Pith. sign in

REVIEW 1 major objections 2 minor 1 cited by

Evaluating Chunking Strategies for Retrieval-Augmented Generation on Academic Texts

T0 review · 1 major / 2 minor · reviewed 2026-07-03 · grok-4.3

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

Pith's one-line read Cluster-based chunking did not outperform simpler fixed-size or recursive chunking for RAG on academic theses.

desk verdict Cluster chunking adds no advantage over fixed or recursive methods in this RAG test on theses, but the finding rests on RAGAs metrics the authors flag as limited in reliability. read the letter →

arxiv 2607.01852 v1 pith:AJ5RA3L2 submitted 2026-07-02 cs.IR cs.AIcs.CL

Valentin J. J. Kreileder , Johannes Reisinger , Andreas Fischer This is my paper · ORCID
classification cs.IRcs.AIcs.CL
keywords chunkingstrategiesretrieval-augmentedgenerationRAGsemanticRAGAsacademictextsevaluationframework
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

The paper tests whether cluster-based semantic chunking improves retrieval and answer quality in retrieval-augmented generation systems compared with fixed-size and recursive chunking. Experiments run on long structured academic theses and score outputs with the RAGAs framework. Results indicate no advantage for the cluster approach under the tested conditions. RAGAs faithfulness scores showed limited reliability, and performance varied sharply between fixed and document-specific questions, likely tied to document formatting and preprocessing.

What carries the argument

Direct comparison of three chunking methods (cluster-based semantic, fixed-size, recursive) inside a RAG pipeline, scored by RAGAs metrics on academic theses.

What would settle it

Human raters scoring the same RAG answers for faithfulness and relevance, finding cluster-based chunking produces clearly superior results on the tested theses.

Watch

Extended reading notes

Core claim

Under the tested configuration, cluster-based chunking did not outperform simpler strategies. RAGAs-based faithfulness shows limited reliability in this setup. Performance on fixed versus document-specific questions varied substantially, likely related to the formatting of documents and preprocessing.

Load-bearing premise

The RAGAs framework supplies reliable measurements of faithfulness and answer quality for academic theses and the chosen question types.

Editorial extensions

If this is right

  • Simpler chunking methods can be used without loss of quality for RAG on structured academic documents.
  • RAGAs faithfulness scores may not track human-perceived answer quality reliably on thesis-length texts.
  • Question design and document preprocessing choices affect measured RAG performance more than chunking method.
  • Fixed-size chunking remains a practical baseline for academic RAG applications.

Reading between the lines

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

  • Teams building RAG systems for academic content may save compute by defaulting to basic chunking unless domain-specific tests prove otherwise.
  • RAG evaluation frameworks require additional validation steps when applied to long-form technical documents.
  • The observed gap between fixed and document-specific questions points to a need for standardized test sets that control for document structure.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

1 major / 2 minor

Summary. The manuscript evaluates cluster-based semantic chunking against fixed-size and recursive chunking for RAG on long academic theses, using the RAGAs framework to measure retrieval and answer quality. It reports that cluster-based chunking did not outperform the simpler strategies under the tested configuration, while noting that RAGAs faithfulness exhibits limited reliability and that performance varies substantially with document formatting and preprocessing.

Significance. If the comparative result can be placed on firmer evidential ground, the finding would indicate that semantic clustering adds little value over simpler chunking for structured academic documents, which could simplify RAG pipelines in scholarly settings. The work supplies a domain-specific empirical comparison on real theses, a useful data point given the length and hierarchical structure of such texts.

major comments (1)
  1. [Abstract] Abstract: The central claim that cluster-based chunking did not outperform simpler strategies is grounded in RAGAs faithfulness and answer-quality scores, yet the abstract itself states that 'RAGAs based faithfulness shows limited reliability in this setup'. Without independent validation (human evaluation, alternative metrics, or controls for the noted formatting effects), this self-identified limitation renders the 'did not outperform' conclusion insecure.
minor comments (2)
  1. [Abstract] Abstract: The statement that 'performance on fixed versus document specific questions varied substantially' is left without quantitative detail or statistical test; adding effect sizes or significance levels would strengthen the observation.
  2. The manuscript does not describe the exact question-generation procedure, the number of theses, or the LLM used for generation and evaluation; these omissions hinder reproducibility even if the RAGAs limitation is addressed.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on the abstract. We address the concern about the security of our central claim below and propose revisions to better qualify our findings.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that cluster-based chunking did not outperform simpler strategies is grounded in RAGAs faithfulness and answer-quality scores, yet the abstract itself states that 'RAGAs based faithfulness shows limited reliability in this setup'. Without independent validation (human evaluation, alternative metrics, or controls for the noted formatting effects), this self-identified limitation renders the 'did not outperform' conclusion insecure.

