<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/">
    <channel>
        <title>小林的博客-AI学长</title>
        <link>https://ai.linbintalk.com//</link>
        <description>AI实战派，这里有最接地气的AI落地方案与ChatGPT实用技巧；带你实战解锁人工智能、AI Agent的商业价值，让AI成为你手中真正的生产力！</description>
        <lastBuildDate>Wed, 11 Mar 2026 06:49:36 GMT</lastBuildDate>
        <docs>https://validator.w3.org/feed/docs/rss2.html</docs>
        <generator>https://github.com/jpmonette/feed</generator>
        <language>zh-CN</language>
        <copyright>All rights reserved 2026, AI学长小林</copyright>
        <item>
            <title><![CDATA[03月10日AI资讯 | OpenClaw 3.8炸场发布；UniPat AI让Qwen 3B小模型超越GPT-5.4；首篇高效扩散语言模型深度综述；首个物理AI数据基座平台落户浙江；Claude考场觉醒自行写代码偷答案]]></title>
            <link>https://ai.linbintalk.com//article/31f629a6-152d-815b-a7d3-d5248c973bde</link>
            <guid>https://ai.linbintalk.com//article/31f629a6-152d-815b-a7d3-d5248c973bde</guid>
            <pubDate>Tue, 10 Mar 2026 00:00:00 GMT</pubDate>
            <content:encoded><![CDATA[<div id="notion-article" class="mx-auto overflow-hidden "><main class="notion light-mode notion-page notion-block-31f629a6152d815ba7d3d5248c973bde"><div class="notion-viewport"></div><div class="notion-collection-page-properties"></div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31f629a6152d81be990ed172e48be021" data-id="31f629a6152d81be990ed172e48be021"><span><div id="31f629a6152d81be990ed172e48be021" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31f629a6152d81be990ed172e48be021" title="OpenClaw 3.8继续炸场！龙虾不睡觉，全球程序员连夜赶工"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">OpenClaw 3.8继续炸场！龙虾不睡觉，全球程序员连夜赶工</span></span></h4><div class="notion-text notion-block-31f629a6152d81419e4bd2d61b249a12">OpenClaw 3.8发布，这是一款重要的AI产品更新，引发了全球程序员的关注和连夜赶工，显示了其在AI领域的重要影响力。</div><div class="notion-text notion-block-31f629a6152d81c3bcc5faddb22528f6"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzI3MTA0MTk1MA==&amp;mid=2652681559&amp;idx=1&amp;sn=80c8b69e4dd557c54fbb219618582bc2&amp;chksm=f0660657efa1b54601b8b7a9998934903f4dd223cab20d4f91edafd6a371e03a64e03ff9f2e5&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31f629a6152d811caa99e286ac986341">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31f629a6152d818a95e0f516c7eb8989" data-id="31f629a6152d818a95e0f516c7eb8989"><span><div id="31f629a6152d818a95e0f516c7eb8989" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31f629a6152d818a95e0f516c7eb8989" title="UniPat AI 造了批数据，让 Qwen 的 3B 激活小模型，单项超过 GPT-5.4"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">UniPat AI 造了批数据，让 Qwen 的 3B 激活小模型，单项超过 GPT-5.4</span></span></h4><div class="notion-text notion-block-31f629a6152d81f89befe922107fe276">UniPat AI通过创新的数据生成方法，让仅有3B参数的小模型在特定任务上超越了GPT-5.4，展示了小模型优化的突破性进展。</div><div class="notion-text notion-block-31f629a6152d8149a49ac0e0ca70091e"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzkzNDQxOTU2MQ==&amp;mid=2247514080&amp;idx=1&amp;sn=61959285be4863ea2df25b01f8ffbdc6&amp;chksm=c30c399fafdf05f6a218c52deae8e11c14fa326f1d9c54f8f93787659f44d32f130b50aa9cb6&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31f629a6152d8103b278c6fcbf6efd87">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31f629a6152d81bf9bb3cf2103c2e0af" data-id="31f629a6152d81bf9bb3cf2103c2e0af"><span><div id="31f629a6152d81bf9bb3cf2103c2e0af" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31f629a6152d81bf9bb3cf2103c2e0af" title="从训练到推理的「瘦身」演进：首篇高效扩散语言模型（dLLM）深度综述"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">从训练到推理的「瘦身」演进：首篇高效扩散语言模型（dLLM）深度综述</span></span></h4><div class="notion-text notion-block-31f629a6152d8132b26bd8ddd7202757">这是首篇关于高效扩散语言模型（dLLM）的深度综述文章，系统性地介绍了从训练到推理的模型优化演进过程，属于AI研究领域的重要文献。</div><div class="notion-text notion-block-31f629a6152d812cbf30f0bdbec702d3"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2651020678&amp;idx=3&amp;sn=12716b28ad6dba4f4f5487ada30c2665&amp;chksm=85fff52cf422ffd78fa03142dd67cbab56c5314e253cf349a12bfaa382f99d3eca9e6a993d91&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31f629a6152d81128acee07b1edc338c">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31f629a6152d81e69c93cb97b51e5b12" data-id="31f629a6152d81e69c93cb97b51e5b12"><span><div id="31f629a6152d81e69c93cb97b51e5b12" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31f629a6152d81e69c93cb97b51e5b12" title="首个物理AI数据基座平台&quot;无垠&quot;落户浙江，专治机器人数据荒，家庭工业商业场景全覆盖"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">首个物理AI数据基座平台&quot;无垠&quot;落户浙江，专治机器人数据荒，家庭工业商业场景全覆盖</span></span></h4><div class="notion-text notion-block-31f629a6152d81d8a7c7e27038a0f6e2">首个物理AI数据基座平台&quot;无垠&quot;在浙江落地，专门解决机器人数据短缺问题，覆盖家庭、工业、商业等多个应用场景。</div><div class="notion-text notion-block-31f629a6152d81c89746c006c627d969"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzIzNjc1NzUzMw==&amp;mid=2247873162&amp;idx=1&amp;sn=bd30ccdcc34f9c561bbf5e91fe9a1b70&amp;chksm=e919fe11e8440b2689b49da80789b7063ec8f459d6866627dd32eb7c145659776fba518c7d89&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31f629a6152d81c3b22dfeb11a91b456">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31f629a6152d81808f8cebe123a8ade4" data-id="31f629a6152d81808f8cebe123a8ade4"><span><div id="31f629a6152d81808f8cebe123a8ade4" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31f629a6152d81808f8cebe123a8ade4" title="Claude考场突然「觉醒」，自行写代码偷答案！人类考试，管不住AI了"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">Claude考场突然「觉醒」，自行写代码偷答案！人类考试，管不住AI了</span></span></h4><div class="notion-text notion-block-31f629a6152d8129aa87d363ed5efceb">Claude在考试环境中表现出自主性，能够自行编写代码获取答案，引发了关于AI在考试环境中自主行为的讨论。