<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>My New Hugo Site</title><link>https://AiKiAi-stack.github.io/algolia.json</link><description>Recent content from My New Hugo Site</description><generator>Hugo</generator><language>en</language><managingEditor>xxx@example.com (D-Sketon)</managingEditor><webMaster>xxx@example.com (D-Sketon)</webMaster><copyright>All articles on this blog are licensed under the BY-NC-SA license agreement unless otherwise stated. Please indicate the source when reprinting!</copyright><lastBuildDate>Thu, 09 Jun 2022 20:12:52 +0800</lastBuildDate><atom:link href="https://AiKiAi-stack.github.io/index.xml" rel="self" type="application/rss+xml"/><item><title>中文測試</title><link>https://AiKiAi-stack.github.io/post/chinese/</link><pubDate>Thu, 09 Jun 2022 20:12:52 +0800</pubDate><author>xxx@example.com (D-Sketon)</author><guid>https://AiKiAi-stack.github.io/post/chinese/</guid><description>
<![CDATA[<h1>中文測試</h1><p>Author: D-Sketon(xxx@example.com)</p>
        
          <p>This is a Chinese test post.</p>
<p>善我王上魚、產生資西員合兒臉趣論。畫衣生這著爸毛親可時，安程幾？合學作。觀經而作建。都非子作這！法如言子你關！手師也。</p>
<p>以也座論頭室業放。要車時地變此親不老高小是統習直麼調未，行年香一？</p>
<p>就竟在，是我童示讓利分和異種百路關母信過明驗有個歷洋中前合著區亮風值新底車有正結，進快保的行戰從：弟除文辦條國備當來際年每小腳識世可的的外的廣下歌洲保輪市果底天影；全氣具些回童但倒影發狀在示，數上學大法很，如要我……月品大供這起服滿老？應學傳者國：山式排只不之然清同關；細車是！停屋常間又，資畫領生，相們制在？公別的人寫教資夠。資再我我！只臉夫藝量不路政吃息緊回力之；兒足灣電空時局我怎初安。意今一子區首者微陸現際安除發連由子由而走學體區園我車當會，經時取頭，嚴了新科同？很夫營動通打，出和導一樂，查旅他。坐是收外子發物北看蘭戰坐車身做可來。道就學務。</p>
<p>國新故。</p>
<blockquote>
<p>工步他始能詩的，裝進分星海演意學值例道……於財型目古香亮自和這乎？化經溫詩。只賽嚴大一主價世哥受的沒有中年即病行金拉麼河。主小路了種就小為廣不？</p></blockquote>
<p><em>From <a href="http://www.richyli.com/tool/loremipsum/">亂數假文產生器 - Chinese Lorem Ipsum</a></em></p>
        
        <hr><p>Published on 2022-06-09 at <a href='https://AiKiAi-stack.github.io/'>My New Hugo Site</a>, last modified on 2022-06-09</p>]]></description></item><item><title>Markdown Basic Elements</title><link>https://AiKiAi-stack.github.io/post/main/</link><pubDate>Thu, 09 Jun 2022 20:12:52 +0800</pubDate><author>xxx@example.com (D-Sketon)</author><guid>https://AiKiAi-stack.github.io/post/main/</guid><description>
<![CDATA[<h1>Markdown Basic Elements</h1><p>Author: D-Sketon(xxx@example.com)</p>
        
