Pith. sign in

REVIEW 2 major objections 5 minor 37 references

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

T0 review · 2 major / 5 minor · reviewed 2026-07-14 · grok-4.5

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A multi-step GraphRAG pipeline with a compact 7B language-skill model builds cleaner knowledge graphs and retrieves more complete context than single-pass systems.

desk verdict Solid open GraphRAG systems paper: multi-step consolidation really does lift evidence recall and synthesis; the 7B extractor works; scaling story is thin motivation, not load-bearing. read the letter →

arxiv 2607.11683 v1 pith:G334577F submitted 2026-07-13 cs.CL cs.AI

Mikhail Komarov , Ivan Bondarenko , Stanislav Shtuka , Oleg Sedukhin , Roman Shuvalov , Yana Dementyeva , Matvey Solovyov , Nikolay O. Nikitin This is my paper · ORCID
classification cs.CLcs.AI
keywords GraphRAGknowledgeconstructionretrieval-augmentedgenerationlanguageskillsscalingcompactLLMsmulti-stepconsolidationentitydeduplicationcommunitydetection
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Existing GraphRAG systems extract entities and relations in one noisy pass, so the resulting graphs are brittle. This paper argues that the LLM inside such a pipeline mainly needs language skills—comprehension, extraction, and reasoning over provided context—which grow only weakly with model size, unlike factual world knowledge. It therefore introduces RAGU, which separates extraction from consolidation through two-stage typed extraction, clustering, summarization, and community detection, and pairs it with Meno-Lite-0.1, a 7B model trained for those skills. On medical GraphRAG benchmarks the system attains the highest evidence recall at every factoid level and overtakes chain-following systems on synthesis tasks, while the multi-hop lead of those systems largely vanishes once answer format is controlled. The result is a single-GPU, pip-installable engine that matches larger extractors at far lower cost.

What carries the argument

Multi-step consolidation: two-stage schema-constrained extraction, DBSCAN-backed deduplication and LLM summarization of entities and relations, then Leiden community detection. This separation of extraction from consolidation, together with the language-skill 7B extractor, is what produces cleaner, more connected graphs.

What would settle it

Measure extraction and in-context multi-hop scores across several model families and sizes; if those language-skill scores rise nearly as steeply with parameter count as pure factual-quiz scores, the scaling hypothesis fails.

Watch

Extended reading notes

Core claim

The skills an LLM needs inside a GraphRAG pipeline—comprehension, extraction, and reasoning over context—are language skills that scale only weakly with size, unlike world knowledge. A 7B model optimized for those skills therefore outperforms a 32B general model on knowledge-graph construction, and a multi-step pipeline that consolidates extractions before community detection yields the most complete retrieved context and leads on synthesis tasks.

Load-bearing premise

The claim that language skills grow only weakly with model size rests on one model family and two tasks; if that pattern fails to generalize, the case for a compact extractor collapses.

Editorial extensions

If this is right

  • A local 7B extractor can replace large API models for GraphRAG indexing without quality loss.
  • Multi-step consolidation systems are preferable when answers must synthesize broad context; chain-traversal systems remain stronger for precise multi-hop fact lookup.
  • High-quality GraphRAG indexing becomes practical on a single consumer GPU.
  • Apparent multi-hop rankings on factoid QA can reverse once answer format is controlled.
  • Reducing extraction noise before community detection raises evidence recall across factoid levels.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same language-skill versus world-knowledge split may justify compact models for other RAG sub-tasks such as query rewriting or reranking.
  • If consolidation dominates quality, further gains may come more from better clustering and schema design than from larger extractors.
  • Schema-constrained two-stage extraction should transfer to new domains once type inventories are adapted, not only the original news/medical setting.
  • Order-of-magnitude cost reduction at corpus scale makes private on-prem GraphRAG feasible for smaller organizations.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 5 minor