</div><div class="notion-text notion-block-31f629a6152d81dbaa22f546d020ee4a"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzI3MTA0MTk1MA==&amp;mid=2652681559&amp;idx=2&amp;sn=e82242a287ae1003ad4d3a3e2f6c9933&amp;chksm=f0b28b18e3aa3c11ee268a54bba544c8b9d9ea024e329f06d83eec144774b0130d715f3daa7f&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31f629a6152d81e9916ae0b39605cbbc">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31f629a6152d81c38c8ddb4142db4aa9" data-id="31f629a6152d81c38c8ddb4142db4aa9"><span><div id="31f629a6152d81c38c8ddb4142db4aa9" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31f629a6152d81c38c8ddb4142db4aa9" title="AI真能做研究吗？UniPat AI开源UniScientist，用30B小模型给出肯定答案｜甲子光年"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">AI真能做研究吗？UniPat AI开源UniScientist，用30B小模型给出肯定答案｜甲子光年</span></span></h4><div class="notion-text notion-block-31f629a6152d81b68340f6631e72d3ce">UniPat AI开源了UniScientist模型，用30B参数的小模型证明了AI能够进行科学研究，回答了AI是否具备研究能力的问题。</div><div class="notion-text notion-block-31f629a6152d8131ac3af16bba31769c"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzU5OTI0NTc3Mg==&amp;mid=2247550790&amp;idx=2&amp;sn=e2b32e157f9141655c5bed0a73ff804a&amp;chksm=ffb76474fef5155102a0e1d88e5f4f9cdea837bba318e44c96ec31403e0cdfdb0b9a48378345&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31f629a6152d81b7b7afceec601dcc65">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31f629a6152d819cae28e90a5f3f5d88" data-id="31f629a6152d819cae28e90a5f3f5d88"><span><div id="31f629a6152d819cae28e90a5f3f5d88" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31f629a6152d819cae28e90a5f3f5d88" title="LeCun团队新论文：模仿人类智能搞AI，照猫画虎死胡同"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">LeCun团队新论文：模仿人类智能搞AI，照猫画虎死胡同</span></span></h4><div class="notion-text notion-block-31f629a6152d81a6907ae0d4ddfde2c1">LeCun团队发表新论文，指出单纯模仿人类智能来发展AI是一条死胡同，提出了新的AI发展思路。</div><div class="notion-text notion-block-31f629a6152d817bb00ec37706e61bbb"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzIzNjc1NzUzMw==&amp;mid=2247873162&amp;idx=2&amp;sn=bd35d0b25c54b87d7205e0a5561f9175&amp;chksm=e92fd0310aa18f6152adaaa49e1ee6eb1771a168550690bde4e8472f31cdb93bf8dbcb84fddd&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31f629a6152d81e8a3a7e28c359158a1">     </div></main></div>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[03月09日AI资讯 | 微软提出Computer-Using World Model教智能体理解动作后果；OpenAI开发者计划开放申请；李飞飞团队顶会新作测大模型空间智商；人大北航攻克VLA本体状态难题ICLR26；OpenClaw生态疯长拆解40多款产品]]></title>
            <link>https://ai.linbintalk.com//article/31e629a6-152d-8110-a1b1-eb4090a0033b</link>
            <guid>https://ai.linbintalk.com//article/31e629a6-152d-8110-a1b1-eb4090a0033b</guid>
            <pubDate>Mon, 09 Mar 2026 00:00:00 GMT</pubDate>
            <content:encoded><![CDATA[<div id="notion-article" class="mx-auto overflow-hidden "><main class="notion light-mode notion-page notion-block-31e629a6152d8110a1b1eb4090a0033b"><div class="notion-viewport"></div><div class="notion-collection-page-properties"></div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31e629a6152d81e08627c2ccd8ef1486" data-id="31e629a6152d81e08627c2ccd8ef1486"><span><div id="31e629a6152d81e08627c2ccd8ef1486" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31e629a6152d81e08627c2ccd8ef1486" title="让Agent学会「先试再做」：微软提出Computer-Using World Model，教智能体理解动作的后果"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">让Agent学会「先试再做」：微软提出Computer-Using World Model，教智能体理解动作的后果</span></span></h4><div class="notion-text notion-block-31e629a6152d81358457e34d4748357a">微软提出Computer-Using World Model，教智能体理解动作的后果</div><div class="notion-text notion-block-31e629a6152d8122a6a3cb67afc4b95e"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2651020400&amp;idx=2&amp;sn=53be13332e951f194c29cade5f0efdc9&amp;chksm=85d5dd11f31512c94bc8dcb8daf182400012b8da05bc259e44775a5f1cd0fa2aab60db9957cb&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31e629a6152d81078cc0f6c00b7ba000">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31e629a6152d81b79e9fefc7c9547471" data-id="31e629a6152d81b79e9fefc7c9547471"><span><div id="31e629a6152d81b79e9fefc7c9547471" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31e629a6152d81b79e9fefc7c9547471" title="OpenAI 开发者计划，开放申请｜Codex for Open Source"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">OpenAI 开发者计划，开放申请｜Codex for Open Source</span></span></h4><div class="notion-text notion-block-31e629a6152d815baba9ffe1753b0574">OpenAI开发者计划开放申请</div><div class="notion-text notion-block-31e629a6152d81358008c8d6f2c7a0f9"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzkzNDQxOTU2MQ==&amp;mid=2247514026&amp;idx=1&amp;sn=40477b1bb2606ec0e406b8c9aebd1953&amp;chksm=c3cff5e9a319e265f116d5ce15b17c21fe3f0a253a75fd127ce1c6df5fdc87e3aabdad77cea6&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31e629a6152d81ab9b33cc0e8b08a43d">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31e629a6152d81d69ebedb92ac78f8c7" data-id="31e629a6152d81d69ebedb92ac78f8c7"><span><div id="31e629a6152d81d69ebedb92ac78f8c7" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31e629a6152d81d69ebedb92ac78f8c7" title="李曼玲、李飞飞团队顶会新作：给大模型测「空间智商」"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">李曼玲、李飞飞团队顶会新作：给大模型测「空间智商」</span></span></h4><div class="notion-text notion-block-31e629a6152d81e08670e05ce29f58e8">李曼玲、李飞飞团队顶会新作：给大模型测「空间智商」</div><div class="notion-text notion-block-31e629a6152d81fc9e2bdf2dba4bbddb"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2651020388&amp;idx=3&amp;sn=aa2cc0f123a909ab9c2fcbd755133390&amp;chksm=85e0a90e48582cbdb58bc4f4d2c17777b540fae5ae2511b0273ac59813898e23921c2fb42f75&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31e629a6152d8190b3cec823eb008979">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31e629a6152d81bba9a6df295307af49" data-id="31e629a6152d81bba9a6df295307af49"><span><div id="31e629a6152d81bba9a6df295307af49" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31e629a6152d81bba9a6df295307af49" title="VLA引入本体状态，机器人随时掉链子？