    Authors: We agree that the abstract's phrasing could more clearly signal the tentative nature of the finding given the metric limitation we ourselves identify. The manuscript already states both the empirical observation ('Under the tested configuration, cluster-based chunking did not outperform simpler strategies') and the caveat ('RAGAs based faithfulness shows limited reliability in this setup') in the same paragraph, and the results section further discusses substantial performance variation linked to document formatting and preprocessing. Nevertheless, to strengthen the presentation, we will revise the abstract to frame the result more explicitly as an observation within the RAGAs framework rather than a definitive comparative outcome, and we will add a brief clause noting that independent validation would be valuable for future work. This change preserves the paper's honest reporting of both the result and its limitations without overstating generalizability. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: pure empirical evaluation with no derivations or self-referential reductions

full rationale

The paper conducts an empirical comparison of chunking strategies (cluster-based, fixed-size, recursive) on academic theses, measuring outcomes via the external RAGAs framework. No equations, fitted parameters, uniqueness theorems, or derivations are present that could reduce to inputs by construction. The central claim rests on experimental results rather than any self-definitional or self-citation chain. Limitations in RAGAs reliability are explicitly noted but do not create circularity in the reported comparisons.

Assumptions & free parameters 0 free parameters · 0 assumptions · 0 invented entities

Empirical evaluation study; contains no free parameters, mathematical axioms, or invented entities.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Evaluating Chunking Strategies for Retrieval-Augmented Generation on Academic Texts." pith.science (2026). https://pith.science/paper/AJ5RA3L2

@misc{pith2026260701852,
  author       = {Pith},
  title        = {Pith review of: Evaluating Chunking Strategies for Retrieval-Augmented Generation on Academic Texts},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AJ5RA3L2}},
  note         = {Machine review of arXiv:2607.01852}
}
read the original abstract

Retrieval-Augmented Generation (RAG) systems use the question-answering capabilities of Large Language Models (LLMs) to access information outside their parameters. We evaluate if cluster-based semantic chunking improves retrieval and answer quality compared to fixed-size and recursive chunking evaluating on long, structured academic theses using the Retrieval Augmented Generation Assessment (RAGAs) framework. RAGAs based faithfulness shows limited reliability in this setup. Performance on fixed versus document specific questions varied substantially, likely related to the formatting of documents and preprocessing. Under the tested configuration, cluster-based chunking did not outperform simpler strategies.

Figures

Figures reproduced from arXiv: 2607.01852 by the authors.

Figure 1
Figure 1. Used chunking methods. Cluster-Based Chunking: Cluster-based chunking is a method where we combine semantically similar sentences to more coherent chunks ( [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 4
Figure 4. Example: Fixed question retrieval snippet Context F1 Results For fixed questions, context f1 medians are 0 across all chunkers. These first five questions target general information in the preliminaries. Even after cleaning, preliminary artifacts and dot leaders survive, polluting both indexing and retrieval. For free questions, context f1 medians reach approximately 0.5 for recursive chunking and 0.3 for fixed-size… view at source ↗
Figure 2
Figure 2. TF-IDF cosine similarity with bigrams as terms. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figures from the paper (3 more)
Figure 3
Figure 3. Figure 3: Pipeline concerning the evaluation of chunking methods using RAGAS [PITH_FULL_IMAGE:figures/full_fig_p003_3.png]
Figure 5
Figure 5. Figure 5: Boxplots displaying context F1 scores for the evaluated chunking strategies. [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]
Figure 6
Figure 6. Figure 6: Boxplots displaying AQS scores for the evaluated chunking strategies. [PITH_FULL_IMAGE:figures/full_fig_p004_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Cross-Attention Calibrated Deduplication for Retrieval-Augmented Generation System

    cs.CL 2026-07 conditional novelty 5.5 of 10

    CACD deduplicates RAG chunks via cross-encoder scores, attention-entropy NIS, and majority vote, dropping ~9.75% of chunks on SQuAD faster than cosine filtering.