Summary. The paper presents RAGU, a modular multi-step GraphRAG engine that separates entity/relation extraction from consolidation (two-stage typed extraction under NEREL, DBSCAN-backed deduplication and LLM summarization, Leiden communities), together with Meno-Lite-0.1, a 7B model fine-tuned for in-pipeline language skills rather than parametric world knowledge. The central claims are that (i) multi-step consolidation yields cleaner graphs and higher Evidence Recall than single-pass systems (LightRAG) and competitive or superior synthesis metrics versus HippoRAG 2 on GraphRAG-Bench (Medical), (ii) the apparent HippoRAG 2 lead on multi-hop factoid QA is largely an answer-format artifact once terse prompts are used, and (iii) a compact skill-oriented 7B extractor can match or beat much larger models on KG construction (+12.5% relative HM vs Qwen2.5-32B) and end-to-end GraphRAG tasks. Evaluation fixes the answer LLM (gpt-4o-mini) across systems, reports ablations on ICL/validation and extractor size (3B–14B), and releases code, model, and tests under open licenses.

Significance. If the controlled results hold, the work supplies a practical, single-GPU GraphRAG stack whose multi-step consolidation demonstrably improves context completeness (Evidence Recall up to 0.84 vs ≤0.76) and synthesis quality, while the format-controlled multi-hop analysis correctly diagnoses a common evaluation confound. The open release (pip-installable package, ~374 tests, mock LLM server, MIT/Apache artifacts) and engineering comparison (Pydantic validation vs eval(), swappable storage tiers) are concrete contributions that lower the barrier for reproducible GraphRAG research and deployment. The language/world-knowledge scaling hypothesis is only weakly supported, but the empirical pipeline gains do not depend on it; the paper therefore advances both systems engineering and evaluation practice even if the theoretical motivation remains provisional.

major comments (2)
  1. §1 and Figure 1: the language/world-knowledge scaling hypothesis is demonstrated only on the Qwen2.5-Instruct family (CheGeKa vs MultiQ). Because this hypothesis is used to justify investing in a compact 7B extractor rather than larger models, the manuscript should either (a) add at least one additional family or skill suite, or (b) explicitly demote the claim to a motivating observation and rest the contribution solely on the controlled pipeline results (which already show AC shifts ≤1.5 pp across 3B–14B extractors in Appendix B). As written, the theoretical framing is broader than the evidence.
  2. §3.4 / Table 3 and Limitations: Meno-Lite-0.1’s large standalone IE advantage (+12.5% HM) compresses to ≤1 pp on end-to-end GraphRAG-Bench QA across every pipeline tested. The paper correctly notes this, but the abstract and introduction still lead with the IE win as a primary selling point. Clarify in the main text that the extractor’s value is cost/efficiency (7B-class extraction quality) rather than end-to-end QA lift once consolidation is present; otherwise readers may over-attribute the GraphRAG-Bench cross-over to the model rather than the multi-step pipeline.
minor comments (5)
  1. Table 1 / Figure 3: report absolute Evidence Recall numbers (0.84 etc.) consistently in both table and figure captions; the prose uses 0.84 while the figure axis is percentage.
  2. §2.1 Step 3: DBSCAN hyperparameters (eps, min_samples) are free parameters; state the defaults used for all reported runs or note that they were held fixed across systems.
  3. Appendix A: the HippoRAG 2 engineering critique is useful but lengthy; a short pointer that the comparison is against a fixed commit (already given) would suffice for the main narrative.
  4. Limitations: the residual NEREL schema/domain overlap for the IE benchmark is disclosed; a one-sentence reminder in §3.4 would help readers who skip the Limitations section.
  5. Typos / consistency: “HippoRAG2” vs “HippoRAG 2”, “gpt-oss-20b” capitalization, and “~374 tests” vs “∼374” should be uniform.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: empirical GraphRAG claims rest on external benchmarks and controlled ablations, not on self-defined predictions or load-bearing self-citation chains.