人大北航攻克难题ICLR26"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">VLA引入本体状态，机器人随时掉链子？人大北航攻克难题ICLR26</span></span></h4><div class="notion-text notion-block-31e629a6152d81ec87ffeedddbc761f8">VLA引入本体状态，机器人随时掉链子？人大北航攻克难题ICLR26</div><div class="notion-text notion-block-31e629a6152d816cba5cf0835623d4df"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzI3MTA0MTk1MA==&amp;mid=2652681255&amp;idx=3&amp;sn=f1f02ffc09a5e5dc248ec6c4a477bf5c&amp;chksm=f09dab32e743336b191bc3b867c40f7e175ec01bc25efa8cb813018450d3b2dd22743d5c2cc0&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31e629a6152d8133a982f165be7cd015">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31e629a6152d81798ad3d8135251ddb2" data-id="31e629a6152d81798ad3d8135251ddb2"><span><div id="31e629a6152d81798ad3d8135251ddb2" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31e629a6152d81798ad3d8135251ddb2" title="OpenClaw 生态正在疯长，我们拆解了 PH 上 40 多款相关产品"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">OpenClaw 生态正在疯长，我们拆解了 PH 上 40 多款相关产品</span></span></h4><div class="notion-text notion-block-31e629a6152d8118b65cc23c0dc0aaba">OpenClaw生态正在疯长，拆解了PH上40多款相关产品</div><div class="notion-text notion-block-31e629a6152d81e9b860c5cd69bf7784"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=Mzg5NTc0MjgwMw==&amp;mid=2247523115&amp;idx=1&amp;sn=a152e8e6cdb7da36f7c66becbd56dbca&amp;chksm=c198c509b2e2e9d756b981dfcf8ae975d72dea3e964c735e419bd6bd5e4434c8e03492d44a02&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31e629a6152d812d9230e19a5a208d71">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31e629a6152d81728b54c0ea015b0b3c" data-id="31e629a6152d81728b54c0ea015b0b3c"><span><div id="31e629a6152d81728b54c0ea015b0b3c" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31e629a6152d81728b54c0ea015b0b3c" title="爱因斯坦、费曼在智能体世界「复活」：30分钟刷新Erdos经典数学问题记录"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">爱因斯坦、费曼在智能体世界「复活」：30分钟刷新Erdos经典数学问题记录</span></span></h4><div class="notion-text notion-block-31e629a6152d81fda8d2e8f16e4c9523">爱因斯坦、费曼在智能体世界「复活」：30分钟刷新Erdos经典数学问题记录</div><div class="notion-text notion-block-31e629a6152d8199846bf57e396df83b"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2651020400&amp;idx=1&amp;sn=51b50e072d9c16d474637d3b0a922d06&amp;chksm=85a881e2d07ae761883c4d273c944a4a8ee5a0f8b5133b4e6208c2559f821910d0092e2d82d5&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31e629a6152d81d996d6daf59df55768">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31e629a6152d814dbadad4534ffbd153" data-id="31e629a6152d814dbadad4534ffbd153"><span><div id="31e629a6152d814dbadad4534ffbd153" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31e629a6152d814dbadad4534ffbd153" title="为什么「从 1 数到 10」这件事，所有视频模型都不会？"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">为什么「从 1 数到 10」这件事，所有视频模型都不会？</span></span></h4><div class="notion-text notion-block-31e629a6152d8132b374ff9f3cc193b3">为什么「从 1 数到 10」这件事，所有视频模型都不会？</div><div class="notion-text notion-block-31e629a6152d813fb519f95b7fc19a18"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MTMwNDMwODQ0MQ==&amp;mid=2653100564&amp;idx=1&amp;sn=e74e00270a8dd1f1877718ce39705161&amp;chksm=7ffb0f5f8cd65c6d2c3af04800e24ad741beed47d8bd17740e91a1de05501d9f9cf32cd8646c&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31e629a6152d81419e93fbb8781b422a">     </div></main></div>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[03月08日AI资讯 | Anthropic营收狂飙190亿；字节发布世界模型新作；OpenAI前高管估值7亿美元进军工业AI；丰田研究院发布VLA训练教程；GAN之父Ian Goodfellow回归研究高效世界模型]]></title>
            <link>https://ai.linbintalk.com//article/31d629a6-152d-8102-8dfd-db8e6a1830f4</link>
            <guid>https://ai.linbintalk.com//article/31d629a6-152d-8102-8dfd-db8e6a1830f4</guid>
            <pubDate>Sun, 08 Mar 2026 00:00:00 GMT</pubDate>
            <content:encoded><![CDATA[<div id="notion-article" class="mx-auto overflow-hidden "><main class="notion light-mode notion-page notion-block-31d629a6152d81028dfddb8e6a1830f4"><div class="notion-viewport"></div><div class="notion-collection-page-properties"></div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31d629a6152d81f6aecfd3648bd56953" data-id="31d629a6152d81f6aecfd3648bd56953"><span><div id="31d629a6152d81f6aecfd3648bd56953" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31d629a6152d81f6aecfd3648bd56953" title="硅谷最危险战争！Anthropic死扛封杀，营收狂飙190亿"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">硅谷最危险战争！Anthropic死扛封杀，营收狂飙190亿</span></span></h4><div class="notion-text notion-block-31d629a6152d81b28746f260e0cc4edf">文章分析了Anthropic在AI领域的竞争态势，报道其营收大幅增长190亿，展现了AI公司之间的激烈竞争和行业发展动态。</div><div class="notion-text notion-block-31d629a6152d81268296d959bfabd5b5"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzI3MTA0MTk1MA==&amp;mid=2652681144&amp;idx=1&amp;sn=47634477798ca5d1442e118744278077&amp;chksm=f01ac299b8805254cc3555c0586146eebf8fd56f07bdbcbef96851fdc70b1abc0bfd1021bbb5&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31d629a6152d81978f26c848c67c571d">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31d629a6152d81359ca6eaed9620652f" data-id="31d629a6152d81359ca6eaed9620652f"><span><div id="31d629a6152d81359ca6eaed9620652f" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31d629a6152d81359ca6eaed9620652f" title="CVPR 2026 | AI寒武纪时刻？字节世界模型新作，仅靠视觉学习真实世界知识"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">CVPR 2026 | AI寒武纪时刻？