<?xml version="1.0" encoding="UTF-8"?>
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9" xmlns:news="http://www.google.com/schemas/sitemap-news/0.9">
  <url>
    <loc>https://api.makesence.top/</loc>
    <changefreq>weekly</changefreq>
    <priority>1.0</priority>
  </url>
  <url>
    <loc>https://api.makesence.top/models</loc>
    <changefreq>weekly</changefreq>
    <priority>0.9</priority>
  </url>
  <url>
    <loc>https://api.makesence.top/pricing</loc>
    <changefreq>weekly</changefreq>
    <priority>0.9</priority>
  </url>
  <url>
    <loc>https://api.makesence.top/news</loc>
    <changefreq>daily</changefreq>
    <priority>0.8</priority>
  </url>
  <url>
    <loc>https://api.makesence.top/docs</loc>
    <changefreq>monthly</changefreq>
    <priority>0.7</priority>
  </url>
  <url>
    <loc>https://api.makesence.top/contact</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
  </url>
  <url>
    <loc>https://api.makesence.top/claude-code</loc>
    <changefreq>monthly</changefreq>
    <priority>0.8</priority>
  </url>
  <url>
    <loc>https://api.makesence.top/cursor</loc>
    <changefreq>monthly</changefreq>
    <priority>0.8</priority>
  </url>
  <url>
    <loc>https://api.makesence.top/cline</loc>
    <changefreq>monthly</changefreq>
    <priority>0.8</priority>
  </url>
  <url>
    <loc>https://api.makesence.top/codex</loc>
    <changefreq>monthly</changefreq>
    <priority>0.8</priority>
  </url>
  <url>
    <loc>https://api.makesence.top/cherry-studio</loc>
    <changefreq>monthly</changefreq>
    <priority>0.7</priority>
  </url>
  <url>
    <loc>https://api.makesence.top/trae</loc>
    <changefreq>monthly</changefreq>
    <priority>0.7</priority>
  </url>
  <url>
    <loc>https://api.makesence.top/windsurf</loc>
    <changefreq>monthly</changefreq>
    <priority>0.7</priority>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_191203_Claude_Code_万字终极教程|全</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Claude Code 万字终极教程|全网最,看完这篇不</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_183807_多模态大模型业界现状、综述及行业应用</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>多模态大模型业界现状、综述及行业应用</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_181817_多模态大模型业界现状、综述及行业应用</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>多模态大模型业界现状、综述及行业应用</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_173748_多模态大模型业界现状、综述及行业应用</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>多模态大模型业界现状、综述及行业应用</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_171527_Claude_Code_万字终极教程|全</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Claude Code 万字终极教程|全网最,看完这篇不</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_170802_RAG、LangChain、Agent_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>RAG、LangChain、Agent 到底有什么关系?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_170039_在Azure上免费创建OpenAI环境并</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>在Azure上免费创建OpenAI环境并避开API调用的</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_163902_Claude_Code_万字终极教程|全</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Claude Code 万字终极教程|全网最,看完这篇不</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_163139_Claude_Code_万字终极教程|全</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Claude Code 万字终极教程|全网最,看完这篇不</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_162413_多模态大模型业界现状、综述及行业应用</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>多模态大模型业界现状、综述及行业应用</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_154851_聊聊Hy3-preview的ChatTe</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>聊聊Hy3-preview的ChatTemplate设计</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_153112_Claude_Code_万字终极教程|全</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Claude Code 万字终极教程|全网最,看完这篇不</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_152339_多模态大模型业界现状、综述及行业应用</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>多模态大模型业界现状、综述及行业应用</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_151505_别被"5分钟写个App"误导:Curso</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>别被"5分钟写个App"误导:Cursor深度实践</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_144343_Token_消耗降低_90%:OpenC</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Token 消耗降低 90%</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_143258_你天天在填的那个__v1_chat_co</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>你天天在填的那个 /v1/chat/completion</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_142535_大语言模型-逻辑能力横评_25-11月榜</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大语言模型-逻辑能力横评 25</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_135222_万字长文解析Agent框架中的上下文管理</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>解析Agent框架中的上下文管理策略</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_134325_如何用AI学会所有东西:基于Obsidi</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何用AI学会所有东西:基于Obsidian+Claud</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_133602_AI_Agent一个月极速入门野路子!(</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>AI Agent一个月极速入门野路子!附入门关键策略+保</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_130110_LLM-based_Agents_sur</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>别让LLM当光杆司令：Agent才是它的手脚和感官</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_125438_AIGC_查重让毕业生崩溃,学生自己写的</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>AIGC 查重让毕业生崩溃,学生自己写的段落被判定 AI</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_125022_当我不再手写代码后,才真正发现编程的价值</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>当我不再手写代码后,才真正发现编程的价值</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_124307_必看:微软166页论文解读_GPT-4V</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>:微软166页论文解读 GPT</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_123103_因制作「奥特曼怀孕」等视频多名博主被罚,</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>因制作奥特曼怀孕等视频多名博主被罚,AI 二创的伦理</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_120724_ICLR'24_大语言模型智能体最新研究</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>ICLR'24 大语言模型智能体最新研究进展</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_115202_GPT-4推理提升1750%!普林斯顿清</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>GPT-4推理提升1750%!普林斯顿清华姚班校友提出全</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_112519_首次成功用CNN自动生成代码:北大研究者</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>首次成功用CNN自动生成代码:北大研究者搞定了炉石传说</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_111041_大模型超详细解读_(目录)</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型超详细解读 (目录)</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_101537_OpenAI罕见发论文:我们找到了AI幻</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>OpenAI罕见发论文:我们找到了AI幻觉的罪魁祸首</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_100825_速度堪比Adam,准确率媲美SGD,还能</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>速度堪比Adam,准确率媲美SGD,还能稳定训练GAN</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_094822_OpenAI_免费开放_ChatGPT_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>OpenAI 免费开放 ChatGPT 搜索,无需注册</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_085342_LoRA:大型语言模型的低秩适配器</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>LoRA:大型语言模型的低秩适配器</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_083928_抖音公开了他们的推荐算法原理,强烈推荐一</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>抖音公开了他们的推荐算法原理,一读</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_082018_图解大模型微调系列之:大模型低秩适配器L</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>图解大模型微调系列之:大模型低秩适配器LoRA原理篇</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_080134_冰毒配方脱口而出,过去时态让GPT-4o</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>冰毒配方脱口而出,过去时态让GPT</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_075520_五万字综述!Prompt-Tuning:</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>五万字综述!Prompt-Tuning:一种新的微调范式</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_074144_如何看待《Science》称大模型「幻觉</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何看待《Science》称大模型幻觉难根除?