <generator>Hugo</generator>
    <language>zh-cn</language>
    <lastBuildDate>Mon, 20 Jul 2026 17:00:00 +0800</lastBuildDate>
    <atom:link href="https://blog.jiaqiguo.xyz/tags/asr/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>把 90 分钟会议塞进一次生成：深入浅出 MOSS Transcribe Diarize</title>
      <link>https://blog.jiaqiguo.xyz/posts/moss-transcribe-diarize/</link>
      <pubDate>Mon, 20 Jul 2026 17:00:00 +0800</pubDate>
      <guid>https://blog.jiaqiguo.xyz/posts/moss-transcribe-diarize/</guid>
      <description>会议转写不只是听清每句话，还要回答谁在何时说了什么。MOSS Transcribe Diarize 用一条 128k 上下文的自回归序列同时生成文字、说话人和时间戳，并以真实录音与可控模拟对话训练。本文拆解它的任务立意、0.9B 开源架构、时间标尺、数据模拟、三项指标与四组实验，也辨清论文尚未测到的时间戳精度。</description>
    </item>
  </channel>
</rss>