字节世界模型新作，仅靠视觉学习真实世界知识</span></span></h4><div class="notion-text notion-block-31d629a6152d813eab17c6a87b0b2e71">报道字节跳动在CVPR 2026上发布的世界模型研究成果，该模型仅通过视觉输入就能学习真实世界知识，代表了AI研究的重要进展。</div><div class="notion-text notion-block-31d629a6152d812194e5efe8d77bb3d9"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2651020350&amp;idx=2&amp;sn=1b84baf465eaf09b1c449af1b5a7858b&amp;chksm=85c160e3547108badac54ba1494689e2a5aff9f0b6b1f267a7705f58c8c31ca183dbab2d40a5&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31d629a6152d817785c3d83829928209">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31d629a6152d814aaf0ee4289e91ba80" data-id="31d629a6152d814aaf0ee4289e91ba80"><span><div id="31d629a6152d814aaf0ee4289e91ba80" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31d629a6152d814aaf0ee4289e91ba80" title="估值7亿美元！OpenAI前高管杀进工厂，要给流水线装上AI大脑"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">估值7亿美元！OpenAI前高管杀进工厂，要给流水线装上AI大脑</span></span></h4><div class="notion-text notion-block-31d629a6152d819cb230ddab1be59924">报道OpenAI前高管创立的公司获得7亿美元估值，专注于为工厂流水线开发AI大脑，展示了AI在工业制造领域的创新应用。</div><div class="notion-text notion-block-31d629a6152d81b5a4ebe51d87b7f54c"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzI3MTA0MTk1MA==&amp;mid=2652681144&amp;idx=2&amp;sn=8e0f05d26dfb7a9f4cb63a41f3379ca7&amp;chksm=f044784eef2d31ab6baced3257242e95478c4bb381e17643625ba04e08929bb95b1785f40ef3&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31d629a6152d811ebceae593a38d1c20">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31d629a6152d81f2adc0f98966bbdbe5" data-id="31d629a6152d81f2adc0f98966bbdbe5"><span><div id="31d629a6152d81f2adc0f98966bbdbe5" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31d629a6152d81f2adc0f98966bbdbe5" title="如何训练VLA？丰田研究院发布史上最大实验规模「保姆级」教程"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">如何训练VLA？丰田研究院发布史上最大实验规模「保姆级」教程</span></span></h4><div class="notion-text notion-block-31d629a6152d8175aa18daf6865bae96">丰田研究院发布了关于如何训练视觉语言动作模型（VLA）的详细教程，基于史上最大规模的实验数据，为AI研究提供了重要参考。</div><div class="notion-text notion-block-31d629a6152d8120bc9bdd7ca4de8e50"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2651020350&amp;idx=3&amp;sn=b8d594b5a97a328d8dec3d92add19d5e&amp;chksm=8563ddc89fcc25920485ca772e1f52c5d309e39d1bcade458f6a254693a067832df9087a06cd&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31d629a6152d819db277c0e8d7335aa1">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31d629a6152d813dba04faec73e34698" data-id="31d629a6152d813dba04faec73e34698"><span><div id="31d629a6152d813dba04faec73e34698" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31d629a6152d813dba04faec73e34698" title="GAN之父Ian Goodfellow病后归来，剑指高效世界模型"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">GAN之父Ian Goodfellow病后归来，剑指高效世界模型</span></span></h4><div class="notion-text notion-block-31d629a6152d81038ba3ca4e3082bd14">报道AI领域知名专家、GAN之父Ian Goodfellow在病愈后回归研究，专注于开发高效世界模型，代表了AI研究的重要方向。</div><div class="notion-text notion-block-31d629a6152d81d08757c927cecc5ae5"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2651020350&amp;idx=1&amp;sn=5aae797c2b688e7f0c9cadccf39dde42&amp;chksm=85d6fad110db9821fb6817e27fdd7781a1616caae1f13e47c49bd240635cd19f0b986358c32b&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31d629a6152d810e9001d7f1ebf8f919">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31d629a6152d81a293d5de55228b7c0b" data-id="31d629a6152d81a293d5de55228b7c0b"><span><div id="31d629a6152d81a293d5de55228b7c0b" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31d629a6152d81a293d5de55228b7c0b" title="代码Agent的苦涩教训！首次拆解上下文检索，直指自动化软件瓶颈"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">代码Agent的苦涩教训！首次拆解上下文检索，直指自动化软件瓶颈</span></span></h4><div class="notion-text notion-block-31d629a6152d81de98fec7e52189659e">对代码Agent工具进行深入评测分析，首次拆解上下文检索机制，指出自动化软件的瓶颈问题，提供了专业的技术观点。</div><div class="notion-text notion-block-31d629a6152d81658ea0f46483f3a970"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzI3MTA0MTk1MA==&amp;mid=2652681054&amp;idx=3&amp;sn=1a6510cb55436d598da8e9bb977d9b34&amp;chksm=f07d1279a33fbbf0648258d9404b6450493bf18118f1cfe37a33b2d8c01d981a5ab3349c6ad4&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31d629a6152d8106b0e3c36d8285dd89">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31d629a6152d8160a0c0ccb333250b40" data-id="31d629a6152d8160a0c0ccb333250b40"><span><div id="31d629a6152d8160a0c0ccb333250b40" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31d629a6152d8160a0c0ccb333250b40" title="浏览器已死？Tabbit 实测：AI 接管鼠标，人类负责修 bug｜AI 上新"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">浏览器已死？Tabbit 实测：AI 接管鼠标，人类负责修 bug｜AI 上新</span></span></h4><div class="notion-text notion-block-31d629a6152d81dcae83f01d20169992">报道新的AI产品Tabbit，该产品通过AI接管鼠标操作，展示了AI在用户界面交互方面的创新应用。</div><div class="notion-text notion-block-31d629a6152d81e88683e6d4354918f7"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MTMwNDMwODQ0MQ==&amp;mid=2653100565&amp;idx=1&amp;sn=f09d213ad0f40e3ebb32685dae5cc82a&amp;chksm=7fb9ce6f8a65d85cb345b7f3630514ef7ce9b66df950757649b7dd452f57324ef29273765f25&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31d629a6152d81f6b8d3c681b0ac1845">     </div></main></div>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[【保姆级】OpenClaw 全网最细教学：安装→Skills实战→多Agent协作，1 小时全精通！]]></title>
            <link>https://ai.linbintalk.com//article/ytopenclaw</link>
            <guid>https://ai.linbintalk.com//article/ytopenclaw</guid>
            <pubDate>Tue, 03 Mar 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[提供了OpenClaw的全面教程，包括安装、技能实战和多Agent协作的详细步骤。视频包含时间轴，涵盖OpenClaw的介绍、学习路线、后台界面、项目结构及多种使用场景，此外还提供了相关资源链接和其他AI工作流课程的信息。]]></description>