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_072102_ChatGPT_背后的“功臣”——RLH</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>ChatGPT 背后的“功臣”——RLHF 技术详解</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_071626_给ChatGPT施咒,让它为你制作Ank</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>给ChatGPT施咒,让它为你制作Anki卡片—</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_065313_“评测即科学”:首篇大语言模型评测的综述</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>“评测即科学”:首篇大语言模型评测的综述,一文带你了解大</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_064402_在一张_24_GB_的消费级显卡上用_R</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>在一张 24 GB 的消费级显卡上用 RLHF 微调 2</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_062558_70款ChatGPT插件评测:惊艳的开发</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>70款ChatGPT插件评测</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_061536_可解释性:从频域角度解释卷积解码神经网络</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>可解释性:从频域角度解释卷积解码神经网络的表达瓶颈</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_055826_复旦大学邱锡鹏教授团队:Transfor</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>复旦大学邱锡鹏教授团队:Transformer最新综述</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_055127_Google新作:注意力机制是否真的能够</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>注意力权重≠决策因果，谷歌实验戳破可解释性泡沫</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_054757_非常详细!万字长文带你了解Attenti</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>!带你了解Attention,从MHA到DeepSeek</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_053649_万字长文图解_Harness_工程:从_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>图解 Harness 工程</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_051741_2024年大模型Alignment偏好优</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型Alignment偏好优化技术PPO,DPO, S</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_045802_当我尝试AI写作!玩着玩着就把稿费赚了!</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>当我尝试AI写作!玩着玩着就把稿费赚了!附AI写小说全套</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_045116_深度学习在医疗影像领域涉及可解释性相关的</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>深度学习在医疗影像领域涉及可解释性相关的工作有哪些?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_043616_开源大模型推理引擎现状及常见推理优化方法</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>开源大模型推理引擎现状及常见推理优化方法</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_035946_关于大模型推理的量化算法总结</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>关于大模型推理的量化算法</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_035123_Informer:_一个基于Transf</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Informer: 一个基于Transformer的效率</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_032401_AI工程范式的三次演化:Prompt_E</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>AI工程范式的三次演化:Prompt Engineeri</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_030225_超100篇!CVPR_2020最全GAN</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>超100篇!CVPR 2020GAN生成对抗网络论文汇总!</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_023334_Anthropic一条推文,引发了全球A</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Anthropic指控中国公司蒸馏，翻车考古发现自己是“蒸馏世家”</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_022646_Vibe_Coding_AReaL:零手</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Vibe Coding AReaL</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_015928_重读_ReAct_论文:_LLM_Age</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>重读ReAct发现：多写一行思考，成功率从45%翻到71%</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_013621_GPT系列详解:GPT1-GPT2-GP</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>GPT系列详解:GPT1-GPT2-GPT3</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_012843_(万字长文)图解大模型的推理,理解大模型</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>图解大模型的推理,理解大模型推理过程,理解什么是测试时计</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_011742_大模型推理加速:KV_Cache_和_G</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型推理加速:KV Cache 和 GQA</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_010551_从技术角度讲,ChatGPT的表达能力为</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>从技术角度讲,ChatGPT的表达能力为什么逊色于 Cl</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_001947_无大算力时,作为学生,LLM_还有哪些值</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>无大算力时,作为学生,LLM 还有哪些值得做的研究?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260623_001208_[LLM性能优化]聊聊长文本推理性能优化</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>LLM性能优化聊聊长文本推理性能优化方向</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_185709_“AI图像生成模型”与“关键技术”迭代历</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>4款AI画图模型实测：手部正确率最高仅72%，最低65%</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_184642_如何去除AI写作里的AI味儿</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>AI写作去AI味儿：一个技巧让AIGC检测率从40%降到个位数</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_183810_OpenClaw_Token_永久自由且</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>OpenClaw Token 永久自由且免费!Ollam</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_182422_《ChatGPT_Prompt_Engi</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>《ChatGPT Prompt Engineering</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_181535_生成式推荐是不是一个伪范式</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>生成式推荐是不是一个伪范式?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_175859_Claude_Code对外公开了官方内部</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Claude Code对外公开了官方内部最佳实践!核心贡</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_174152_机器学习中的优化方法</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>机器学习中的优化方法</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_173706_生成对抗网络(GAN)</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>我花十年踩遍GAN的坑，模式崩溃最致命</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_172611_MLLM理解多模态信息的过程比你想象得复</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>MLLM理解多模态信息的过程比你想象得复杂1—</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_171240_大模型百倍推理加速之KV_cache篇</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型百倍推理加速之KV cache篇</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_163811_26年1月开源模型汇总,太卷了,这几个已</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>26年1月开源模型汇总,太卷了,这几个已经追平闭源了!</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_163430_开源大模型与闭源大模型的差距是在缩小还是</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>开源大模型与闭源大模型的差距是在缩小还是在扩大?关键因素</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_162637_中国科学院团队首篇LLM模型压缩综述:细</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>中国科学院团队首篇LLM模型压缩综述</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_161543_如何评价_OpenAI_的超级对话模型_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何评价 OpenAI 的超级对话模型 ChatGPT ?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_155721_Transformer深度剖析</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Transformer训练时间仅为LSTM的1/5，成本降80%</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_154633_薅光天下免费_token,OpenCla</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>薅光天下免费 token,OpenClaw 自由不是梦!