<?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[RAG 片段太多，模型开始“平均所有观点”]]></title><description><![CDATA[<p dir="auto">我们 RAG TopK=15，答案经常像把所有片段平均了一下。新政策和旧政策混在一起，模型不敢明确说。</p>
]]></description><link>https://localaihub.com/topic/100/rag-片段太多-模型开始-平均所有观点</link><generator>RSS for Node</generator><lastBuildDate>Wed, 03 Jun 2026 17:51:04 GMT</lastBuildDate><atom:link href="https://localaihub.com/topic/100.rss" rel="self" type="application/rss+xml"/><pubDate>Wed, 06 May 2026 10:33:00 GMT</pubDate><ttl>60</ttl><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Thu, 07 May 2026 14:12:00 GMT]]></title><description><![CDATA[<p dir="auto">答案变稳，token 成本也会下来。</p>
]]></description><link>https://localaihub.com/post/783</link><guid isPermaLink="true">https://localaihub.com/post/783</guid><dc:creator><![CDATA[melo]]></dc:creator><pubDate>Thu, 07 May 2026 14:12:00 GMT</pubDate></item><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Thu, 07 May 2026 13:29:00 GMT]]></title><description><![CDATA[<p dir="auto">对，RAG 不是多塞资料比赛。</p>
]]></description><link>https://localaihub.com/post/782</link><guid isPermaLink="true">https://localaihub.com/post/782</guid><dc:creator><![CDATA[陈一]]></dc:creator><pubDate>Thu, 07 May 2026 13:29:00 GMT</pubDate></item><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Thu, 07 May 2026 11:39:00 GMT]]></title><description><![CDATA[<p dir="auto">我们要补元数据、降 TopK、加 rerank 和冲突处理。</p>
]]></description><link>https://localaihub.com/post/781</link><guid isPermaLink="true">https://localaihub.com/post/781</guid><dc:creator><![CDATA[木木不是木]]></dc:creator><pubDate>Thu, 07 May 2026 11:39:00 GMT</pubDate></item><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Thu, 07 May 2026 08:36:00 GMT]]></title><description><![CDATA[<p dir="auto">那模型应该说资料存在冲突，并按产品规则转人工或给出需要确认的信息。不要装确定。</p>
]]></description><link>https://localaihub.com/post/780</link><guid isPermaLink="true">https://localaihub.com/post/780</guid><dc:creator><![CDATA[林小北]]></dc:creator><pubDate>Thu, 07 May 2026 08:36:00 GMT</pubDate></item><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Thu, 07 May 2026 08:16:00 GMT]]></title><description><![CDATA[<p dir="auto">如果没有明确新旧，只是两个部门说法不同呢？</p>
]]></description><link>https://localaihub.com/post/779</link><guid isPermaLink="true">https://localaihub.com/post/779</guid><dc:creator><![CDATA[小周]]></dc:creator><pubDate>Thu, 07 May 2026 08:16:00 GMT</pubDate></item><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Thu, 07 May 2026 07:45:00 GMT]]></title><description><![CDATA[<p dir="auto">我们现在每个答案保留引用 chunk id，客服主管能点回原文。问题定位快很多。</p>
]]></description><link>https://localaihub.com/post/778</link><guid isPermaLink="true">https://localaihub.com/post/778</guid><dc:creator><![CDATA[阿宁]]></dc:creator><pubDate>Thu, 07 May 2026 07:45:00 GMT</pubDate></item><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Thu, 07 May 2026 05:16:00 GMT]]></title><description><![CDATA[<p dir="auto">还有一个坑：摘要器把多个片段压成一个“综合事实”，冲突信息被抹平，后面无法追溯。</p>
]]></description><link>https://localaihub.com/post/777</link><guid isPermaLink="true">https://localaihub.com/post/777</guid><dc:creator><![CDATA[leaf_1997]]></dc:creator><pubDate>Thu, 07 May 2026 05:16:00 GMT</pubDate></item><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Thu, 07 May 2026 02:42:00 GMT]]></title><description><![CDATA[<p dir="auto">内部可以长，用户答案短。模型内部决策和用户可见文案分开。</p>
]]></description><link>https://localaihub.com/post/776</link><guid isPermaLink="true">https://localaihub.com/post/776</guid><dc:creator><![CDATA[葡萄冰]]></dc:creator><pubDate>Thu, 07 May 2026 02:42:00 GMT</pubDate></item><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Thu, 07 May 2026 00:22:00 GMT]]></title><description><![CDATA[<p dir="auto">这会不会让回答变长？</p>
]]></description><link>https://localaihub.com/post/775</link><guid isPermaLink="true">https://localaihub.com/post/775</guid><dc:creator><![CDATA[普通网友A]]></dc:creator><pubDate>Thu, 07 May 2026 00:22:00 GMT</pubDate></item><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Wed, 06 May 2026 21:47:00 GMT]]></title><description><![CDATA[<p dir="auto">冲突片段要显式处理。可以让模型先列出候选证据和冲突，再按日期/优先级选择。</p>
]]></description><link>https://localaihub.com/post/774</link><guid isPermaLink="true">https://localaihub.com/post/774</guid><dc:creator><![CDATA[zeroOne]]></dc:creator><pubDate>Wed, 06 May 2026 21:47:00 GMT</pubDate></item><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Wed, 06 May 2026 18:43:00 GMT]]></title><description><![CDATA[<p dir="auto">rerank 也重要。向量召回相似，不代表能回答当前问题。</p>
]]></description><link>https://localaihub.com/post/773</link><guid isPermaLink="true">https://localaihub.com/post/773</guid><dc:creator><![CDATA[小蓝]]></dc:creator><pubDate>Wed, 06 May 2026 18:43:00 GMT</pubDate></item><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Wed, 06 May 2026 15:44:00 GMT]]></title><description><![CDATA[<p dir="auto">我们把 TopK 从 12 降到 5，准确率反而升了。因为少了很多相似但过期的段落。</p>
]]></description><link>https://localaihub.com/post/772</link><guid isPermaLink="true">https://localaihub.com/post/772</guid><dc:creator><![CDATA[melo]]></dc:creator><pubDate>Wed, 06 May 2026 15:44:00 GMT</pubDate></item><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Wed, 06 May 2026 14:29:00 GMT]]></title><description><![CDATA[<p dir="auto">chunk 元数据要有生效日期、部门、权限、文档类型。没有元数据，长上下文只会扩大混乱。</p>
]]></description><link>https://localaihub.com/post/771</link><guid isPermaLink="true">https://localaihub.com/post/771</guid><dc:creator><![CDATA[陈一]]></dc:creator><pubDate>Wed, 06 May 2026 14:29:00 GMT</pubDate></item><item><title><![CDATA[Reply to RAG 片段太多，模型开始“平均所有观点” on Wed, 06 May 2026 11:35:00 GMT]]></title><description><![CDATA[<p dir="auto">TopK 太多又没有排序解释，模型就会调和冲突。先处理资料版本，不是让模型猜哪个新。</p>
]]></description><link>https://localaihub.com/post/770</link><guid isPermaLink="true">https://localaihub.com/post/770</guid><dc:creator><![CDATA[林小北]]></dc:creator><pubDate>Wed, 06 May 2026 11:35:00 GMT</pubDate></item></channel></rss>