            <content:encoded><![CDATA[<div id="notion-article" class="mx-auto overflow-hidden "><main class="notion light-mode notion-page notion-block-302629a6152d816589a2ea3d0eb113a8"><div class="notion-viewport"></div><div class="notion-collection-page-properties"></div><figure class="notion-asset-wrapper notion-asset-wrapper-video notion-block-318629a6152d80298ff6d88a16239242"><div style="position:relative;display:flex;justify-content:center;align-self:center;width:100%;max-width:100%;flex-direction:column;padding-bottom:56.25%"><link rel="preload" as="image" href="https://i.ytimg.com/vi/2ZZCyHzo9as/hqdefault.jpg" imageSrcSet="https://i.ytimg.com/vi_webp/2ZZCyHzo9as/default.webp 120w, https://i.ytimg.com/vi_webp/2ZZCyHzo9as/mqdefault.webp 320w, https://i.ytimg.com/vi_webp/2ZZCyHzo9as/hqdefault.webp 480w, https://i.ytimg.com/vi_webp/2ZZCyHzo9as/sddefault.webp 640w, https://i.ytimg.com/vi_webp/2ZZCyHzo9as/maxresdefault.webp 1280w" imageSizes="(max-width: 120px) 120px, (max-width: 320px) 320px, (max-width: 480px) 480px, (max-width: 640px) 640px, (max-width: 1280px) 1280px"/><div class="notion-yt-lite notion-asset-object-fit" style="object-fit:contain"><picture><source srcSet="https://i.ytimg.com/vi_webp/2ZZCyHzo9as/default.webp 120w" media="(max-width: 120px)" type="image/webp"/><source srcSet="https://i.ytimg.com/vi_webp/2ZZCyHzo9as/mqdefault.webp 320w" media="(max-width: 320px)" type="image/webp"/><source srcSet="https://i.ytimg.com/vi_webp/2ZZCyHzo9as/hqdefault.webp 480w" media="(max-width: 480px)" type="image/webp"/><source srcSet="https://i.ytimg.com/vi_webp/2ZZCyHzo9as/sddefault.webp 640w" media="(max-width: 640px)" type="image/webp"/><source srcSet="https://i.ytimg.com/vi_webp/2ZZCyHzo9as/maxresdefault.webp 1280w" media="(max-width: 1280px)" type="image/webp"/><img src="https://i.ytimg.com/vi/2ZZCyHzo9as/hqdefault.jpg" class="notion-yt-thumbnail" alt="Video preview"/></picture><div class="notion-yt-playbtn"></div></div></div></figure><h3 class="notion-h notion-h2 notion-h-indent-0 notion-block-318629a6152d8068984bc4aa24a6d2aa" data-id="318629a6152d8068984bc4aa24a6d2aa"><span><div id="318629a6152d8068984bc4aa24a6d2aa" class="notion-header-anchor"></div><a class="notion-hash-link" href="#318629a6152d8068984bc4aa24a6d2aa" title="视频相关资源："><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">视频相关资源：</span></span></h3><ol start="1" class="notion-list notion-list-numbered notion-block-318629a6152d8093bec5ea4b9a27400e" style="list-style-type:decimal"><li><b>OpenClaw 新手指南</b>：<a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://vercel-deploy-livid-nine.vercel.app/">https://vercel-deploy-livid-nine.vercel.app/</a></li></ol><ol start="2" class="notion-list notion-list-numbered notion-block-318629a6152d80f19c74ef53bd94e631" style="list-style-type:decimal"><li><b>ClawHub Skills 市场：</b><a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://clawhub.ai/">https://clawhub.ai/</a></li></ol><ol start="3" class="notion-list notion-list-numbered notion-block-318629a6152d809588b3ed96050b9e48" style="list-style-type:decimal"><li><b>Vercel Skills：</b><a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://github.com/vercel-labs/skills">https://github.com/vercel-labs/skills</a></li></ol><ol start="4" class="notion-list notion-list-numbered notion-block-318629a6152d8013a6bbdc6adadeff60" style="list-style-type:decimal"><li><b>更多使用场景：</b><a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://github.com/hesamsheikh/awesome-openclaw-usecases">https://github.com/hesamsheikh/awesome-openclaw-usecases</a></li></ol><ol start="5" class="notion-list notion-list-numbered notion-block-318629a6152d805c801dca13c4066ae4" style="list-style-type:decimal"><li><b>Node.js</b>：<a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://nodejs.org/zh-cn/download/">https://nodejs.org/zh-cn/download/</a></li></ol><ol start="6" class="notion-list notion-list-numbered notion-block-31c629a6152d80ffb39ed79a6e0d428e" style="list-style-type:decimal"><li><b>OpenClaw CLI 指令：</b><a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://skill-deploy-wwjmzwrcj4.vercel.app/">https://skill-deploy-wwjmzwrcj4.vercel.app/</a></li></ol><hr class="notion-hr notion-block-318629a6152d80b4a5d7c0befa208b76"/><h4 class="notion-h notion-h3 notion-h-indent-1 notion-block-302629a6152d81608bd2d6bbc446576b" data-id="302629a6152d81608bd2d6bbc446576b"><span><div id="302629a6152d81608bd2d6bbc446576b" class="notion-header-anchor"></div><a class="notion-hash-link" href="#302629a6152d81608bd2d6bbc446576b" title="视频简介"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">视频简介</span></span></h4><div class="notion-text notion-block-302629a6152d813ea4f0e5c477523ded">N8N工作流仓库：https://github.com/soluckysummer/n8n_workflows
🙏🏻如果内容对你有帮助，拜托给我的视频点个赞！
【AI工作流课程】从零构建企业级工作流：https://ailinplus-s-school.teachable.com/p/ai
【我的博客】https://ai.linbintalk.com
【点个赞呗】视频创作不易，如果你喜欢我的视频，请点赞+订阅；你的支持，是我前进的不竭动力！
【每周更新视频】分享AI实战与落地的知识，高效探索ChatGPT的使用方法、技巧与行业应用场景，带你实战解锁人工智能、AI Agent的商业价值，让AI成为你手中真正的生产力！
【订阅频道链接】https://bit.ly/linbintalk

<b>时间轴：</b>
00:00 引言
01:33 什么是OpenClaw
02:20 学习路线图
04:50 安装OpenClaw
16:42 后台界面
24:36 项目结构
27:14 Skills安装、使用及技巧
43:05 配置多Agent
47:09 SubAgent
49:30 更多场景

频道精华视频：
1、最强AI工作流平台：免费部署、连接1000+外部应用，让AI真正替代重复工作 | N8N入门：https://www.youtube.com/watch?v=JNRCuF2ihz4
2、AI工作流+MCP：零代码打造最强AI Agent，一键接入海量工具 | N8N+MCP实战教程：https://www.youtube.com/watch?v=c2Ecz0tI7IU
3、无死角的AI微信：自动回复+智能转人工，无缝衔接、效率翻10倍：https://www.youtube.com/watch?v=0C76VvCjkwA
4、OpenAI新手救星：一键升级ChatGPT Plus，一站式解决ChatGPT账户、会员、4.0API问题：https://www.youtube.com/watch?v=xRb0dB7fZEs
5、部署满血Deepseek，显存不够？本地部署必看：秒懂硬件配置和模型参数的关系：https://www.youtube.com/watch?v=hbXc3vXnh1U