Reference graph

Works this paper leans on

73 extracted references · 73 canonical work pages · cited by 1 Pith paper

  1. [1]

    and Raman, Marc M

    Mirsky, Grace M. and Raman, Marc M. , month = nov, year =. Data. 2025 3rd. doi:10.1109/FLLM67465.2025.11390957 , abstract =

    work page doi:10.1109/fllm67465.2025.11390957 2025

  2. [2]

    Campello, Ricardo J. G. B. and Moulavi, Davoud and Zimek, Arthur and Sander, Jörg , month = jul, year =. Hierarchical. doi:10.1145/2733381 , file =

    work page doi:10.1145/2733381

  3. [3]

    Campello, Ricardo J. G. B. and Moulavi, Davoud and Sander, Joerg , editor =. Density-. Advances in. 2013 , keywords =. doi:10.1007/978-3-642-37456-2_14 , abstract =

    work page doi:10.1007/978-3-642-37456-2_14 2013

  4. [4]

    Database Syst

    ACM Trans. Database Syst. , author =. 2017 , pages =. doi:10.1145/3068335 , abstract =

    work page doi:10.1145/3068335 2017

  5. [5]

    Wang, Yining and Wang, Liwei and Li, Yuanzhi and He, Di and Liu, Tie-Yan and Chen, Wei , month = apr, year =. A. doi:10.48550/arXiv.1304.6480 , abstract =

    work page Pith review arXiv doi:10.48550/arxiv.1304.6480

  6. [6]

    Edge, Darren and Trinh, Ha and Cheng, Newman and Bradley, Joshua and Chao, Alex and Mody, Apurva and Truitt, Steven and Metropolitansky, Dasha and Ness, Robert Osazuwa and Larson, Jonathan , month = feb, year =. From. doi:10.48550/arXiv.2404.16130 , abstract =

    work page Pith review arXiv doi:10.48550/arxiv.2404.16130

  7. [7]

    IEEE Transactions on Big Data , author =

    Billion-. IEEE Transactions on Big Data , author =. 2021 , keywords =. doi:10.1109/TBDATA.2019.2921572 , abstract =

    work page doi:10.1109/tbdata.2019.2921572 2021

  8. [8]

    Improved

    Ioffe, Sergey , month = dec, year =. Improved. 2010. doi:10.1109/ICDM.2010.80 , abstract =

    work page doi:10.1109/icdm.2010.80 2010

Show all 73 references
  1. [9]

    WIREs Data Mining and Knowledge Discovery , author =

    Algorithms for hierarchical clustering: an overview , volume =. WIREs Data Mining and Knowledge Discovery , author =. 2012 , pages =. doi:10.1002/widm.53 , abstract =

    2012 doi

  2. [10]

    and Matsuura, Kenji , year =

    Willmott, Cort J. and Matsuura, Kenji , year =. Advantages of the mean absolute error (. Climate Research , publisher =

  3. [11]

    The American Journal of Psychology , author =

    The. The American Journal of Psychology , author =. 1904 , pages =. doi:10.2307/1412159 , language =

    1904 doi

  4. [12]

    scikit-learn , file =

  5. [13]

    Proceedings of

    Riedl, Martin and Biemann, Chris , editor =. Proceedings of. 2012 , pages =

    2012

  6. [14]

    Financial

    Yepes, Antonio Jimeno and You, Yao and Milczek, Jan and Laverde, Sebastian and Li, Renyu , month = mar, year =. Financial. doi:10.48550/arXiv.2402.05131 , abstract =

    doi

  7. [15]

    Kreileder, Valentin and Fischer, Andreas and Reisinger, Johannes , month = jan, year =. Thesis-

  8. [16]

    2026 , note =

    vibrantlabsai/ragas , copyright =. 2026 , note =

    2026

  9. [17]

    and Raghavan, Prabhakar and Schütze, Hinrich , month = jul, year =

    Manning, Christopher D. and Raghavan, Prabhakar and Schütze, Hinrich , month = jul, year =. Introduction to. Cambridge Aspire website , publisher =. doi:10.1017/CBO9780511809071 , note =

    doi

  10. [18]

    Document

    Wang, Zhitong and Gao, Cheng and Xiao, Chaojun and Huang, Yufei and Si, Shuzheng and Luo, Kangyang and Bai, Yuzhuo and Li, Wenhao and Duan, Tangjian and Lv, Chuancheng and Lu, Guoshan and Chen, Gang and Qi, Fanchao and Sun, Maosong , editor =. Document. Findings of the. 2025 ,...

