RAGU：一種採用緊湊型領域自適應大語言模型的多步圖RAG引擎

RAGU: A Multi-Step GraphRAG Engine with a Compact Domain-Adapted LLM

July 13, 2026

作者: Mikhail Komarov, Ivan Bondarenko, Stanislav Shtuka, Oleg Sedukhin, Roman Shuvalov, Yana Dementyeva, Matvey Solovyov, Nikolay O. Nikitin

cs.AI

摘要

圖形檢索增強生成（GraphRAG）能為大型語言模型注入結構化知識，但現有系統僅透過單次提取過程建構知識圖譜，導致實體雜訊多且檢索脆弱。RAGU 作為開源模組化圖形檢索增強生成引擎，透過將提取與整合分離來解決此問題：實體與關係經過兩階段類型化提取、基於 DBSCAN 的去重、大型語言模型摘要以及 Leiden 社群檢測。一個關鍵洞察推動了緊湊型提取器的設計：管線內大型語言模型所需的技能——理解、提取、基於上下文推理——這些語言技能隨模型規模增長的幅度遠低於事實性世界知識。據此，我們訓練了 Meno-Lite-0.1，這是一個針對語言技能優化的 7B 模型，在知識圖譜建構上超越 Qwen2.5-32B（相對和諧平均提升 12.5%），並在英文圖形檢索增強生成任務上與其持平。在 GraphRAG-Bench（醫學）上，RAGU 在每個事實層級檢索到最完整的上下文（證據召回率最高達 0.84，對比 ≤0.76），並在合成任務上超越 HippoRAG2；而在多跳事實性問答中，HippoRAG2 的明顯優勢被證明主要是答案格式造成的假象。RAGU 可通過 `pip install graph_ragu` 安裝，運行於單張 GPU，並以 MIT 授權釋出。原始碼公開於 https://github.com/RaguTeam/RAGU，而 Meno-Lite-0.1 模型可從 https://huggingface.co/bond005/meno-lite-0.1 取得。

English

Graph retrieval-augmented generation (GraphRAG) enhances large language models with structured knowledge, yet existing systems construct knowledge graphs in a single extraction pass, producing noisy entities and brittle retrieval. RAGU, an open-source modular GraphRAG engine, addresses this by separating extraction from consolidation: entities and relations pass through two-stage typed extraction, DBSCAN-backed deduplication, LLM summarization, and Leiden community detection. A key insight motivates a compact extractor: the skills an in-pipeline LLM needs - comprehension, extraction, reasoning over context - are language skills that grow only weakly with model size, unlike factual world knowledge. Accordingly, we train Meno-Lite-0.1, a 7B model optimized for language skills, which outperforms Qwen2.5-32B on knowledge-graph construction (+12.5% relative harmonic mean) and matches it on English GraphRAG tasks. On GraphRAG-Bench (Medical), RAGU retrieves the most complete context at every factoid level (evidence recall up to 0.84 vs. leq0.76) and overtakes HippoRAG2 on synthesis tasks; on multi-hop factoid QA, the apparent HippoRAG2 advantage is shown to be largely an answer-format artifact. RAGU is installable via pip install graph_ragu, runs on a single GPU, and is released under MIT. The source code is publicly available at https://github.com/RaguTeam/RAGU, and the Meno-Lite-0.1 model can be obtained from https://huggingface.co/bond005/meno-lite-0.1.