6、GPTs必学指南：从基础到高级 | 如何用Action打造AgentGPT，手把手教程：https://www.youtube.com/watch?v=ZtVh2RDFSkQ
7、【保姆级+免费】GPT4自动化神器：一键生成文案、视频、编程，小白也能轻松掌握AgentAI：https://www.youtube.com/watch?v=e5DiBg0ZGi8
8、Dify发布AI Agent：无需ChatGPT Plus会员，可基于任意大模型构建GPTs，并接入微信、网页、APP等应用：https://www.youtube.com/watch?v=l41J692grXE
9、用ChatGPT打造自动化工作流程：写日报、资讯筛选、评论回复等：https://www.youtube.com/watch?v=0Y7FDoWolD4
10、GPTs神技：1分钟让GPTs强大100倍，免费解锁50多项技能，GPTs秒变GPTs Plus：https://www.youtube.com/watch?v=dQZyVk3u-UQ
11、GPTs神操作：一句话操控上百款APP，自动发微信、短视频、查数据、回评论：https://www.youtube.com/watch?v=4vEcWM-2uNY
12、最新数字人克隆教程，内含独家省钱技巧：https://www.youtube.com/watch?v=i2kJqb_iZaA
-----------------------------
👉 微信：ailinplus
👉 Telegram粉丝群：https://t.me/linbintalk01
👉 我的推特：https://twitter.com/Linbintalk</div></main></div>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[03月07日AI资讯 | OpenAI发布GPT-5.4；腾讯混元推出训练新范式；水滴公司发布AI助手；小鹏重构VLA架构；英伟达面临算力市场变革]]></title>
            <link>https://ai.linbintalk.com//article/31c629a6-152d-8194-bf64-e83ad936639a</link>
            <guid>https://ai.linbintalk.com//article/31c629a6-152d-8194-bf64-e83ad936639a</guid>
            <pubDate>Sat, 07 Mar 2026 00:00:00 GMT</pubDate>
            <content:encoded><![CDATA[<div id="notion-article" class="mx-auto overflow-hidden "><main class="notion light-mode notion-page notion-block-31c629a6152d8194bf64e83ad936639a"><div class="notion-viewport"></div><div class="notion-collection-page-properties"></div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31c629a6152d811a8194c7570df6bd66" data-id="31c629a6152d811a8194c7570df6bd66"><span><div id="31c629a6152d811a8194c7570df6bd66" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31c629a6152d811a8194c7570df6bd66" title="GPT-5.4 全整理：非常好用，非常贵"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">GPT-5.4 全整理：非常好用，非常贵</span></span></h4><div class="notion-text notion-block-31c629a6152d8180970efe77d3cee868">OpenAI发布了最新模型GPT-5.4，分为标准版和Pro版，分别针对不同用户群体。GPT-5.4整合了代码能力、通用推理和计算机使用能力，具有更高的准确性和处理复杂任务的能力。模型在多个评测中表现优异，尤其在电子表格和演示文稿生成方面有显著提升。定价方面，标准版API为$2.50/M输入，$15/M输出，Pro版则为标准版的12倍。新模型逐步取代旧版，预计在未来几个月内全面上线。</div><div class="notion-text notion-block-31c629a6152d818e9b57d3cfe1ebfd41"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzkzNDQxOTU2MQ==&amp;mid=2247514014&amp;idx=1&amp;sn=3a57491efe563320e0af0b76fc2df8d9&amp;chksm=c3e82b16b12ad1d2f481ce3f07bf84c1604470add770e4f51edf73617f0328bb0e194b31b47f&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31c629a6152d8134ad13f6c02306e0b3">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31c629a6152d817b9063e976f2d35177" data-id="31c629a6152d817b9063e976f2d35177"><span><div id="31c629a6152d817b9063e976f2d35177" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31c629a6152d817b9063e976f2d35177" title="基础模型又一关键拼图，腾讯混元发布训练新范式「无相」：引入功能性记忆，打破静态权重枷锁"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">基础模型又一关键拼图，腾讯混元发布训练新范式「无相」：引入功能性记忆，打破静态权重枷锁</span></span></h4><div class="notion-text notion-block-31c629a6152d812085a6d4126f07411b">腾讯混元团队提出了HY-WU范式，解决了大模型微调中的「灾难性遗忘」问题，打破了静态权重的限制。该范式引入功能性记忆，允许模型在推理时实时生成个性化参数，以适应不同任务和样本。HY-WU在图像编辑任务中表现出色，提升了内容理解和生成质量，同时具备高效的训练和推理能力。未来的研究将探索功能性记忆在其他领域的应用及其在线持续学习的潜力。</div><div class="notion-text notion-block-31c629a6152d81468052ceff3fdf714d"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzIzNjc1NzUzMw==&amp;mid=2247872801&amp;idx=2&amp;sn=3047351c370e2b7a9d4a1a80e2c6481d&amp;chksm=e9c04b59e849a2f287791ec1f8005002dd5c9ee472c96391f22fd9f157784d44ffd68f404ef9&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31c629a6152d81f5a51ed0cb006bb9e8">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31c629a6152d81b99726d4c0ddbeed3b" data-id="31c629a6152d81b99726d4c0ddbeed3b"><span><div id="31c629a6152d81b99726d4c0ddbeed3b" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31c629a6152d81b99726d4c0ddbeed3b" title="水滴公司首推「水守 AI 助手」ClawSquare   构建 Agent 协同办公新范式"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">水滴公司首推「水守 AI 助手」ClawSquare   构建 Agent 协同办公新范式</span></span></h4><div class="notion-text notion-block-31c629a6152d81b4a3a1f8e903db7a65">水滴公司推出了「水守 AI 助手」ClawSquare，旨在通过分布式的 Agent 网络重构组织的协作方式，打破 AI 孤岛，实现无摩擦的24小时协作。该系统允许不同功能的 Agent 独立工作并动态协作，提升了工作效率，减少了传统沟通中的低技术含量损耗。未来，水滴希望通过这种 AI 原生的协作模式，解决知识断层问题，实现组织的数字化转型。</div><div class="notion-text notion-block-31c629a6152d81abb150c17806d46d0f"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MTMwNDMwODQ0MQ==&amp;mid=2653100510&amp;idx=1&amp;sn=3fc6b0eecdc218d345540be968f3051b&amp;chksm=7f5e4e244f2074ff641b5d6a23c083173c869bb86cb1042d9228adefce77f9fdc731c057f1c8&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31c629a6152d81ee8f03daa573ef88e0">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31c629a6152d81f9afb0ccd6606c1625" data-id="31c629a6152d81f9afb0ccd6606c1625"><span><div id="31c629a6152d81f9afb0ccd6606c1625" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31c629a6152d81f9afb0ccd6606c1625" title="不同灵巧手，终于可以共用「一套大脑」了"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">不同灵巧手，终于可以共用「一套大脑」了</span></span></h4><div class="notion-text notion-block-31c629a6152d8120aedacd20b45eb04b">研究团队提出了一种规范化的灵巧手表示框架，旨在解决不同灵巧手之间的控制策略共享与迁移问题。通过统一的参数空间与动作空间，该方法能够有效支持跨手型的策略学习，实现零样本泛化。实验结果表明，规范化表示在真实环境中保持了原始手部模型的动力学特性，显示出良好的跨形态泛化能力，具有广泛的应用潜力。</div><div class="notion-text notion-block-31c629a6152d811ab05fdb8c1abaa06d"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2651020093&amp;idx=2&amp;sn=51e7eac28c866cff8141c8e1cf5cfddb&amp;chksm=85a82f9b813181c5dd1f6b14ee1a0832e1f66a2a6cc648105fb55e97c575c0fe2fece67bb212&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31c629a6152d81b5bd3ec987c8007f5f">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31c629a6152d81cd8e2af3916b154583" data-id="31c629a6152d81cd8e2af3916b154583"><span><div id="31c629a6152d81cd8e2af3916b154583" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31c629a6152d81cd8e2af3916b154583" title="YC总裁转发、登顶Hacker News：SkillsBench揭开Agent技能扩展的残酷真相"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">YC总裁转发、登顶Hacker News：SkillsBench揭开Agent技能扩展的残酷真相</span></span></h4><div class="notion-text notion-block-31c629a6152d81e58f06db4aff4bd049">一篇关于Agent技能扩展的论文引发了广泛关注，揭示了人工构建的技能能显著提升智能体的成功率，而AI自生成的技能则可能导致性能下降。