</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_153023_claude_code_+_GLM,保姆</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>claude code + GLM,保姆级配置教程</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_152250_抱抱脸:ChatGPT背后的算法——RL</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>抱抱脸:ChatGPT背后的算法—</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_151519_大模型训练:Megatron-Core中</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>实测改三个开关，70B模型训练快30%</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_150800_如果你是大模型算法的面试官,你会问哪些问</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>分布式训练多机比单机慢？面试官揭秘80%的人踩过的坑</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_144200_大模型推理性能如何优化</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型推理性能如何优化?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_141656_2026最夯副业!二胎宝妈在家靠AI写作</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>2026最夯副业!二胎宝妈在家靠AI写作3个月变现2.6</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_135845_大语言模型-编程能力测评_25-10月榜</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>GPT-5 Mini全面碾压Sonnet4.5：大模型编程能力10月榜，新测试方法揭真相</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_134015_能否写出一个能根据自我需要而进行编程的程</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>能否写出一个能根据自我需要而进行编程的程序?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_133347_GPT-4核心技术探秘</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>GPT-4用MoE架构：16个专家分工，训练成本降至1/6</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_132022_【手撕NSA】DeepSeek新作-原生</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>手撕NSADeepSeek新作-原生稀疏注意力-超长文附代码</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_124534_一文带你详细了解:大模型MoE架构(含D</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>一文带你了解:大模型MoE架构含DeepSeek MoE详解</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_124039_混合专家模型_(MoE)_详解</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>7个小模型组队击败700亿参数大模型，推理成本降90%</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_123020_大语言模型LLaMA,_ChatGLM,</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大语言模型LLaMA, ChatGLM, BLOOM 的</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_122341_ReLE中文大模型能力评测榜单(持续更新</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>ReLE中文大模型能力评测榜单持续更新</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_115213_🐰大模型分布式训练篇——从零实现_Ten</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>🐰大模型分布式训练篇——从零实现 Tensor Para</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_114221_他把Claude_Code给玩通关了</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>他把Claude Code给玩通关了</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_113204_涌现是伪科学吗</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>涌现是伪科学吗?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_112735_大语言模型的涌现能力:现象与解释</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大语言模型的涌现能力:现象与解释</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_111226_打开模型Zero-Shot新范式:Ins</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>FLAN实测：19/25任务碾压零样本GPT-3，7倍提升怎么来的</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_110317_DeepSeek_V4.1下月发布500</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>DeepSeek V4.1下月发布500亿融资抢先到位</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_104927_[通俗易读]LLM训练-从显存占用分析到</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>通俗易读LLM训练-从显存占用分析到DeepSpeed</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_104227_拆解追溯_ChatGPT各项能力的起源</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>指令微调不注入新能力，只是解锁ChatGPT已有能力</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_102953_LLM推理量化:FP8_versus_I</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>LLM推理量化:FP8 versus INT8</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_100416_深度解读:Policy_Gradient</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>:Policy Gradient,PPO及PPG</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_094530_LORA微调系列(一):LORA和它的基</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>全参微调7B模型要80GB显存，LoRA凭什么只用5GB？</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_093631_工业界主流大语言模型后训练(Post-T</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Midtrain用600B token训练，代码能力飙升但通用能力不降</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_085052_新人写网文把控不住节奏怎么办</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>拆了200本爆款后我发现：节奏不是快慢，是信息差</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_084345_大语言模型推理性能优化综述</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>32B模型首token从4.3秒到？实测通信优化降38%延迟</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_083618_多模态:模型架构总结</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>多模态:模型架构</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_081024_位置编码之路:SIN->ALiBi->R</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>位置编码之路:SIN-&gt;ALiBi-&gt;RoPE -&gt;PI</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_075207_Harness_Engineering_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>所有人都以为AI越用越笨，其实是你没给它“记忆系统”</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_074056_万字长篇!超全Stable_Diffus</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>我花3年踩遍SD参数坑，从原理到实操一次说清</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_072857_2025_年终总结:LLM_推理系统创新</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>推理吞掉80%算</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_072036_BERT模型蒸馏完全指南(原理_技巧_代</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>300M模型蒸成30M，延迟降90%精度仅掉0.5%</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_071238_API是什么:_一篇讲透API</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>你天天用API，但99%的人理解错了它的本质</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_065603_大模型「幻觉」,看这一篇就够了_|_哈工</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>模型把姚明编成足球运动员</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_064606_LLM强化学习算法演进之路:MC->TD</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>TD比MC快4倍收敛，PPO比TRPO省80%显存</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_064243_从数学的角度来说,强化学习究竟是在干什么</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>我花了3个月调参车还是不会过环岛，一个优化问题而已</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_063432_OpenClaw_Token_终极免费自</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>所有人都以为Ollama免费，其实它比API贵10倍</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_062645_万字长文总结检索增强_LLM</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>RAG vs 纯LLM：一个能翻书，一个靠瞎编，差距大到离谱</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_061639_MCP_(Model_Context_P</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>RAG像盲人摸象，Function Calling累成狗，MCP怎么解决？</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_055847_大模型推理加速调研(框架、方法)</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型瓶颈不是算力，是内存带宽</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_055105_仅凭ai真的能做好复杂项目吗</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>9天重写100万行代码，但AI项目编译通过率仅27%</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_054321_MoE(Mixture-of-Exper</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>实测：MoE用1/2算力撬动4倍参数，效果反超稠密模型</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_053319_如何看待DeepSeek连续多次降价</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>API价格降到打印纸的十分之一：DeepSeek连续降价背后的算盘</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_052937_万字长文理解纯视觉感知算法_——_BEV</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>纯视觉感知NDS达56.9%，BEVFormer用时序信息把速度误差减半</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_052039_旋转式位置编码_(RoPE)_知识总结</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>RoPE实测：rotate_half比稀疏矩阵快10倍，训练省几天</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_050949_GPT-4o_实现真正的多模态大一统了吗</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>GPT-4o 实现真正的多模态大一统了吗?