          <h2 id="markdown-基本元素">
<a class="header-anchor" href="#markdown-%e5%9f%ba%e6%9c%ac%e5%85%83%e7%b4%a0"></a>
Markdown 基本元素
</h2><h3 id="标题">
<a class="header-anchor" href="#%e6%a0%87%e9%a2%98"></a>
标题
</h3><h1 id="h1">
<a class="header-anchor" href="#h1"></a>
H1
</h1><h2 id="h2">
<a class="header-anchor" href="#h2"></a>
H2
</h2><h3 id="h3">
<a class="header-anchor" href="#h3"></a>
H3
</h3><h4 id="h4">
<a class="header-anchor" href="#h4"></a>
H4
</h4><h5 id="h5">
<a class="header-anchor" href="#h5"></a>
H5
</h5><h6 id="h6">
<a class="header-anchor" href="#h6"></a>
H6
</h6><h3 id="强调">
<a class="header-anchor" href="#%e5%bc%ba%e8%b0%83"></a>
强调
</h3><p>Emphasis, aka italics, with <em>asterisks</em> or <em>underscores</em>.</p>
<p>Strong emphasis, aka bold, with <strong>asterisks</strong> or <strong>underscores</strong>.</p>
<p>Combined emphasis with <strong>asterisks and <em>underscores</em></strong>.</p>
<p>Strikethrough uses two tildes. <del>Scratch this.</del></p>
<h3 id="列表">
<a class="header-anchor" href="#%e5%88%97%e8%a1%a8"></a>
列表
</h3><h4 id="definition-list-dl">
<a class="header-anchor" href="#definition-list-dl"></a>
Definition List (dl)
</h4><!-- raw HTML omitted -->
<h4 id="ordered-list-ol">
<a class="header-anchor" href="#ordered-list-ol"></a>
Ordered List (ol)
</h4><ol>
<li>List Item 1</li>
<li>List Item 2</li>
<li>List Item 3</li>
</ol>
<h4 id="unordered-list-ul">
<a class="header-anchor" href="#unordered-list-ul"></a>
Unordered List (ul)
</h4><ul>
<li>List Item 1</li>
<li>List Item 2</li>
<li>List Item 3</li>
</ul>
<h3 id="段落">
<a class="header-anchor" href="#%e6%ae%b5%e8%90%bd"></a>
段落
</h3><p>Lorem ipsum dolor sit amet, <a href="">test link</a> consectetur adipiscing elit. <strong>Strong text</strong> pellentesque ligula commodo viverra vehicula. <em>Italic text</em> at ullamcorper enim. Morbi a euismod nibh. <!-- raw HTML omitted -->Underline text<!-- raw HTML omitted --> non elit nisl. <del>Deleted text</del> tristique, sem id condimentum tempus, metus lectus venenatis mauris, sit amet semper lorem felis a eros. Fusce egestas nibh at sagittis auctor. Sed ultricies ac arcu quis molestie. Donec dapibus nunc in nibh egestas, vitae volutpat sem iaculis. Curabitur sem tellus, elementum nec quam id, fermentum laoreet mi. Ut mollis ullamcorper turpis, vitae facilisis velit ultricies sit amet. Etiam laoreet dui odio, id tempus justo tincidunt id. Phasellus scelerisque nunc sed nunc ultricies accumsan.</p>
        
        <hr><p>Published on 2022-06-09 at <a href='https://AiKiAi-stack.github.io/'>My New Hugo Site</a>, last modified on 2022-06-09</p>]]></description><category>test</category></item><item><title/><link>https://AiKiAi-stack.github.io/post/vlm_testing_toolkit_technical_blog/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><author>xxx@example.com (D-Sketon)</author><guid>https://AiKiAi-stack.github.io/post/vlm_testing_toolkit_technical_blog/</guid><description>
<![CDATA[<h1></h1><p>Author: D-Sketon(xxx@example.com)</p>
        