    2025 doi

  11. [19]

    ACM Comput

    A. ACM Comput. Surv. , author =. 2023 , pages =. doi:10.1145/3606367 , abstract =

    2023 doi

  12. [20]

    1973 , pages =

    The Computer Journal , author =. 1973 , pages =. doi:10.1093/comjnl/16.1.30 , abstract =

    1973 doi

  13. [21]

    Qu, Renyi and Tu, Ruixuan and Bao, Forrest Sheng , editor =. Is. Findings of the. 2025 , pages =. doi:10.18653/v1/2025.findings-naacl.114 , abstract =

    2025 doi

  14. [22]

    Evaluation of

    Yu, Hao and Gan, Aoran and Zhang, Kai and Tong, Shiwei and Liu, Qi and Liu, Zhaofeng , year =. Evaluation of. doi:10.1007/978-981-96-1024-2_8 , abstract =

    doi

  15. [23]

    Proceedings of the 18th

    Es, Shahul and James, Jithin and Espinosa Anke, Luis and Schockaert, Steven , editor =. Proceedings of the 18th. 2024 , pages =. doi:10.18653/v1/2024.eacl-demo.16 , abstract =

    2024 doi

  16. [24]

    Information Storage and Retrieval , author =

    A definition of relevance for information retrieval , volume =. Information Storage and Retrieval , author =. 1971 , pages =. doi:10.1016/0020-0271(71)90024-6 , abstract =

    1971 doi

  17. [25]

    How to split text based on semantic similarity

  18. [26]

    Chase, Harrison , month = oct, year =

  19. [27]

    Günther, Michael and Mohr, Isabelle and Williams, Daniel James and Wang, Bo and Xiao, Han , month = jul, year =. Late. doi:10.48550/arXiv.2409.04701 , abstract =

    doi

  20. [28]

    Rethinking

    Bhat, Sinchana Ramakanth and Rudat, Max and Spiekermann, Jannis and Flores-Herr, Nicolas , month = may, year =. Rethinking. doi:10.48550/arXiv.2505.21700 , abstract =

    doi

  21. [29]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , author =

    n-. IEEE Transactions on Pattern Analysis and Machine Intelligence , author =. 1979 , keywords =. doi:10.1109/TPAMI.1979.4766902 , abstract =

    1979 doi

  22. [30]

    Étude comparative de la distribution florale dans une portion des

    Jaccard, Paul , year =. Étude comparative de la distribution florale dans une portion des. Bulletin de la Société Vaudoise des Sciences Naturelles , publisher =. doi:10.5169/seals-266450 , number =

    doi

  23. [31]

    Procedia Computer Science , author =

    Matching. Procedia Computer Science , author =. 2024 , keywords =. doi:10.1016/j.procs.2024.03.039 , abstract =

    2024 doi

  24. [32]

    Training

    Nussbaum, Zach and Duderstadt, Brandon , month = mar, year =. Training. doi:10.48550/arXiv.2502.07972 , abstract =

    doi

  25. [33]

    Thakur, Nandan and Reimers, Nils and Rücklé, Andreas and Srivastava, Abhishek and Gurevych, Iryna , month = aug, year =

  26. [34]

    Caspari, Laura and Dastidar, Kanishka Ghosh and Zerhoudi, Saber and Mitrovic, Jelena and Granitzer, Michael , month = jul, year =. Beyond. doi:10.48550/arXiv.2407.08275 , abstract =

    doi

  27. [35]

    Measurement of

    Wang, Jiapeng and Dong, Yihong , month = sep, year =. Measurement of. Information , publisher =. doi:10.3390/info11090421 , abstract =

    doi

  28. [36]

    Chroma , file =

    Chroma , url =. Chroma , file =

  29. [37]

    Engineering at Meta , month = mar, year =

    Faiss:. Engineering at Meta , month = mar, year =

  30. [38]

    The vector database to build knowledgeable

  31. [39]

    , editor =

    Yang, Zhilin and Qi, Peng and Zhang, Saizheng and Bengio, Yoshua and Cohen, William and Salakhutdinov, Ruslan and Manning, Christopher D. , editor =. Proceedings of the 2018. 2018 , pages =. doi:10.18653/v1/D18-1259 , abstract =

    2018 doi

  32. [40]

    2025 , note =

    allenai/scidocs , url =. 2025 , note =

    2025

  33. [41]