技术难度多高?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_050140_qwen2.5-3.7系列总结</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>我花了三天把Qwen从2.5测到3，发现它会偷看答案了</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_045405_关于_LoRA_的大模型训练工程师面试题</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>实测对比：LoRA显存降至1/5，但知识注入F1低8个点</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_044657_全景解读_LLM_后训练技术</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>全景解读 LLM 后训练技术</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_043359_AI正在吃掉自己的粮食</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>AI正在吃掉自己的粮食，而且粮食越来越馊</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_041857_一文看完多模态:从视觉表征到多模态大模型</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>一文看完多模态:从视觉表征到多模态大模型</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_040951_Kimi_K2_Thinking模型发布</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Kimi K2 Thinking模型发布并开源,该模型哪</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_035346_MCP协议看这篇就够了:3W字长文,从入</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>MCP协议看这篇就够了:3W字长文,从入门到精通,从精通</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_034002_一文讲明白大模型分布式逻辑(从GPU通信</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>一文讲明白大模型分布式逻辑从GPU通信原语到Megatr</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_033213_清华00后揪出AI幻觉元凶:仅0.1%神</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>清华00后揪出AI幻觉元凶：0.1%神经元在“讨好”你</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_031838_用Agent评测思路管理AI_Codin</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>用Agent评测思路管理AI Coding —</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_031404_多模态对齐与融合:综述_Multimod</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>我花了3个月踩的坑：多模态对齐做不对，融合就是和稀泥</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_030026_一文对比4种_RLHF_算法:PPO,_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>PPO烧钱、GRPO翻车？实测4种RLHF算法，省下80%显存的代价是啥</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_025255_[必读]_LLM_应用开发全栈指南</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>必读 LLM 应用开发全栈指南</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_024423_RAG_2.0_深入解读</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>RAG 2.0 深入解读</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_023937_图解大模型训练之:Megatron源码解</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>图解大模型训练之:Megatron源码解读2,模型并行</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_023059_深入浅出完整解析Stable_Diffu</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>U-Net在扩散模型中收敛速度比Transformer快一倍</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_022717_深入浅出完整解析ControlNet核心</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>所有人都以为权重初始化要随机，ControlNet却靠全零卷积实现精细控制</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_021343_DeepSeek-R1_模型发布,性能对</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>DeepSeek-R1 模型发布,性能对标 OpenAI</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_020533_claude_code源码优秀解读整理</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>claude code源码优秀解读整理</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_020042_发现一超级Prompt:直接激活GPT-</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>160行代码4天获2700星，这个prompt让AI突然开窍</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_014330_一文讲懂Agent及主流Agent框架介</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>3天实测9大Agent框架：从原型到生产，我选它</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_013908_2026_年普通人学习_AI_的方法指南</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>2026 年普通人学习 AI 的方法指南!</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_012456_大规模多模态训练的断点续训</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>断点续训最大的坑不是模型权重，而是数据加载状态</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_012132_(万字长文)说说大模型中的推理加速技术</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>GQA砍掉60% KV Cache，精度仅降0.5%，这笔账你会算吗</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_010951_深入浅出完整解析LoRA(Low-Ran</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>仅用3%参数微调大模型，效果几乎不变</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_010231_超全&超细的生成式推荐业界落地实践梳理</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Meta实测CTR跌3个点，生成式推荐到底行不行？</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260622_000509_腾讯高管称今年大部分代码都由_AI_生成</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>腾讯高管称今年大部分代码都由 AI 生成,这会对软件开发</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_234311_为什么至今仍未有任何主流游戏为NPC接入</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>为什么至今仍未有任何主流游戏为NPC接入大语言模型LLM?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_232950_2025年,大模型推理加速技术的学习路线</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>,大模型推理加速技术的学习路线是什么?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_230850_什么是LoRA模型,如何使用和训练LoR</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>什么是LoRA模型,如何使用和训练LoRA模型?你想要的</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_230128_细说复旦大学智能体综述AI-Agent(</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>细说复旦大学智能体综述AI-Agent二更</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_223636_AI编程工具里的_Skill、MCP、W</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>AI编程工具里的 Skill、MCP、Workflow、</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_223240_Claude_独立攻克图论猜想,高德纳对</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Claude 独立攻克图论猜想,高德纳对此震惊发文,反映</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_222451_AI_Agent(或者LLM_Agent</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>AI Agent或者LLM Agent深度讲解—</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_214909_深度拆解_Superpowers_和_g</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>深度拆解 Superpowers 和 gstack</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_212052_垂直领域_Agent_落地:为什么我放弃</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>垂直领域 Agent 落地</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_210250_号外!号外!GPT-4技术细节大揭秘!</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>号外!号外!GPT-4技术细节大揭秘!</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_205448_全套资料下载_|_一文了解边缘AI_(E</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>全套资料下载 | 一文了解边缘AI Edge AI 前沿</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_204408_用AI写小说3个月8k,但90%的人第一</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>用AI写小说月入8k，第一步不是让AI写而是建角色库</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_202721_这波可以,终于有内行人把_GPT-4_说</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>这波可以,终于有内行人把 GPT-4 说透了</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_202236_Github_Copilot_全方位上手</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Github Copilot 全方位上手通关指南</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_200225_大模型算法面经:Function_Cal</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型算法面经:Function Call、MCP、A2A</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_193608_大模型分布式训练并行技术(一)-概述</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型分布式训练并行技术一-概述</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_185142_深度学习的可解释性研究(一)——_让模型</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>深度学习的可解释性研究一—— 让模型具备说人话的能力</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_184816_万字长文概览深度学习的可解释性研究</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>概览深度学习的可解释性研究</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_181541_AI_Infra学习路线</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>AI Infra学习路线</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_180757_Qwen2-VL多模态大模型微调实战(完</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Qwen2-VL多模态大模型微调实战代码</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_180008_推荐一个可交互的_Attention_可</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>推荐一个可交互的 Attention 可视化工具!