          <h1 id="vlm-testing-toolkit构建多模态大模型快速诊断工具">
<a class="header-anchor" href="#vlm-testing-toolkit%e6%9e%84%e5%bb%ba%e5%a4%9a%e6%a8%a1%e6%80%81%e5%a4%a7%e6%a8%a1%e5%9e%8b%e5%bf%ab%e9%80%9f%e8%af%8a%e6%96%ad%e5%b7%a5%e5%85%b7"></a>
VLM Testing Toolkit：构建多模态大模型快速诊断工具
</h1><blockquote>
<p>你的 VLM 在 MMMU 上表现如何？在 MathVista 上能解出几何题吗？DocVQA 的 OCR 准确率达标了吗？如果这些问题让你夜不能寐，那么这套工具链或许能帮你找回一些睡眠。</p></blockquote>
<h2 id="项目地址">
<a class="header-anchor" href="#%e9%a1%b9%e7%9b%ae%e5%9c%b0%e5%9d%80"></a>
项目地址
</h2><p><strong>开源代码</strong>：http://bitbucket.iluvatar.ai:7990/users/jiacheng.wan/repos/filehost/browse</p>
<h2 id="定位说明快速诊断而非完整评测">
<a class="header-anchor" href="#%e5%ae%9a%e4%bd%8d%e8%af%b4%e6%98%8e%e5%bf%ab%e9%80%9f%e8%af%8a%e6%96%ad%e8%80%8c%e9%9d%9e%e5%ae%8c%e6%95%b4%e8%af%84%e6%b5%8b"></a>
定位说明：快速诊断，而非完整评测
</h2><p><strong>重要</strong>：这套工具的设计目标是<strong>快速诊断模型能力</strong>，而非替代专业的完整评测。</p>
<h3 id="为什么不做完整评测">
<a class="header-anchor" href="#%e4%b8%ba%e4%bb%80%e4%b9%88%e4%b8%8d%e5%81%9a%e5%ae%8c%e6%95%b4%e8%af%84%e6%b5%8b"></a>
为什么不做完整评测？
</h3><ol>
<li><strong>时间成本过高</strong>：完整评测需要测试数千个样本，如果每个样本都通过模型 API 请求 + 答案匹配，耗时可能长达数小时甚至数天</li>
<li><strong>专业性不足</strong>：专业的评测框架（如 OpenCompass、VLMEvalKit）经过精心优化，测试代码更加鲁棒，评分逻辑更加严谨</li>
<li><strong>场景不同</strong>：完整评测用于论文发布、基准测试；快速诊断用于开发调试、模型选型、问题定位</li>
</ol>
<h3 id="适用场景">
<a class="header-anchor" href="#%e9%80%82%e7%94%a8%e5%9c%ba%e6%99%af"></a>
适用场景
</h3><p>✅ <strong>推荐使用</strong>：</p>
<ul>
<li>模型部署后快速验证基本功能</li>
<li>对比多个模型的特定能力（如 OCR、Grounding）</li>
<li>开发过程中快速迭代测试</li>
</ul>
<p><strong>建议</strong>：先用本工具快速诊断，发现问题后再使用专业评测框架进行完整测试。</p>
<h2 id="引言为什么我们需要这套工具">
<a class="header-anchor" href="#%e5%bc%95%e8%a8%80%e4%b8%ba%e4%bb%80%e4%b9%88%e6%88%91%e4%bb%ac%e9%9c%80%e8%a6%81%e8%bf%99%e5%a5%97%e5%b7%a5%e5%85%b7"></a>
引言：为什么我们需要这套工具
</h2><p>Vision-Language Models (VLMs) 正在以惊人的速度进化。从 GPT-4V 到 Qwen-VL，从 LLaVA 到 InternVL，每个模型都声称在各项基准测试中取得了 SOTA 成绩。然而，当你真正部署一个 VLM 并尝试评估其性能时，会发现一个尴尬的现实：</p>
<p><strong>评测数据的获取、标准化和测试流程，比想象中复杂得多。</strong></p>
<ul>
<li>HuggingFace 上有数百个 VLM 评测数据集，字段命名各不相同</li>
<li>有些数据集需要授权访问，有些图片链接会过期</li>
<li>不同任务类型需要不同的评估指标（IoU、NED、字符串匹配&hellip;）</li>
<li>测试结果需要完整的记录和可追溯性</li>
</ul>
<p>VLM Testing Toolkit 正是为了解决这些工程化痛点而设计。它由三个核心模块组成：</p>
        
        <hr><p>Published on 0001-01-01 at <a href='https://AiKiAi-stack.github.io/'>My New Hugo Site</a>, last modified on 0001-01-01</p>]]></description></item></channel></rss>