    Zhang, Yue and Li, Yafu and Cui, Leyang and Cai, Deng and Liu, Lemao and Fu, Tingchen and Huang, Xinting and Zhao, Enbo and Zhang, Yu and Xu, Chen and Chen, Yulong and Wang, Longyue and Luu, Anh Tuan and Bi, Wei and Shi, Freda and Shi, Shuming , month = sep, year =. Siren's. d...

    doi arXiv

  34. [42]

    Zhang, Yanzhao and Li, Mingxin and Long, Dingkun and Zhang, Xin and Lin, Huan and Yang, Baosong and Xie, Pengjun and Yang, An and Liu, Dayiheng and Lin, Junyang and Huang, Fei and Zhou, Jingren , month = jun, year =. Qwen3. doi:10.48550/arXiv.2506.05176 , abstract =

    doi arXiv

  35. [43]

    2024 , file =

    nomic-ai/nomic-embed-text-v1 ·. 2024 , file =

    2024

  36. [44]

    Arize AI , file =

    Understanding and. Arize AI , file =

  37. [45]

    Analytics Vidhya , author =

    8. Analytics Vidhya , author =. 2025 , file =

    2025

  38. [46]

    Zhong, Zijie and Liu, Hanwen and Cui, Xiaoya and Zhang, Xiaofan and Qin, Zengchang , month = jun, year =. Mix-of-. doi:10.48550/arXiv.2406.00456 , abstract =

    doi

  39. [47]

    IEEE Transactions on Knowledge and Data Engineering , author =

    Neural. IEEE Transactions on Knowledge and Data Engineering , author =. 2022 , keywords =. doi:10.1109/TKDE.2020.2983360 , abstract =

    2022 doi

  40. [48]

    and Lin, Kevin and Hewitt, John and Paranjape, Ashwin and Bevilacqua, Michele and Petroni, Fabio and Liang, Percy , month = nov, year =

    Liu, Nelson F. and Lin, Kevin and Hewitt, John and Paranjape, Ashwin and Bevilacqua, Michele and Petroni, Fabio and Liang, Percy , month = nov, year =. Lost in the. doi:10.48550/arXiv.2307.03172 , abstract =

    doi arXiv

  41. [49]

    https://openaccess.nhh.no/nhh-xmlui/bitstream/handle/11250/3178510/no.nhh\

  42. [50]

    doi:10.48550/arXiv.2203.05794 , abstract =

    Grootendorst, Maarten , month = mar, year =. doi:10.48550/arXiv.2203.05794 , abstract =

    doi arXiv

  43. [51]

    https://simg.baai.ac.cn/paperfile/25a43194-c74c-4cd3-b60f-0a1f27f8b8af.pdf , url =

  44. [52]

    Retrieval-

    Lewis, Patrick and Perez, Ethan and Piktus, Aleksandra and Petroni, Fabio and Karpukhin, Vladimir and Goyal, Naman and Küttler, Heinrich and Lewis, Mike and Yih, Wen-tau and Rocktäschel, Tim and Riedel, Sebastian and Kiela, Douwe , month = apr, year =. Retrieval-. doi:10.48550...

    2005 doi arXiv

  45. [53]

    https://arxiv.org/pdf/2005.11401 , url =

    2005 arXiv

  46. [54]

    Retrieval-

    Gao, Yunfan and Xiong, Yun and Gao, Xinyu and Jia, Kangxiang and Pan, Jinliu and Bi, Yuxi and Dai, Yi and Sun, Jiawei and Wang, Meng and Wang, Haofen , month = mar, year =. Retrieval-. doi:10.48550/arXiv.2312.10997 , abstract =

    doi arXiv

  47. [55]

    Bavarian Journal of Applied Sciences , author =

    Topical. Bavarian Journal of Applied Sciences , author =. 2023 , note =. doi:10.25929/1rjp-d197 , abstract =

    2023 doi

  48. [56]

    doi:10.48550/arXiv.2210.07316 , abstract =

    Muennighoff, Niklas and Tazi, Nouamane and Magne, Loïc and Reimers, Nils , month = mar, year =. doi:10.48550/arXiv.2210.07316 , abstract =

    doi arXiv

  49. [57]

    Retrieval-based

    Asai, Akari and Min, Sewon and Zhong, Zexuan and Chen, Danqi , editor =. Retrieval-based. Proceedings of the 61st. 2023 , pages =. doi:10.18653/v1/2023.acl-tutorials.6 , abstract =