我的T</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_171939_群魔乱舞:MoE大模型详解</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>群魔乱舞:MoE大模型详解</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_171530_如何正确复现_Instruct_GPT_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何正确复现 Instruct GPT / RLHF?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_170733_盘一盘,2017年Transformer</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>盘一盘,2017年Transformer之后,LLM领域</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_164622_在2020年,你觉得计算机视觉有哪些值得</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>在2020年,你觉得计算机视觉有哪些值得研究的领域?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_162705_理解LLM位置编码:RoPE</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>理解LLM位置编码:RoPE</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_154439_AnimationGPT——一个AIGC</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>AnimationGPT—</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_153117_深入解析Function_Calling</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>深入解析Function Calling、MCP和Ski</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_151238_AI_coding如果真的可以,这算是程</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>AI coding如果真的可以,这算是程序员是革了自己的</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_145345_深度学习在医学图像处理中的应用</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>深度学习在医学图像处理中的应用</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_144254_基于大模型_+_知识库的_Code_Re</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>基于大模型 + 知识库的 Code Review 实践</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_142921_[大模型_28]_deepseek-r1</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>我花了三周复现R1：跳过SFT推理混乱，加上冷启动才稳住</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_142308_故事熔炉(StoryForge)——一款</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>故事熔炉StoryForge—</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_133159_基于大语言模型构建知识问答系统</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>基于大语言模型构建知识问答系统</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_132758_Claude_Code在通用Agent上</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Claude Code在通用Agent上的探索Anthr</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_131531_【多模态大模型】Qwen2-VL解剖</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>多模态大模型Qwen2-VL解剖</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_130512_Anthropic年收300亿,碾压Op</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Anthropic年收300亿,碾压OpenAI,为什么</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_123928_一文辨析清楚LORA、Prompt_Tu</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>一文辨析清楚LORA、Prompt Tuning、P</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_122811_2026年最新_ChatGPT_Imag</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>2026年最新 ChatGPT Images 2.0 使</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_120948_Claude_Code_最佳实践指南</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Claude Code 最佳实践指南</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_093945_一篇文章讲清大语言模型发展史</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>一篇文章讲清大语言模型发展史</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_093213_大模型推理加速:KV_Cache_Spa</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型推理加速:KV Cache Sparsity稀疏化方法</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_091910_LLM_推理提速:Attention与F</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>LLM 推理提速:Attention与FFN分离AFD方</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_082915_如何看待_Anthropic_发布的_C</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何看待 Anthropic 发布的 Claude 4</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_081254_[ICLR2024]_LLM安全_隐私_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>ICLR2024 LLM安全/隐私/越狱/幻觉相关论文合辑</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_075539_深度学习与医学图像分析</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>我花了两周跑遍主流医学图像模型，发现没一个省心</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_074101_Claude_Code_+_OpenSp</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>我被AI坑了两个月才明白：先定规范再动手，效率翻倍</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_072657_通向AGI之路:大型语言模型(LLM)技</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>通向AGI之路:大型语言模型LLM技术精要</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_071243_大模型微调(finetune)方法总结</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型微调finetune方法</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_070522_中科院一区顶刊_|_DilateForm</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>中科院一区顶刊 | DilateFormer</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_065132_OpenAI首次揭秘GPT训练细节</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>OpenAI揭秘训练细节，但最值钱的数据配方没给</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_062044_从自编码器到生成对抗网络:一文纵览无监督</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>从自编码器到生成对抗网络:一文纵览无监督学习研究现状</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_060015_如何评价_Anthropic_发布的_C</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何评价 Anthropic 发布的 Claude Fa</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_054825_靠AI写作额外收入几W是什么体验!小白怎</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>靠AI写作额外收入几W是什么体验!小白怎么入门AI写作挣</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_053632_从Claude_Code入手看Agent</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>从Claude Code入手看Agent框架设计思路</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_045430_通俗解读大模型微调(Fine_Tunin</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>通俗解读大模型微调Fine Tuning</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_043248_大模型SFT数据精选方法串讲:IFD、S</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型SFT数据精选方法串讲</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_042136_深入解析多模态大模型-主要技术和最新发展</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>深入解析多模态大模型-主要技术和最新发展综述</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_035720_Transformer学习笔记二:Sel</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Transformer学习二</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_034950_从零训练一个多模态LLM:预训练+指令微</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>从零训练一个多模态LLM:预训练+指令微调+对齐+融合多</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_033108_智能风控中的机器学习</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>智能风控中的机器学习</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_031550_Anthropic_被曝_300_亿美元</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Anthropic 被曝 300 亿美元融资,投前估值超</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_031105_LLM大语言模型之Generate_In</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>LLM大语言模型之Generate/Inference生</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_030330_再来聊聊强化学习在自动驾驶中的应用</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>再来聊聊强化学习在自动驾驶中的应用</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_025605_LLM为什么宁可瞎编也不说“我不知道”</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>LLM为什么宁可瞎编也不说“我不知道”?