Two Jev passes helped our Graph RAG pipeline outperform the LLM rerankers we tested, with higher rec...

TL;DR · AI 摘要
Milvus团队通过双阶段Jev方法提升Graph RAG召回率,Recall@5达86.61%,比单阶段Jev提升6.6个百分点。
核心要点
- 双阶段Jev在MuSiQue和2Wiki数据集上将Recall@5从80.04%提升至86.61%
- 单阶段Jev存在关系筛选后文本仍可能无用的问题
- 双阶段Jev增加0.85秒模型调用时间但显著提升召回效果
结构提纲
按章节快速跳转。
- §技术背景
介绍Vector Graph RAG项目及LLM过滤的性能瓶颈
用Jev替代LLM过滤关系但召回率不足
分两阶段过滤和排序,提升召回效果
Jev筛选关系并整合向量检索结果
Jev对候选文本进行排序优化
- ·实验结果
双阶段Jev在2000个问题上提升6.6%召回率
思维导图
用一张图看清主题之间的关系。
查看大纲文本(无障碍 / 无 JS 友好)
- Vector Graph RAG优化
- 方法对比
- LLM过滤
- 单阶段Jev
- 双阶段Jev
- 实验数据
- Recall@5 86.61%
- 时间增加0.85s
金句 / Highlights
值得收藏与分享的关键句。
双阶段Jev在MuSiQue和2Wiki数据集上将Recall@5提升至86.61%
单阶段Jev存在筛选后文本仍可能无用的问题,如metformin案例
双阶段Jev增加0.85秒模型调用时间但显著提升召回效果
Milvus on X: "Two Jev passes helped our Graph RAG pipeline outperform the LLM rerankers we tested, with higher recall and less time spent on model calls. Earlier, we released 𝗩𝗲𝗰𝘁𝗼𝗿 𝗚𝗿𝗮𝗽𝗵 𝗥𝗔𝗚, an open-source project that uses Milvus for multi-hop retrieval without a separate graph da… / X
@milvusio
Two Jev passes helped our Graph RAG pipeline outperform the LLM rerankers we tested, with higher recall and less time spent on model calls. Earlier, we released 𝗩𝗲𝗰𝘁𝗼𝗿 𝗚𝗿𝗮𝗽𝗵 𝗥𝗔𝗚, an open-source project that uses Milvus for multi-hop retrieval without a separate graph database. During retrieval, an LLM filters the relationships found through graph expansion. That LLM step could involve hundreds of candidates, slowing retrieval. To address this, we tested two Jev-based alternatives and compared both with the original setup. 1️⃣ 𝗦𝗶𝗻𝗴𝗹𝗲-𝘀𝘁𝗮𝗴𝗲 𝗝𝗲𝘃 We replaced the LLM with Jev to filter the relationships. Filtering got faster, but recall still lagged behind our stronger LLM baseline on complex multi-hop questions. A selected relationship can still lead to unhelpful text. A relationship about metformin treating diabetes could come from a drug history article that says little about side effects. 2️⃣ 𝗧𝘄𝗼-𝘀𝘁𝗮𝗴𝗲 𝗝𝗲𝘃 • 𝗦𝘁𝗮𝗴𝗲 𝟭: Jev selects graph relationships. Milvus retrieve their source passages, add results from direct vector search, and remove duplicates. • 𝗦𝘁𝗮𝗴𝗲 𝟮: Jev ranks those passages by how useful they are for answering the question. Across 2,000 questions from MuSiQue and 2Wiki, two-stage Jev 1.13 improved average Recall@5 from 80.04% to 86.07% over single-stage Jev, adding 0.85 seconds of model-call time. Jev 1.14 reached 86.61%. Check out the repo to see how we pushed Vector Graph RAG to the Pareto frontier of speed and quality! 👇
github.com/zilliztech/vec…
3:30 PM · Nov 6, 2026
·
542
Views
4
6
7