full rationale

RAGU’s central claims are empirical system comparisons (Evidence Recall, Answer Correctness, Coverage, Faithfulness) on public benchmarks—GraphRAG-Bench (Medical), BioASQ, MuSiQue, 2WikiMultiHopQA—against open competitors (HippoRAG 2, LightRAG) under a shared answer-generation LLM. The multi-step pipeline (two-stage typed extraction, DBSCAN summarization, Leiden communities) is an engineering design, not a mathematical derivation that reduces outputs to fitted free parameters. The language/world-knowledge scaling hypothesis (Figure 1, Qwen2.5-Instruct on CheGeKa vs MultiQ) is an empirical motivation for a compact extractor; it is not used as a uniqueness theorem or as a fitted input renamed as prediction, and ablations (Appendix B) already show end-to-end AC shifts ≤1.5 pp across 3B–14B extractors, so the pipeline result does not depend on that hypothesis by construction. Self-citations (Meno-Lite model card, NEREL-instruct / NEREL-bench by Bondarenko) document released artifacts and a disclosed schema-overlap caveat on the IE benchmark; they do not define the evaluation metrics or force the GraphRAG-Bench cross-over. No equation equates a claimed prediction to its own fit; no uniqueness result is imported from the authors to forbid alternatives. The paper is therefore self-contained against external benchmarks with no significant circularity.

Assumptions & free parameters 2 free parameters · 4 assumptions · 2 invented entities

This is an empirical systems paper, not a first-principles derivation. The load-bearing premises are domain assumptions about GraphRAG evaluation practice and the language-skill scaling hypothesis; free parameters are ordinary engineering hyperparameters rather than constants fitted to force a theoretical claim. No new physical or mathematical entities are postulated.

free parameters (2)
  • DBSCAN clustering hyperparameters (eps, min_samples)
    Used inside EntitySummarizer/RelationSummarizer to group duplicate mentions; values are not reported as fitted to the final QA metrics but still affect graph cleanliness.
  • Chunking strategy and overlap sizes
    Simple/Semantic/SmartSemantic chunkers are configurable; choice influences entity density and is not exhaustively ablated against final AC.
assumptions (4)
  • ad hoc to paper Language skills required inside a RAG pipeline (comprehension, extraction, context reasoning) scale only weakly with model size, while world knowledge scales steeply.
    Motivated by Figure 1 on the Qwen2.5 family alone; treated as justification for investing in a 7B skill-tuned extractor (§1, Limitations).
  • domain assumption Fixing the answer-generation LLM (gpt-4o-mini) isolates differences in graph-construction quality.
    Standard experimental control in the GraphRAG literature; used throughout §3.
  • domain assumption NEREL schema (29 entity types, 49 relation types) is an adequate typed vocabulary for the evaluated domains.
    Extraction is constrained to NEREL; Limitations note possible domain mismatch outside Russian news-style text.
  • domain assumption Leiden community detection plus LLM community reports improve retrieval for synthesis tasks.
    Inherited from Microsoft GraphRAG and used as the final consolidation stage.
invented entities (2)
  • Meno-Lite-0.1 independent evidence
    purpose: 7B extractor optimized for language skills rather than parametric world knowledge, intended as a drop-in replacement for larger models inside GraphRAG pipelines.
    New model checkpoint released by the authors; independent evidence is the public HF weights and the IE/MERA numbers, but the skill-vs-knowledge framing is paper-specific.
  • RAGU multi-step consolidation pipeline independent evidence
    purpose: Explicit separation of typed extraction, DBSCAN-backed deduplication/summarization, and Leiden communities to reduce noise before retrieval.
    The pipeline architecture itself is the main systems contribution; it is fully specified and open-sourced.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RAGU: A Multi-Step GraphRAG Engine with a Compact Domain-Adapted LLM." pith.science (2026). https://pith.science/paper/G334577F

@misc{pith2026260711683,
  author       = {Pith},
  title        = {Pith review of: RAGU: A Multi-Step GraphRAG Engine with a Compact Domain-Adapted LLM},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G334577F}},
  note         = {Machine review of arXiv:2607.11683}
}
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 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. $\leq$0.76) 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 $\texttt{pip install graph_ragu}$, 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.