一个信息论的回答</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_024829_LLM训练-pretrain</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>LLM训练-pretrain</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_024151_试完这个神级Prompt,我发现Clau</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>我花了两天实测李继刚的汉语新解Prompt，证实Claude 3.5确实断层领先</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_022654_大模型创业:两种极端与一种经验</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型创业:两种极端与一种经验</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_022134_终于学会用ChatGPT写小说了!</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>终于学会用ChatGPT写小说了!</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_021340_在中国内地稳定使用_Claude:从注册</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>在中国内地稳定使用 Claude:从注册到订阅的方案</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_020646_arxiv_新论文解释语言模型产生幻觉的</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>arxiv 新论文解释语言模型产生幻觉的原因,为什么LL</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_020220_有没有来锐评一下大模型各大研究方向的</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>有没有来锐评一下大模型各大研究方向的?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_015504_【深度学习数学基础】序章_+_目录(已完</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>深度学习数学基础序章 + 已完结,共30章</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_013944_万字长文图解_Claude_Code_剖</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>图解 Claude Code 剖析源码</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_013531_国外大神逆向了_Claude_Code,</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>国外大神逆向了 Claude Code,发现它好用的秘密</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_012800_如何看待Anthropic公司在Chat</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何看待Anthropic公司在ChatGPT4.5推出</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_012410_GPT等大模型的“涌现”能力是玄学吗</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>GPT等大模型的“涌现”能力是玄学吗?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_010113_多模态大模型的训练策略对比</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>多模态大模型的训练策略对比</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_005553_【多模态大模型】Qwen2.5-VL解剖</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>多模态大模型Qwen2.5-VL解剖</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_004859_AI绘画技术的背景、发展、应用及其对艺术</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>AI绘画技术的背景、发展、应用及其对艺术创作、社会与文化</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_003415_御三家大模型横评(Claude,_Gem</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>御三家大模型横评Claude, Gemini, GPT和</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_002945_大模型API订阅套餐、API费用对比:O</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型API订阅套餐、API费用对比</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_002122_如何系统入门大模型微调并进行相关的实践</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何系统入门大模型微调并进行相关的实践?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_001235_要研究深度学习的可解释性(Interpr</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>要研究深度学习的可解释性Interpretability</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260621_000022_72_页_PPT,带你梳理神经网络完整架</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>72 页 PPT,带你梳理神经网络架构含 PyTorch</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_235039_【深度学习-第2篇】CNN卷积神经网络3</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>深度学习-第2篇CNN卷积神经网络30分钟入门!足够通俗</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_233640_本周最火AutoGPT!GitHub3.</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>本周最火AutoGPT!GitHub3.6万星,解决复杂</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_233311_如何看本周最火的AutoGPT</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何看本周最火的AutoGPT?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_232519_大模型结构的进化(一):LLaMA_3.</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型结构的进化一:LLaMA 3.1结构及影响解析</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_231741_LORA:大模型轻量级微调</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>LORA:大模型轻量级微调</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_230740_深度好文!最全的大模型_RAG_技术概览</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>深度好文!的大模型 RAG 技术概览</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_230251_陈巍:ChatGPT大模型技术精要——发</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>陈巍:ChatGPT大模型技术精要—</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_225459_如何评价AI检测系统将《荷塘月色》判为6</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何评价AI检测系统将《荷塘月色》判为62.88%疑似A</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_225028_大家如何看待用_AI_写小说</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大家如何看待用 AI 写小说?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_224315_基于_MCP_的_AI_Agent_应用</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>基于 MCP 的 AI Agent 应用开发实践</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_223833_【LLM技术报告】DeepSeek-V3</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>LLM技术报告DeepSeek-V3技术报告全文</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_222536_LLM_in_Medical_Domai</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>LLM in Medical Domain</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_221827_包教包会!从零实现基于Transform</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>包教包会!从零实现基于Transformer的语音识别A</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_221342_基于PyTorch,用搭积木的方式实现完</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>基于PyTorch,用搭积木的方式实现的Transfor</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_215534_GPT___GPT-2___GPT-3_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>GPT / GPT-2 / GPT</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_214752_大模型输出答案前「思考」的「思维链」机制</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型输出答案前思考的思维链机制,本质上是何种原理?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_214133_深度学习算法工程师面经(微软、阿里、商汤</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>深度学习算法工程师面经微软、阿里、商汤、旷视、滴滴、华为</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_212828_Claude_Code_还是_Codex</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Claude Code 还是 Codex?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_212037_利用GPT创作学术论文的过程记录</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>利用GPT创作学术论文的过程记录</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_211353_Claude_Code__源码里有意思设</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Claude Code 源码里有意思设定</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_210403_目标检测最新方向:推翻固有设置,不再一成</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>目标检测最新方向:推翻固有设置,不再一成不变Anchor</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_210000_NLP算法面试必备!PTMs:NLP预训</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>NLP算法面试必备!PTMs:NLP预训练模型的</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_204954_大模型微调总结</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型微调总结</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_204339_多模态大模型_CLIP,_BLIP,_B</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>多模态大模型 CLIP, BLIP, BLIP2, LL</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_203610_英伟达把GPT-4塞进我的世界,打游戏快</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>英伟达把GPT-4塞进我的世界,打游戏快15倍</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_202616_深度学习实现缺陷检测算法汇总</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>深度学习实现缺陷检测算法汇总</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_201918_初始化图像:扩散模型中被忽视的控制机制</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>初始化图像:扩散模型中被忽视的控制机制</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_200922_大语言模型推理加速:硬件视角的全面解析</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大语言模型推理加速:硬件视角的解析</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_200137_大模型推理夯实:并行策略图解</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型推理夯实:并行策略图解</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_195308_主流大语言模型从预训练到微调的技术原理</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>主流大语言模型从预训练到微调的技术原理</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_194804_预训练大语言模型的三种微调技术总结:fi</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>预训练大语言模型的三种微调技术</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_194113_【万字长文】LLaMA,_ChatGLM</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>LLaMA, ChatGLM, BLOOM的参数高效微调实践</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_193056_DeepSeek-R1_技术报告解读</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>DeepSeek-R1 技术报告解读</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_192141_杜克大学最新《可解释机器学习》综述论文,</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>杜克大学最新《可解释机器学习》综述论文,80页pdf阐述</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_190140_万字长文!