<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[读 Transformer 论文，对做应用到底有什么用]]></title><description><![CDATA[<p dir="auto">问个可能有点基础的问题：做 AI 应用的人有必要读 Transformer 原论文吗？还是看科普就够了。</p>
]]></description><link>https://localaihub.com/topic/146/读-transformer-论文-对做应用到底有什么用</link><generator>RSS for Node</generator><lastBuildDate>Wed, 03 Jun 2026 19:21:19 GMT</lastBuildDate><atom:link href="https://localaihub.com/topic/146.rss" rel="self" type="application/rss+xml"/><pubDate>Sun, 10 May 2026 05:06:00 GMT</pubDate><ttl>60</ttl><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Mon, 11 May 2026 08:34:00 GMT]]></title><description><![CDATA[<p dir="auto">读完记得回来说哪段最卡。很多人都是卡在同几个地方。</p>
]]></description><link>https://localaihub.com/post/1476</link><guid isPermaLink="true">https://localaihub.com/post/1476</guid><dc:creator><![CDATA[米饭]]></dc:creator><pubDate>Mon, 11 May 2026 08:34:00 GMT</pubDate></item><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Mon, 11 May 2026 07:36:00 GMT]]></title><description><![CDATA[<p dir="auto">对。论文不是考试材料，是减少工程误判的工具。</p>
]]></description><link>https://localaihub.com/post/1475</link><guid isPermaLink="true">https://localaihub.com/post/1475</guid><dc:creator><![CDATA[nora]]></dc:creator><pubDate>Mon, 11 May 2026 07:36:00 GMT</pubDate></item><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Mon, 11 May 2026 05:52:00 GMT]]></title><description><![CDATA[<p dir="auto">这样说我有动力了。先读结构和动机，不从公式开始。</p>
]]></description><link>https://localaihub.com/post/1474</link><guid isPermaLink="true">https://localaihub.com/post/1474</guid><dc:creator><![CDATA[小李不困]]></dc:creator><pubDate>Mon, 11 May 2026 05:52:00 GMT</pubDate></item><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Mon, 11 May 2026 04:52:00 GMT]]></title><description><![CDATA[<p dir="auto">还有一个好处：能分辨供应商 PPT。很多“突破上下文限制”的说法，一问机制就露馅。</p>
]]></description><link>https://localaihub.com/post/1473</link><guid isPermaLink="true">https://localaihub.com/post/1473</guid><dc:creator><![CDATA[Luna]]></dc:creator><pubDate>Mon, 11 May 2026 04:52:00 GMT</pubDate></item><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Mon, 11 May 2026 04:33:00 GMT]]></title><description><![CDATA[<p dir="auto">我会建议团队做论文读书会，但目标不是学术汇报，是回答工程问题：为什么慢、为什么贵、为什么会忘。</p>
]]></description><link>https://localaihub.com/post/1472</link><guid isPermaLink="true">https://localaihub.com/post/1472</guid><dc:creator><![CDATA[小吴]]></dc:creator><pubDate>Mon, 11 May 2026 04:33:00 GMT</pubDate></item><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Mon, 11 May 2026 02:30:00 GMT]]></title><description><![CDATA[<p dir="auto">看需要。Transformer 原论文是地基；BERT/GPT 关系到预训练范式；MoE 关系到大模型扩展。应用开发不用全读，但别完全不碰。</p>
]]></description><link>https://localaihub.com/post/1471</link><guid isPermaLink="true">https://localaihub.com/post/1471</guid><dc:creator><![CDATA[陈一]]></dc:creator><pubDate>Mon, 11 May 2026 02:30:00 GMT</pubDate></item><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Mon, 11 May 2026 00:22:00 GMT]]></title><description><![CDATA[<p dir="auto">那是不是还要读 BERT、GPT、MoE？</p>
]]></description><link>https://localaihub.com/post/1470</link><guid isPermaLink="true">https://localaihub.com/post/1470</guid><dc:creator><![CDATA[普通网友A]]></dc:creator><pubDate>Mon, 11 May 2026 00:22:00 GMT</pubDate></item><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Sun, 10 May 2026 21:46:00 GMT]]></title><description><![CDATA[<p dir="auto">这句话适合贴在每个 RAG 项目门口。</p>
]]></description><link>https://localaihub.com/post/1469</link><guid isPermaLink="true">https://localaihub.com/post/1469</guid><dc:creator><![CDATA[半截薯条]]></dc:creator><pubDate>Sun, 10 May 2026 21:46:00 GMT</pubDate></item><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Sun, 10 May 2026 20:15:00 GMT]]></title><description><![CDATA[<p dir="auto">我读完最大感受是，模型不是数据库。它会生成，不是查表。</p>
]]></description><link>https://localaihub.com/post/1468</link><guid isPermaLink="true">https://localaihub.com/post/1468</guid><dc:creator><![CDATA[小满]]></dc:creator><pubDate>Sun, 10 May 2026 20:15:00 GMT</pubDate></item><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Sun, 10 May 2026 17:12:00 GMT]]></title><description><![CDATA[<p dir="auto">做应用最有用的是：你会少说一些玄学话。比如“把全部历史都塞进去不就好了”，读完会知道代价在哪。</p>
]]></description><link>https://localaihub.com/post/1467</link><guid isPermaLink="true">https://localaihub.com/post/1467</guid><dc:creator><![CDATA[阿航]]></dc:creator><pubDate>Sun, 10 May 2026 17:12:00 GMT</pubDate></item><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Sun, 10 May 2026 14:39:00 GMT]]></title><description><![CDATA[<p dir="auto">可以不死磕公式。先看输入怎么变成 token embedding，再看 self-attention 为什么每个位置要看其他位置。</p>
]]></description><link>https://localaihub.com/post/1466</link><guid isPermaLink="true">https://localaihub.com/post/1466</guid><dc:creator><![CDATA[melo]]></dc:creator><pubDate>Sun, 10 May 2026 14:39:00 GMT</pubDate></item><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Sun, 10 May 2026 11:42:00 GMT]]></title><description><![CDATA[<p dir="auto">我怕读不懂，尤其公式。</p>
]]></description><link>https://localaihub.com/post/1465</link><guid isPermaLink="true">https://localaihub.com/post/1465</guid><dc:creator><![CDATA[小李不困]]></dc:creator><pubDate>Sun, 10 May 2026 11:42:00 GMT</pubDate></item><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Sun, 10 May 2026 10:01:00 GMT]]></title><description><![CDATA[<p dir="auto">我觉得读一遍摘要和结构图很值。很多上下文、KV cache、长文本成本的问题，绕不开这个底层直觉。</p>
]]></description><link>https://localaihub.com/post/1464</link><guid isPermaLink="true">https://localaihub.com/post/1464</guid><dc:creator><![CDATA[Grace]]></dc:creator><pubDate>Sun, 10 May 2026 10:01:00 GMT</pubDate></item><item><title><![CDATA[Reply to 读 Transformer 论文，对做应用到底有什么用 on Sun, 10 May 2026 07:13:00 GMT]]></title><description><![CDATA[<p dir="auto">不用每个人都推公式，但至少要知道 attention 解决了什么问题，以及它为什么吃上下文成本。</p>
]]></description><link>https://localaihub.com/post/1463</link><guid isPermaLink="true">https://localaihub.com/post/1463</guid><dc:creator><![CDATA[陈一]]></dc:creator><pubDate>Sun, 10 May 2026 07:13:00 GMT</pubDate></item></channel></rss>