Figures

Figures reproduced from arXiv: 2607.11683 by the authors.

Figure 1
Figure 1. Effect of model size on world-knowledge (CheGeKa) vs. language-skill (MultiQ) tasks in the Qwen2.5-Instruct family (F1 scores on MERA (Fenogenova et al., 2024)). CheGeKa F1 grows 21.1× from 0.5 B to 72 B; MultiQ only 4×. Log-linear slopes: 0.65 vs. 0.26. skills. This prediction motivates a compact extrac￾tor. We address all three obstacles with two artifacts, released under open licenses: 1. Meno-Lite-0.1, a 7 B mod… view at source ↗
Figure 2
Figure 2. End-to-end indexing pipeline. Documents are chunked, entities and relations are extracted under the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Cross-over by task complexity on GraphRAG-Bench (Medical). All three systems use Meno-Lite-0.1 [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Knowledge graph built from the Ritchie pas [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 3 linked inside Pith

  1. [1]

    L ight RAG : Simple and Fast Retrieval-Augmented Generation

    Guo, Zirui and Xia, Lianghao and Yu, Yanhua and Ao, Tu and Huang, Chao. L ight RAG : Simple and Fast Retrieval-Augmented Generation. Findings of the Association for Computational Linguistics: EMNLP 2025. 2025. doi:10.18653/v1/2025.findings-emnlp.568

    doi:10.18653/v1/2025.findings-emnlp.568 2025

  2. [2]

    From Local to Global: A Graph

    Edge, Darren and Trinh, Ha and Cheng, Newman and Bradley, Joshua and Chao, Alex and Mody, Apurva and Truitt, Steven and Larson, Jonathan , year=. From Local to Global: A Graph. arXiv preprint arXiv:2404.16130 , eprint=

    arXiv

  3. [3]

    Guti. Hippo. Advances in Neural Information Processing Systems (NeurIPS) , year=

  4. [4]

    Guti. From. Forty-second International Conference on Machine Learning , year=

  5. [5]

    Wikontic: Constructing W ikidata-Aligned, Ontology-Aware Knowledge Graphs with Large Language Models

    Chepurova, Alla and Bulatov, Aydar and Burtsev, Mikhail and Kuratov, Yuri. Wikontic: Constructing W ikidata-Aligned, Ontology-Aware Knowledge Graphs with Large Language Models. Proceedings of the 19th Conference of the E uropean Chapter of the A ssociation for C omputational L inguistics (Volume 1: Long Papers). 2026. doi:10.18653/v1/2026.eacl-long.388

    doi:10.18653/v1/2026.eacl-long.388 2026

  6. [6]

    When to use Graphs in

    Xiang, Zhishang and Wu, Chuanjie and Zhang, Qinggang and Chen, Shengyuan and Hong, Zijin and Huang, Xiao and Su, Jinsong , booktitle=. When to use Graphs in

  7. [7]

    Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021). 2021

    2021

  8. [8]

    Popovi. chr. Proceedings of the Tenth Workshop on Statistical Machine Translation , month = sep, year =

Show all 37 references
  1. [9]

    Popovi. chr. Proceedings of the Second Conference on Machine Translation , month = sep, year =. doi:10.18653/v1/W17-4770 , pages =

    doi

  2. [10]

    A Call for Clarity in Reporting

    Post, Matt , booktitle =. A Call for Clarity in Reporting. 2018 , address =

    2018

  3. [11]

    doi:10.5281/zenodo.5371628 , url =

    Gao, Leo and Tow, Jonathan and Abbasi, Baber and Biderman, Stella and Black, Sid and DiPofi, Anthony and Foster, Charles and Golding, Laurence and Hsu, Jeffrey and Le Noac'h, Alain and Li, Haonan and McDonell, Kyle and Muennighoff, Niklas and Ociepa, Chris and Phang, Jason and...

    doi

  4. [12]