大模型(LLM)推理优化技术总</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>!大模型LLM推理优化技术</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_185442_大模型的幻觉问题调研:_LLM_Hall</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型的幻觉问题调研: LLM Hallucinatio</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_184412_ChatGPT中的提示工程(Prompt</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>ChatGPT中的提示工程Prompt怎么做?DAIR.</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_183941_Transformer_&_Bert_相</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Transformer &amp; Bert 相关问题复盘及</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_183249_一文带你熟悉lora微调各类参数,轻松上</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>一文带你熟悉lora微调各类参数,轻松上手deepsee</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_182832_大模型参数高效微调(PEFT)</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>LoRA微调参数少99.6%，效果反超全量微调</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_182213_垂直领域大模型的思考</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>垂直领域大模型的思考</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_181821_大模型推理加速技术的学习路线是什么</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型推理加速技术的学习路线是什么?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_181050_Mixture_of_Experts(M</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Mixture of ExpertsMoE学习</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_180502_大模型思维链(Chain-of-Thou</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型思维链Chain-of-Thought技术原理</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_175639_SFT、RLHF、DPO、IFT_——_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>SFT、RLHF、DPO、IFT —</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_175154_真·万字长文:可能是全网最晚的chatg</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>真·:可能是全网最晚的chatgpt技术</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_174440_Claude_Code_完整源码泄露了,</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Claude Code 源码泄露了,花了一天读完全部源码</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_173738_[大模型_40]_张祥雨访谈_多模态大模</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型 40 张祥雨访谈 多模态大模型研究的挣扎史和未来</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_173026_OpenAI_开放_GPT-3.5_Tu</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>OpenAI 开放 GPT</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_171616_陈巍:GPT-4核心技术分析报告(2)—</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>陈巍:GPT-4核心技术分析报告2——GPT</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_170337_RLHF_对齐之_REINFORCE++</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>RLHF 对齐之 REINFORCE++ 算法</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_165047_大模型参数高效微调技术原理综述(一)-背</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型参数高效微调技术原理综述一-背景、参数高效微调简介</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_164615_为什么基于_Transformer_的大</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>GPU高效矩阵乘法详解：Transformer凭什么碾压LSTM 3000倍效率差</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_163856_从Claude_Code入手看Agent</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>从Claude Code入手看Agent框架设计思路基础篇</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_163238_国内外知名大模型及应用——模型_应用维度</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>国内外知名大模型及应用——模型/应用维度2026/06/17</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_162648_AIGC是什么普通人如何自学AI应用</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>零基础学AIGC，我靠这4个教训避开了所有坑</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_162040_Transformer结构及其应用详解-</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Transformer结构及其应用详解-</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_161230_大模型高效微调-LoRA原理详解和训练过</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型高效微调-LoRA原理详解和训练过程深入分析</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_160341_万字长文总结多模态大模型最新进展(Mod</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>多模态大模型最新进展Modality Bridging篇</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_155310_「从夯到拉」2026年AI编程工具全景测</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>从夯到拉2026年AI编程工具全景测评</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_154604_Claude_Code_源码泄露,我从中</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Claude Code 源码泄露,我从中扒出了 Anth</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_154023_2023年神秘而难以理解的大模型强化学习</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>神秘而难以理解的大模型强化学习技术</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_153302_2023年能够解决复杂问题的思维链技术:</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>能够解决复杂问题的思维链技术:Cot,ToT,GoT,AoT</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_152618_多模态大模型主流架构介绍:从_LLaVA</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>多模态大模型主流架构介绍:从 LLaVA 到 Qwen3</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_151325_基于深度学习的自然语言处理在_2016_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>基于深度学习的自然语言处理在 2016 年有哪些值得期待</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_150639_国内AI大模型已近80个,哪个最有前途</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>国内AI大模型已近80个,哪个最有前途?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_145928_如何评价Claude_Fable_5全球</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何评价Claude Fable 5全球暂停访问?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_145232_一文搞懂大模型RAG应用(附实践案例)</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>一文搞懂大模型RAG应用附实践案例</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_144523_Andrej_Kaparthy:如何用L</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Andrej Kaparthy</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_144037_五部门联合发布《人工智能拟人化互动服务管</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>五部门联合发布《人工智能拟人化互动服务管理暂行办法》,提</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_143454_再论大模型位置编码及其外推性(万字长文)</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>再论大模型位置编码及其外推性</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_142634_别再花钱调APIKey了!2026最全免</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>别再花钱调APIKey了!2026免费大模型,国内外直连</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_142206_如何评价_OpenAI_最新发布的_GP</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>GPT-5.5 token消耗降了40%，SWE-Bench却几乎原地踏步</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_141606_0%完成率!Claude、GPT、Gem</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>0%完成率!Claude、GPT、Gemini 全灭,S</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_140953_为什么_LLM_仅预测下一词,就能「涌现</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>为什么 LLM 仅预测下一词,就能涌现出高级能力?