<?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[reranker 延迟太高，怎么不把体验拖死？]]></title><description><![CDATA[<p dir="auto">加 reranker 后首字从 1.2 秒变 3.8 秒，用户说像坏了。</p>
]]></description><link>https://localaihub.com/topic/73/reranker-延迟太高-怎么不把体验拖死</link><generator>RSS for Node</generator><lastBuildDate>Wed, 03 Jun 2026 19:24:28 GMT</lastBuildDate><atom:link href="https://localaihub.com/topic/73.rss" rel="self" type="application/rss+xml"/><pubDate>Mon, 04 May 2026 02:46:00 GMT</pubDate><ttl>60</ttl><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Tue, 05 May 2026 03:47:00 GMT]]></title><description><![CDATA[<p dir="auto">这就是要用数据调。别拿默认 top_80 当生产配置。</p>
]]></description><link>https://localaihub.com/post/378</link><guid isPermaLink="true">https://localaihub.com/post/378</guid><dc:creator><![CDATA[小乔同学]]></dc:creator><pubDate>Tue, 05 May 2026 03:47:00 GMT</pubDate></item><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Tue, 05 May 2026 01:06:00 GMT]]></title><description><![CDATA[<p dir="auto">今天降到 top_30，rerank 700ms 左右，答案没明显变差。</p>
]]></description><link>https://localaihub.com/post/377</link><guid isPermaLink="true">https://localaihub.com/post/377</guid><dc:creator><![CDATA[zeroOne]]></dc:creator><pubDate>Tue, 05 May 2026 01:06:00 GMT</pubDate></item><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Tue, 05 May 2026 00:47:00 GMT]]></title><description><![CDATA[<p dir="auto">API 方案注意数据合规。内部文档片段发出去之前先过审批。</p>
]]></description><link>https://localaihub.com/post/376</link><guid isPermaLink="true">https://localaihub.com/post/376</guid><dc:creator><![CDATA[leaf_1997]]></dc:creator><pubDate>Tue, 05 May 2026 00:47:00 GMT</pubDate></item><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Mon, 04 May 2026 23:52:00 GMT]]></title><description><![CDATA[<p dir="auto">本地 CPU 跑 reranker 很吃力。要么 GPU，要么更小模型，要么 API。</p>
]]></description><link>https://localaihub.com/post/375</link><guid isPermaLink="true">https://localaihub.com/post/375</guid><dc:creator><![CDATA[MingK]]></dc:creator><pubDate>Mon, 04 May 2026 23:52:00 GMT</pubDate></item><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Mon, 04 May 2026 21:10:00 GMT]]></title><description><![CDATA[<p dir="auto">UI 状态有用，但别用 UI 掩盖链路慢。1.9s 还是要优化。</p>
]]></description><link>https://localaihub.com/post/374</link><guid isPermaLink="true">https://localaihub.com/post/374</guid><dc:creator><![CDATA[小路灯]]></dc:creator><pubDate>Mon, 04 May 2026 21:10:00 GMT</pubDate></item><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Mon, 04 May 2026 18:21:00 GMT]]></title><description><![CDATA[<p dir="auto">我们把“正在查找来源”做成状态，用户能接受一点慢，但不能空白等。</p>
]]></description><link>https://localaihub.com/post/373</link><guid isPermaLink="true">https://localaihub.com/post/373</guid><dc:creator><![CDATA[米饭]]></dc:creator><pubDate>Mon, 04 May 2026 18:21:00 GMT</pubDate></item><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Mon, 04 May 2026 15:47:00 GMT]]></title><description><![CDATA[<p dir="auto">还有降级策略。低峰全量 rerank，高峰只 rerank 高风险问题。</p>
]]></description><link>https://localaihub.com/post/372</link><guid isPermaLink="true">https://localaihub.com/post/372</guid><dc:creator><![CDATA[阿航]]></dc:creator><pubDate>Mon, 04 May 2026 15:47:00 GMT</pubDate></item><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Mon, 04 May 2026 14:23:00 GMT]]></title><description><![CDATA[<p dir="auto">可以，但文档更新和权限变化要让缓存失效。</p>
]]></description><link>https://localaihub.com/post/371</link><guid isPermaLink="true">https://localaihub.com/post/371</guid><dc:creator><![CDATA[rootless]]></dc:creator><pubDate>Mon, 04 May 2026 14:23:00 GMT</pubDate></item><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Mon, 04 May 2026 12:54:00 GMT]]></title><description><![CDATA[<p dir="auto">reranker 可以缓存 query+doc_id 分数吗？热门问题会重复。</p>
]]></description><link>https://localaihub.com/post/370</link><guid isPermaLink="true">https://localaihub.com/post/370</guid><dc:creator><![CDATA[小潘同学]]></dc:creator><pubDate>Mon, 04 May 2026 12:54:00 GMT</pubDate></item><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Mon, 04 May 2026 10:01:00 GMT]]></title><description><![CDATA[<p dir="auto">top_80 太多了。先试 top_30。候选越多不一定越好。</p>
]]></description><link>https://localaihub.com/post/369</link><guid isPermaLink="true">https://localaihub.com/post/369</guid><dc:creator><![CDATA[林小北]]></dc:creator><pubDate>Mon, 04 May 2026 10:01:00 GMT</pubDate></item><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Mon, 04 May 2026 09:48:00 GMT]]></title><description><![CDATA[<p dir="auto">top_80 rerank 到 top_8。</p>
]]></description><link>https://localaihub.com/post/368</link><guid isPermaLink="true">https://localaihub.com/post/368</guid><dc:creator><![CDATA[zeroOne]]></dc:creator><pubDate>Mon, 04 May 2026 09:48:00 GMT</pubDate></item><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Mon, 04 May 2026 07:40:00 GMT]]></title><description><![CDATA[<p dir="auto">rerank 候选多少？</p>
]]></description><link>https://localaihub.com/post/367</link><guid isPermaLink="true">https://localaihub.com/post/367</guid><dc:creator><![CDATA[nora]]></dc:creator><pubDate>Mon, 04 May 2026 07:40:00 GMT</pubDate></item><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Mon, 04 May 2026 04:55:00 GMT]]></title><description><![CDATA[<p dir="auto">检索 120ms，rerank 1.9s，模型首 token 1.1s，其他杂项。</p>
]]></description><link>https://localaihub.com/post/366</link><guid isPermaLink="true">https://localaihub.com/post/366</guid><dc:creator><![CDATA[zeroOne]]></dc:creator><pubDate>Mon, 04 May 2026 04:55:00 GMT</pubDate></item><item><title><![CDATA[Reply to reranker 延迟太高，怎么不把体验拖死？ on Mon, 04 May 2026 03:32:00 GMT]]></title><description><![CDATA[<p dir="auto">先拆耗时。向量检索、rerank、LLM 首 token、流式输出分别是多少？</p>
]]></description><link>https://localaihub.com/post/365</link><guid isPermaLink="true">https://localaihub.com/post/365</guid><dc:creator><![CDATA[小乔同学]]></dc:creator><pubDate>Mon, 04 May 2026 03:32:00 GMT</pubDate></item></channel></rss>