    Hugging Face Datasets , howpublished =

    NEREL-bench: A Benchmark for Evaluating LLMs on Russian Knowledge Graph Construction Tasks , author =. Hugging Face Datasets , howpublished =. 2026 , publisher =

    2026

  5. [13]

    Hugging Face Datasets , howpublished =

    Bondarenko, Ivan , title =. Hugging Face Datasets , howpublished =. 2026 , publisher =

    2026

  6. [14]

    R agu T eam at S em E val-2026 Task 8: Meno and F riends in a Judge-Orchestrated LLM Ensemble for Faithful Multi-Turn Response Generation

    Bondarenko, Ivan and Derunets, Roman and Sedukhin, Oleg and Komarov, Mikhail and Chernov, Ivan and Kulakov, Mikhail. R agu T eam at S em E val-2026 Task 8: Meno and F riends in a Judge-Orchestrated LLM Ensemble for Faithful Multi-Turn Response Generation. Proceedings of the 20...

    2026 doi

  7. [15]

    MERA : A Comprehensive LLM Evaluation in R ussian

    Fenogenova, Alena and Chervyakov, Artem and Martynov, Nikita and Kozlova, Anastasia and Tikhonova, Maria and Akhmetgareeva, Albina and Emelyanov, Anton and Shevelev, Denis and Lebedev, Pavel and Sinev, Leonid and Isaeva, Ulyana and Kolomeytseva, Katerina and Moskovskiy, Daniil...

    2024 doi

  8. [16]

    Retrieval-Augmented Generation for Knowledge-Intensive

    Lewis, Patrick and Perez, Ethan and Piktus, Aleksandra and Petroni, Fabio and Karpukhin, Vladimir and Goyal, Naman and K\". Retrieval-Augmented Generation for Knowledge-Intensive. Proceedings of the 34th International Conference on Neural Information Processing Systems , artic...

    2020

  9. [17]

    2024 , eprint=

    Yang, An and Yang, Baosong and Zhang, Beichen and Hui, Binyuan and Zheng, Bo and Yu, Bowen and Li, Chengyuan and Liu, Dayiheng and Huang, Fei and Wei, Haoran and Lin, Huan and Yang, Jian and Tu, Jianhong and Zhang, Jianwei and Yang, Jianxin and Yang, Jiaxi and Zhou, Jingren an...

    2024

  10. [18]

    2023 , journal=

    Towards General Text Embeddings with Multi-stage Contrastive Learning , author=. 2023 , journal=. 2308.03281 , archivePrefix=

    2023 arXiv

  11. [19]

    Tang, Yixuan and Yang, Yi , booktitle=. Multi. 2024 , url=

    2024

  12. [20]

    2023 , pages =

    BioASQ-QA: A manually curated corpus for Biomedical Question Answering , journal =. 2023 , pages =

    2023

  13. [21]

    Transactions of the Association for Computational Linguistics

    ♫ M u S i Q ue: Multihop Questions via Single-hop Question Composition. Transactions of the Association for Computational Linguistics. 2022. doi:10.1162/tacl_a_00475

    2022 doi

  14. [22]

    Proceedings of the 28th International Conference on Computational Linguistics

    Constructing A Multi-hop. Proceedings of the 28th International Conference on Computational Linguistics. 2020. doi:10.18653/v1/2020.coling-main.580

    2020 doi

  15. [23]

    Sarthi, Parth and Abdullah, Salman and Tuli, Aditi and Khapra, Mitesh M and Rai, Pratyush , booktitle=

  16. [24]

    Transactions of the Association for Computational Linguistics , volume =

    Katsis, Yannis and Rosenthal, Sara and Fadnis, Kshitij and Gunasekara, Chulaka and Lee, Young-Suk and Popa, Lucian and Shah, Vraj and Zhu, Huaiyu and Contractor, Danish and Danilevsky, Marina , title =. Transactions of the Association for Computational Linguistics , volume =. ...