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_140301_垂直领域大模型微调实践经验最全总结</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>垂直领域大模型微调实践经验</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_135551_如何使用Claude_Code实现科研自</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何使用Claude Code实现科研自动化</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_134819_Fable_5_的系统提示词被人扒出来了</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Fable 5 的系统提示词被人扒出来了,精彩,太精彩了</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_134048_全程不写一行代码,我如何用_AI_做出一</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>全程不写一行代码,我如何用 AI 做出一个复杂的小说分析系统</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_133246_目前针对大模型进行量化的方法有哪些</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>目前针对大模型进行量化的方法有哪些?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_132515_Prompt_caching,一篇就够了</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Prompt caching,一篇就够了</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_131642_2025年AI大模型谁能笑到最后</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>AI大模型?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_130854_目前AI编程工具哪个最好用</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>目前AI编程工具哪个最好用?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_125942_GPT-4的多模态能力是如何实现的</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>GPT-4的多模态能力是如何实现的?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_090458_大语言模型推理加速:硬件视角的全面解析_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>算力焦虑里，藏着一个你没注意到的突破口！</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_090244_垂直领域大模型的一些思考及开源模型汇总_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>垂直领域大模型的一些思考及开源模型汇总</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_091259_深度学习与医学图像分析_-_知乎</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>深度学习与医学图像分析</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_100854_Transformer-based模型到</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>20260620 100854 Transformer Based模型到</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_091659_NLP预训练模型(2021版)_-_知乎</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>2021年，NLP预训练模型杀疯了，但工程师差点哭了</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_113119_垂直领域大模型微调实践经验最全总结_-_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>20260620 113119 垂直领域大模型微调实践经验最全总结   </news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_091522_预训练大语言模型的三种微调技术总结:fi</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>预训练大语言模型的三种微调技术总结:fine-tuning、parameter-efficient fine-tuning和prompt-tuning的介绍和对比</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_112053_如果你是大模型算法的面试官,你会问哪些问</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如果你是大模型算法的面试官,你会问哪些问题?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_124642_分析transformer模型的参数量、</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>分析transformer模型的参数量、计算量、中间激活</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_123537_视觉生成超详细解读_(目录)</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>视觉生成超详细解读 (目录)</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_122715_LLM推理优化技术综述:KVCache、</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>LLM推理优化技术综述:KVCache、PageAttention、FlashAttention、MQA、GQA</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_121925_从啥也不会到DeepSpeed————一</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>从啥也不会到DeepSpeed————一篇大模型分布式训练的学习过程总结</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_121056_GPT-4大模型硬核解读!看完成半个专家</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>GPT-4大模型硬核解读!看完成半个专家</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_120311_估算_Claude_Opus_4.5_4</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>估算 Claude Opus 4.5/4.6 和 GPT 的真实大小:它到底有多少参数?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_115342_如何利用AI进行小说创作</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何利用AI进行小说创作?</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_114509_OpenAI_GPT-5.5_测评</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>OpenAI GPT-5.5 测评</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_084934_国内AI大模型已近80个,哪个最有前途</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>国内AI大模型已近80个,哪个最有前途? -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_084003_「从夯到拉」2026年AI编程工具全景测</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>「从夯到拉」2026年AI编程工具全景测评 -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_075423_OpenAI_GPT-5.5_测评_-_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>OpenAI GPT-5.5 测评 -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_083617_如何评价_OpenAI_最新发布的_GP</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何评价 OpenAI 最新发布的 GPT-5.5 模型? -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_053403_清华00后揪出AI幻觉元凶:仅0.1%神</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>清华00后揪出AI幻觉元凶:仅0.1%神经元,一按就老实 -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_064100_Bash_Is_All_Agent_Ne</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>Bash Is All Agent Need:Anthropic 重新定义智能体开发 -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_064521_目前AI编程工具哪个最好用_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>目前AI编程工具哪个最好用? -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_070000_AI安全风险已从说错话升级为系统性欺骗</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>AI 安全风险已从「说错话」升级为更隐蔽的「系统性欺骗」</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_060000_NVIDIA数据飞轮实测_1B小模型干翻70B</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>NVIDIA 数据飞轮实测：1B 小模型干翻 70B，工具调用准确率达到 98%</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_064903_别再花钱调APIKey了!2026最全免</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>别再花钱调APIKey了!2026最全免费大模型合集,国内外直连、不限额度都有 -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_064328_2025年AI大模型谁能笑到最后_-_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>2025年AI大模型谁能笑到最后? -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_031005_ICLR_2026_Efficient_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>ICLR 2026 Efficient Agent 阅读笔记 -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_052620_大模型思维链(Chain-of-Thou</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>大模型思维链(Chain-of-Thought)技术原理 -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_053201_如何利用AI进行小说创作_</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>如何利用AI进行小说创作? -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_041406_多模态大模型主流架构介绍:从_LLaVA</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>多模态大模型主流架构介绍:从 LLaVA 到 Qwen3-VL,解构多模态大模型的演进之路 -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_030709_人人都能看懂的_On-Policy_Di</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>人人都能看懂的 On-Policy Distillation(OPD):原理、源码与工业实践 -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_025935_从技术报告看On-Policy_Dist</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>从技术报告看On-Policy Distillation的崛起: 大模型后训练新范式 -</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://api.makesence.top/news/20260620_030337_近期来自seed,deepseek,de</loc>
    <lastmod>2026-06-23T22:04:46+08:00</lastmod>
    <changefreq>monthly</changefreq>
    <priority>0.6</priority>
    <news:news>
      <news:publication>
        <news:name>MakeSense 智感</news:name>
        <news:language>zh-cn</news:language>
      </news:publication>
      <news:publication_date>2026-06-23T22:04:46+08:00</news:publication_date>
      <news:title>近期来自seed,deepseek,deepmind工作量拉满的几篇文章 -</news:title>
    </news:news>
  </url>
</urlset>