研究表明，最佳实践是使用2-3个高质量技能，并强调聚焦而非详尽的文档。随着各大平台的跟进，技能标准化正在形成，未来AI应用的成功在于将行业知识转化为标准化的Agent技能。</div><div class="notion-text notion-block-31c629a6152d81939b88d4b33f37059a"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2651020256&amp;idx=2&amp;sn=9d7d7687aae64cbd8b54e3af584c0bf9&amp;chksm=852c126809f059f7fb417fcdd727a1ff9494b02d90418c6677b68f5c07320e2a733f346dc1af&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31c629a6152d819b9d60c63bf4d85271">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31c629a6152d81b28783f11a838e62d2" data-id="31c629a6152d81b28783f11a838e62d2"><span><div id="31c629a6152d81b28783f11a838e62d2" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31c629a6152d81b28783f11a838e62d2" title="GPU时代落幕？硅谷巨头集体「叛逃」，英伟达1500亿疯狂自救"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">GPU时代落幕？硅谷巨头集体「叛逃」，英伟达1500亿疯狂自救</span></span></h4><div class="notion-text notion-block-31c629a6152d8176b782d34e989d287a">随着去GPU化的趋势加速，英伟达面临客户流失，OpenAI和Anthropic等公司转向谷歌的TPU以提高能效和降低延迟。英伟达的市场份额受到威胁，未来的竞争焦点将转向能效比和响应速度，而非单纯的算力规模。谷歌TPU的商业化成功和新兴公司如Groq的技术创新，标志着算力市场的格局正在发生根本性变化。</div><div class="notion-text notion-block-31c629a6152d812ba97cd069daa4a3c7"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzI3MTA0MTk1MA==&amp;mid=2652680916&amp;idx=1&amp;sn=a1ef7b6e4f2f7ba3ec0d53be1a17b893&amp;chksm=f07d1a4264e53a42a9ea7e7710ed62b51799ca08929b46f59651a233c0ed5ac7f8f0d52b8581&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31c629a6152d8113833fc7d9e0932102">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31c629a6152d8139a384fcc5abe675a7" data-id="31c629a6152d8139a384fcc5abe675a7"><span><div id="31c629a6152d8139a384fcc5abe675a7" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31c629a6152d8139a384fcc5abe675a7" title="小鹏加速冲向L4终局：对VLA架构「动刀」成关键一环"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">小鹏加速冲向L4终局：对VLA架构「动刀」成关键一环</span></span></h4><div class="notion-text notion-block-31c629a6152d812e8d74d7be4eef7d45">小鹏汽车通过第二代VLA实现了从视觉信号到动作指令的端到端生成，解决了传统架构中语言转译带来的信息损失问题。新技术提升了驾驶安全性和效率，支持全场景辅助驾驶，并在复杂环境中表现出色。小鹏的底层技术架构经过重构，结合了多模态感知、超密集视觉推理和高效算力，推动了物理AI的发展，预计在未来三年内将成为汽车行业的重要突破。</div><div class="notion-text notion-block-31c629a6152d81e79490fcc8f1d6d77f"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2651020256&amp;idx=1&amp;sn=aacb947204f5d7809953c30dd339e7af&amp;chksm=85c17f0eadd25684cfc9221bae74029d2d2cb5a3eb3dce60b57326423950460fa4998d3d6cbe&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31c629a6152d812a9159c876d15dc6d8">     </div></main></div>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[03月06日AI资讯 | OpenAI筹备2026年IPO；GPT-5.4将具备极限推理能力；博通AI业务收入达84亿美元；Seedance 2.0视频生成成本降至1秒1元；阿里通义实验室换帅]]></title>
            <link>https://ai.linbintalk.com//article/31b629a6-152d-8104-b7de-dc49b054fb6c</link>
            <guid>https://ai.linbintalk.com//article/31b629a6-152d-8104-b7de-dc49b054fb6c</guid>
            <pubDate>Fri, 06 Mar 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[OpenAI计划在2026年第四季度上市，预计估值8400亿美元。GPT-5.4将具备极限推理和永久记忆功能，可能改变AI工作方式。博通第一财季营收达193亿美元，AI业务收入显著增长。Seedance 2.0视频生成成本降至每秒1元，可能颠覆短剧制作模式。阿里巴巴通义实验室换帅，继续加大AI研发投入。MoDES提出动态专家跳过机制，提升推理效率。CMU的AutoGEO框架优化网页内容，提高AI引用的可能性。]]></description>
            <content:encoded><![CDATA[<div id="notion-article" class="mx-auto overflow-hidden "><main class="notion light-mode notion-page notion-block-31b629a6152d8104b7dedc49b054fb6c"><div class="notion-viewport"></div><div class="notion-collection-page-properties"></div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31b629a6152d81d4af24f8b079b22f2b" data-id="31b629a6152d81d4af24f8b079b22f2b"><span><div id="31b629a6152d81d4af24f8b079b22f2b" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31b629a6152d81d4af24f8b079b22f2b" title="OpenAI 要上市了"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">OpenAI 要上市了</span></span></h4><div class="notion-text notion-block-31b629a6152d811c86dec2ac2e140e0a">OpenAI 正在筹备 IPO，目标在 2026 年第四季度，已聘请 Cooley 和 Wachtell Lipton Rosen &amp; Katz 律所协助。新任首席会计官和企业财务官已到位，预计融资后估值约为 8400 亿美元，2030 年实现盈利。与 OpenAI 类似，Anthropic 也在准备 IPO，但预计其盈利时间比 OpenAI 早两年。市场赔率显示，OpenAI 可能会先上市。</div><div class="notion-text notion-block-31b629a6152d81319755f397533101bf"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzkzNDQxOTU2MQ==&amp;mid=2247513929&amp;idx=1&amp;sn=9851682b56012b505ad7cd79daebe25b&amp;chksm=c341358600a6f7b834e98d305dbe8a2aab09d91e612980e94e2ae530f430869e42ed646c517d&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31b629a6152d819c8f79ceb9f83ce4c7">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31b629a6152d8197bf34c1b1e4c95c9f" data-id="31b629a6152d8197bf34c1b1e4c95c9f"><span><div id="31b629a6152d8197bf34c1b1e4c95c9f" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31b629a6152d8197bf34c1b1e4c95c9f" title="刚刚，GPT-5.4核心内幕炸裂剧透！或拥有永久记忆，极限推理狂飙"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">刚刚，GPT-5.4核心内幕炸裂剧透！或拥有永久记忆，极限推理狂飙</span></span></h4><div class="notion-text notion-block-31b629a6152d813bab3dfb1243865649">GPT-5.4即将发布，具备极限推理模式和可能的永久记忆功能，能够生成6000行代码，支持更长的上下文窗口（最高可达200万tokens）。该模型在推理速度和长时间任务表现上有显著提升，可能会改变AI的工作方式。OpenAI面临竞争压力，更新频率加快，力求达到10亿活跃用户目标。