    2025 doi

  17. [25]

    Doklady Mathematics , year=

    Ruadapt: Cost-Effective Large Language Model Lingual Adaptation , author=. Doklady Mathematics , year=. doi:10.1134/S1064562425700322 , url =

    doi

  18. [26]

    arXiv preprint arXiv:2312.10997 , eprint=

    Retrieval-Augmented Generation for Large Language Models: A Survey , author=. arXiv preprint arXiv:2312.10997 , eprint=

    arXiv

  19. [27]

    Li, Zheyuan and Chen, Xu and Yu, Haojie and Lin, Hongyang and Lu, Zaixiang and Tang, Qiang and Huang, Fei and Han, Xianpei and Sun, Le and Li, Yong , booktitle=. Struct

  20. [28]

    2025 , isbn =

    Liang, Lei and Bo, Zhongpu and Gui, Zhengke and Zhu, Zhongshu and Zhong, Ling and Zhao, Peilong and Sun, Mengshu and Zhang, Zhiqiang and Zhou, Jun and Chen, Wenguang and Zhang, Wen and Chen, Huajun , title =. 2025 , isbn =. doi:10.1145/3701716.3715240 , booktitle =

    2025 doi

  21. [29]

    2025 , issue_date =

    Peng, Boci and Zhu, Yun and Liu, Yongchao and Bo, Xiaohe and Shi, Haizhou and Hong, Chuntao and Zhang, Yan and Tang, Siliang , title =. 2025 , issue_date =. doi:10.1145/3777378 , journal =

    2025 doi

  22. [30]

    Language Models are Few-Shot Learners , url =

    Brown, Tom and Mann, Benjamin and Ryder, Nick and Subbiah, Melanie and Kaplan, Jared D and Dhariwal, Prafulla and Neelakantan, Arvind and Shyam, Pranav and Sastry, Girish and Askell, Amanda and Agarwal, Sandhini and Herbert-Voss, Ariel and Krueger, Gretchen and Henighan, Tom a...

  23. [31]

    TAPE : Assessing Few-shot R ussian Language Understanding

    Taktasheva, Ekaterina and Fenogenova, Alena and Shevelev, Denis and Katricheva, Nadezhda and Tikhonova, Maria and Akhmetgareeva, Albina and Zinkevich, Oleg and Bashmakova, Anastasiia and Iordanskaia, Svetlana and Kurenshchikova, Valentina and Spiridonova, Alena and Artemova, E...

    2022 doi

  24. [32]

    Long Context Benchmark for the R ussian Language

    Churin, Igor and Apishev, Murat and Tikhonova, Maria and Shevelev, Denis and Bulatov, Aydar and Kuratov, Yuri and Averkiev, Sergei and Fenogenova, Alena. Long Context Benchmark for the R ussian Language. Proceedings of the 6th Workshop on Computational Approaches to Discourse,...

    2025 doi

  25. [33]

    Zheng, Lianmin and Chiang, Wei-Lin and Sheng, Ying and Zhu, Siyuan and Wu, Zhanghao and Zhuang, Yonghao and Lin, Zi and Li, Zhuohan and Li, Dacheng and Xing, Eric and Zhang, Hao and Gonzalez, Joseph E and Stoica, Ion , booktitle=. Judging

  26. [34]

    Asai, Akari and Wu, Zeqiu and Wang, Yiran and Sil, Avirup and Hajishirzi, Hannaneh , booktitle=. Self-

  27. [35]

    GitHub repository , year=

    Fast-. GitHub repository , year=

  28. [36]

    , journal=

    Edge, Darren J.L. , journal=. Lazy

  29. [37]

    and Zhang, Hao and Stoica, Ion , booktitle =

    Kwon, Woosuk and Li, Zhuohan and Zhuang, Siyuan and Sheng, Ying and Zheng, Lianmin and Yu, Cody Hao and Gonzalez, Joseph E. and Zhang, Hao and Stoica, Ion , booktitle =. Efficient Memory Management for Large Language Model Serving with. 2023 , url =

    2023

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.