    2023 doi

  50. [58]

    Proceedings of the 2019

    Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina , editor =. Proceedings of the 2019. 2019 , pages =. doi:10.18653/v1/N19-1423 , abstract =

    2019 doi

  51. [59]

    Contextual

    Bari, Harsh , month = jan, year =. Contextual. 2025. doi:10.1109/SCEECS64059.2025.10940513 , abstract =

    2025 doi

  52. [60]

    Putra, Syopiansyah Jaya and Gunawan, Muhamad Nur and Hidayat, Arief Akbar , month = sep, year =. Feature. 2022 10th. doi:10.1109/CITSM56380.2022.9935873 , abstract =

    2022 doi

  53. [61]

    Robust and

    Rahman, Rifat , month = dec, year =. Robust and. 2020 23rd. doi:10.1109/ICCIT51783.2020.9392738 , abstract =

    2020 doi

  54. [62]

    Wang, Liang and Yang, Nan and Huang, Xiaolong and Jiao, Binxing and Yang, Linjun and Jiang, Daxin and Majumder, Rangan and Wei, Furu , month = feb, year =. Text. doi:10.48550/arXiv.2212.03533 , abstract =

    doi arXiv

  55. [63]

    Sentence-

    Reimers, Nils and Gurevych, Iryna , month = aug, year =. Sentence-. doi:10.48550/arXiv.1908.10084 , abstract =

    1908 doi arXiv

  56. [64]

    Semantic

    Reisinger, Johannes and Fischer, Andreas and Igl, Andreas , month = aug, year =. Semantic. 2025 2nd. doi:10.1109/GACLM67198.2025.11231968 , abstract =

    2025 doi

  57. [65]

    Advances in Neural Information Processing Systems , author =

    Judging. Advances in Neural Information Processing Systems , author =. 2023 , pages =

    2023

  58. [66]

    2024 , keywords =

    sentence-transformers/all-. 2024 , keywords =

    2024

  59. [67]

    Y. Gao, Y. Xiong, X. Gao, K. Jia, J. Pan, Y. Bi, Y. Dai, J. Sun, M. Wang, and H. Wang, `` en Retrieval- Augmented Generation for Large Language Models : A Survey ,'' Mar. 2024, arXiv:2312.10997 [cs]. [Online]. Available: http://arxiv.org/abs/2312.10997

    2024 arXiv

  60. [68]

    R. Qu, R. Tu, and F. S. Bao, ``Is Semantic Chunking Worth the Computational Cost ?'' in Findings of the Association for Computational Linguistics : NAACL 2025 , L. Chiruzzo, A. Ritter, and L. Wang, Eds. 1em plus 0.5em minus 0.4em Albuquerque, New Mexico: Association for Comput...

    2025

  61. [69]

    Günther, I

    M. Günther, I. Mohr, D. J. Williams, B. Wang, and H. Xiao, ``Late Chunking : Contextual Chunk Embeddings Using Long - Context Embedding Models ,'' Jul. 2025, arXiv:2409.04701 [cs]. [Online]. Available: http://arxiv.org/abs/2409.04701

    2025

  62. [70]

    Reisinger, A

    J. Reisinger, A. Fischer, and A. Igl, ``Semantic Document Graphs for Knowledge Retrieval ,'' in 2025 2nd International Generative AI and Computational Language Modelling Conference ( GACLM ) , Aug. 2025, pp. 294--298. [Online]. Available: https://ieeexplore.ieee.org/document/11231968

    2025

  63. [71]

    ``sentence-transformers/all- MiniLM - L6 -v2 · Hugging Face ,'' Jan. 2024. [Online]. Available: https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2

    2024

  64. [72]

    S. Es, J. James, L. Espinosa Anke, and S. Schockaert, `` RAGAs : Automated Evaluation of Retrieval Augmented Generation ,'' in Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics : System Demonstrations , N. Aletras and O...

    2024

  65. [73]

    C. D. Manning, P. Raghavan, and H. Schütze, `` en Introduction to Information Retrieval ,'' Jul. 2008, iSBN: 9780511809071. [Online]. Available: https://www.cambridge.org/highereducation/books/introduction-to-information-retrieval/669D108D20F556C5C30957D63B5AB65C

    2008

Pith tools

Reviewed July 3, 2026 · model on record in the stance chip above.