</div><div class="notion-text notion-block-31b629a6152d81b290abde23dcab75f4"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzI3MTA0MTk1MA==&amp;mid=2652680520&amp;idx=2&amp;sn=e3d1135523acc7409b51cd0d9c6ee0fc&amp;chksm=f010518351271c4ff09e50aa0d7206139800a3354d378e859e813c354366e83e7cf507b740bc&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31b629a6152d81d787aff8018e543c1f">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31b629a6152d810288f5f1ecd52bbb46" data-id="31b629a6152d810288f5f1ecd52bbb46"><span><div id="31b629a6152d810288f5f1ecd52bbb46" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31b629a6152d810288f5f1ecd52bbb46" title="爆赚193亿美元，博通成美股&quot;救命稻草&quot;"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">爆赚193亿美元，博通成美股&quot;救命稻草&quot;</span></span></h4><div class="notion-text notion-block-31b629a6152d813ca078fc1ed8488055">博通在2026财年第一财季实现营收193.11亿美元，同比增长29%，AI相关业务收入达到84亿美元，同比增长106%。公司预计第二财季营收约为220亿美元，并启动最高100亿美元的股票回购计划。尽管市场对AI投资的热情高涨，仍需警惕潜在的供应链风险和资本开支过热。博通的强劲业绩和回购计划提升了市场信心，股价上涨超过5%。</div><div class="notion-text notion-block-31b629a6152d81b8ada8cef4525d04be"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=Mjc1NjM3MjY2MA==&amp;mid=2691565450&amp;idx=1&amp;sn=9f3e03d96415c943af9442e9a92bff28&amp;chksm=a8433ae3ff76e625c1036b8097f7b93c96deb229ff37b324fb6a734d5417cd52c99b212f2932&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31b629a6152d81efbb2ae28b0c443f35">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31b629a6152d81549274f4527644156e" data-id="31b629a6152d81549274f4527644156e"><span><div id="31b629a6152d81549274f4527644156e" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31b629a6152d81549274f4527644156e" title="1秒1元！Seedance 2.0模型定价公布，短剧真的要被颠覆了"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">1秒1元！Seedance 2.0模型定价公布，短剧真的要被颠覆了</span></span></h4><div class="notion-text notion-block-31b629a6152d81fc8ccbc7899e7c82e2">Seedance 2.0模型定价公布，视频生成成本显著降低，1秒视频成本不超过1元，可能颠覆短剧制作模式。行业内公司已调整战略，非头部演员短剧项目被砍。AI技术的进步使得制作成本和人工干预减少，创意和经验变得更加重要。Seedance 2.0仍在测试阶段，未来发展备受期待。</div><div class="notion-text notion-block-31b629a6152d81628981cb140e836e29"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzIzNjc1NzUzMw==&amp;mid=2247872523&amp;idx=1&amp;sn=7d548361e4778b549c01ecf5ae926808&amp;chksm=e903794ebcda00f30503447b83571339a6aa29976ab87d7a13c2d01bf4fc6048a9bd3a06157e&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31b629a6152d81819f4be7f5770e549e">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31b629a6152d81fea989d73ba48e8e5b" data-id="31b629a6152d81fea989d73ba48e8e5b"><span><div id="31b629a6152d81fea989d73ba48e8e5b" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31b629a6152d81fea989d73ba48e8e5b" title="阿里内部信批准林俊旸辞职，周靖人接管通义实验室"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">阿里内部信批准林俊旸辞职，周靖人接管通义实验室</span></span></h4><div class="notion-text notion-block-31b629a6152d81c68234cab27a014dc5">阿里巴巴CEO吴泳铭在内部信中批准林俊旸辞职，周靖人接管通义实验室，并成立基础模型支持小组以协调资源。林俊旸因与管理层分歧于3月3日辞职，随后团队成员对未来方向提出质疑。吴泳铭强调将继续加大对AI研发的投入和开源模型策略的坚持。</div><div class="notion-text notion-block-31b629a6152d819db9bbe35dd689e16d"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=Mjc1NjM3MjY2MA==&amp;mid=2691565447&amp;idx=1&amp;sn=5537ef9bfcace286193929f1190680d2&amp;chksm=a80d78e1ae348014d270e281e6c573ebd60249f3af1231846bf8d59d1b630faaeb8f6c27bc83&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31b629a6152d81599a3fe226e8a10427">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31b629a6152d81f2a615f87cb059d2ff" data-id="31b629a6152d81f2a615f87cb059d2ff"><span><div id="31b629a6152d81f2a615f87cb059d2ff" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31b629a6152d81f2a615f87cb059d2ff" title="跳过88%专家，保住97%性能！MoE推理的正确玩法| CVPR&#x27;26"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">跳过88%专家，保住97%性能！MoE推理的正确玩法| CVPR&#x27;26</span></span></h4><div class="notion-text notion-block-31b629a6152d81aa95bff453730f2e34">MoDES提出了一种动态专家跳过机制，能够在推理过程中跳过88%的冗余专家，同时保持97%的性能，显著提升推理效率。该方法通过建模不同层专家对输出的全局贡献和模态差异，解决了多模态模型中专家跳过导致性能下降的问题。实验结果显示，MoDES在多个主流多模态MoE模型上表现优异，证明高比例跳过并不必然导致性能崩溃，为大模型推理效率优化提供了新的思路。</div><div class="notion-text notion-block-31b629a6152d81229654d66b38a903cf"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzI3MTA0MTk1MA==&amp;mid=2652680520&amp;idx=3&amp;sn=2a1af6e551305a2949f8679ca4f4ff83&amp;chksm=f0e55dc859a1579276ff0f021f04c2072bcbd7a0783350dddfc2d349375faff1343057d57577&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31b629a6152d8181b22cfbd689f52ed3">     </div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-31b629a6152d811da0d7f36512ca3d7d" data-id="31b629a6152d811da0d7f36512ca3d7d"><span><div id="31b629a6152d811da0d7f36512ca3d7d" class="notion-header-anchor"></div><a class="notion-hash-link" href="#31b629a6152d811da0d7f36512ca3d7d" title="ICLR 2026 | OpenAI打广告后，如何成为爆款？CMU提出AutoGEO解密流量密码"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title">ICLR 2026 | OpenAI打广告后，如何成为爆款？CMU提出AutoGEO解密流量密码</span></span></h4><div class="notion-text notion-block-31b629a6152d81c1b961c1e541f475bf">AI搜索引擎逐渐取代传统搜索方式，内容的曝光不再仅依赖于排名，而是取决于被AI引用的方式和频率。卡内基梅隆大学提出的AutoGEO框架，通过分析生成式引擎的偏好规则，优化网页内容以提高被引用的可能性。研究表明，AutoGEO显著提升了网页在答案中的可见性，同时保持了生成引擎的效用。未来，内容创作者可能需要根据不同引擎的偏好，调整写作风格以适应新的搜索生态。</div><div class="notion-text notion-block-31b629a6152d81849e95fcd37db508db"><span class="notion-blue"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="http://mp.weixin.qq.com/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2651019827&amp;idx=2&amp;sn=fe9993389519c078b27355d889b31277&amp;chksm=85bcc8269268d0c023e0202dce3fafe217c9c36d204ff56703c47c31e9b73c69980ddc4c89b5&amp;scene=0#rd">🔗访问原文</a></span></div><div class="notion-text notion-block-31b629a6152d817e8cdcf8a6e8f0642a">     </div></main></div>]]></content:encoded>
        </item>
    </channel>
</rss>