English(EN) Going from top-5 to top-20 retrieval makes every RAG answer 2.7x more expensive

检索量增加2.7倍导致RAG成本飙升，开发者被敦促预算token

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-06 12:00

在检索增强生成（RAG）系统中，将检索文档数量从5个增加到20个，每次回答的成本会增加2.7倍，这与所使用的具体LLM无关。成本的增加主要是由于发送给模型的token数量增加，而不是token的价格。开发者可以通过实施token预算或对检索到的块设置分数截止，而不是依赖固定的top-k结果数量来降低这些成本。 AI

影响使用RAG的开发者应重新评估其检索策略，以平衡答案质量和运营成本。

排序理由该条目讨论了RAG系统的特定优化技术及其成本影响，而不是新的模型发布或重大行业事件。

在 dev.to — LLM tag 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 23 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该条目讨论了RAG系统的特定优化技术及其成本影响，而不是新的模型发布或重大行业事件。
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: infra, product; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

dev.to — LLM tag TIER_1 English(EN) · Andrey Altrouter · 2026-11-06 12:00

Going from top-5 to top-20 retrieval makes every RAG answer 2.7x more expensive

<p>Your support bot answers from your docs. It gets a wrong answer in testing, so you raise <code>top_k</code> from 5 to 20 in the retriever settings, the next answer comes out right, and you ship it. Nobody notices that each answer now costs 2.7 times as much, because the settin…

本文如何被排名

报道来源 [1]

Going from top-5 to top-20 retrieval makes every RAG answer 2.7x more expensive

相关实体

相关话题