@inproceedings{komarov-etal-2026-ragu,
title = "{RAGU}: A Multi-Step {G}raph{RAG} Engine with a Compact Domain-Adapted {LLM}",
author = "Komarov, Mikhail and
Bondarenko, Ivan and
Shtuka, Stanislav and
Sedukhin, Oleg and
Shuvalov, Roman and
Dementyeva, Yana and
Solovyov, Matvey and
Nikitin, Nikolay",
editor = "Agrawal, Sweta and
Abad, Alberto",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: System Demonstrations",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-demos.83/",
pages = "880--891",
ISBN = "979-8-89176-461-3",
abstract = "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 7 B 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. {\ensuremath{\leq}}0.76) and overtakes HippoRAG 2 on synthesis tasks; on multi-hop factoid QA, the apparent HippoRAG 2 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 Apache 2.0. Code and model: https://github.com/RaguTeam/RAGU."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="komarov-etal-2026-ragu">
<titleInfo>
<title>RAGU: A Multi-Step GraphRAG Engine with a Compact Domain-Adapted LLM</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mikhail</namePart>
<namePart type="family">Komarov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ivan</namePart>
<namePart type="family">Bondarenko</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Stanislav</namePart>
<namePart type="family">Shtuka</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Oleg</namePart>
<namePart type="family">Sedukhin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Roman</namePart>
<namePart type="family">Shuvalov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yana</namePart>
<namePart type="family">Dementyeva</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Matvey</namePart>
<namePart type="family">Solovyov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nikolay</namePart>
<namePart type="family">Nikitin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: System Demonstrations</title>
</titleInfo>
<name type="personal">
<namePart type="given">Sweta</namePart>
<namePart type="family">Agrawal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alberto</namePart>
<namePart type="family">Abad</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-461-3</identifier>
</relatedItem>
<abstract>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 7 B 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. \ensuremathłeq0.76) and overtakes HippoRAG 2 on synthesis tasks; on multi-hop factoid QA, the apparent HippoRAG 2 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 Apache 2.0. Code and model: https://github.com/RaguTeam/RAGU.</abstract>
<identifier type="citekey">komarov-etal-2026-ragu</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-demos.83/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>880</start>
<end>891</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T RAGU: A Multi-Step GraphRAG Engine with a Compact Domain-Adapted LLM
%A Komarov, Mikhail
%A Bondarenko, Ivan
%A Shtuka, Stanislav
%A Sedukhin, Oleg
%A Shuvalov, Roman
%A Dementyeva, Yana
%A Solovyov, Matvey
%A Nikitin, Nikolay
%Y Agrawal, Sweta
%Y Abad, Alberto
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: System Demonstrations
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-461-3
%F komarov-etal-2026-ragu
%X 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 7 B 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. \ensuremathłeq0.76) and overtakes HippoRAG 2 on synthesis tasks; on multi-hop factoid QA, the apparent HippoRAG 2 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 Apache 2.0. Code and model: https://github.com/RaguTeam/RAGU.
%U https://aclanthology.org/2026.emnlp-demos.83/
%P 880-891
Markdown (Informal)
[RAGU: A Multi-Step GraphRAG Engine with a Compact Domain-Adapted LLM](https://aclanthology.org/2026.emnlp-demos.83/) (Komarov et al., EMNLP 2026)
ACL
- Mikhail Komarov, Ivan Bondarenko, Stanislav Shtuka, Oleg Sedukhin, Roman Shuvalov, Yana Dementyeva, Matvey Solovyov, and Nikolay Nikitin. 2026. RAGU: A Multi-Step GraphRAG Engine with a Compact Domain-Adapted LLM. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 880–891, Budapest, Hungary. Association for Computational Linguistics.