<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
  <title>MATRIX · WeVisOne-VL</title>
  <link>https://derekwang2002.github.io/ai-intelligence/projects/wevisone-vl/</link>
  <description>AI 模型、Agent、工程与基础设施的重要变化，附一手来源与建议等级。</description>
  <language>zh-CN</language>
  <lastBuildDate>Mon, 02 Nov 2026 16:56:00 GMT</lastBuildDate>
  <atom:link href="https://derekwang2002.github.io/ai-intelligence/feeds/projects/wevisone-vl.xml" rel="self" type="application/rss+xml"/>
  <item>
    <title>Tencent 发布 WeVisOne-VL 统一视觉感知权重</title>
    <link>https://derekwang2002.github.io/ai-intelligence/events/ev-20261023-01/</link>
    <guid isPermaLink="false">ev-20261023-01-published</guid>
    <pubDate>Fri, 23 Oct 2026 16:04:55 GMT</pubDate>
    <category>新事件</category>
    <category>TRIAL</category>
    <description>&lt;p&gt;Tencent 公开 WeVisOne-VL 2B、4B 权重和必需的 960 万参数 pixel decoder，统一支持检测、指代表达定位、计数、point grounding、OCR、分割与 alpha matting。Hugging Face 仓库无需申请并标为 Apache-2.0，但 GitHub 尚未识别出许可证。公开 benchmark 均来自发布方。&lt;/p&gt;&lt;p&gt;建议等级：TRIAL&lt;/p&gt;&lt;p&gt;来源：&lt;a href=&quot;https://github.com/Tencent/WeVisOne-VL&quot;&gt;Tencent WeVisOne-VL&lt;/a&gt; · &lt;a href=&quot;https://huggingface.co/tencent/WeVisOne-VL-4B&quot;&gt;WeVisOne-VL-4B&lt;/a&gt; · &lt;a href=&quot;https://huggingface.co/tencent/WeVisOne-VL-2B&quot;&gt;WeVisOne-VL-2B&lt;/a&gt;&lt;/p&gt;</description>
  </item>
</channel>
</rss>
