Pith. sign in

REVIEW 4 major objections 6 minor 3 cited by

Multi-LLM Text Summarization

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Multiple LLMs generating and then selecting summaries outperforms single-LLM summarization by up to 3x on ROUGE/BLEU metrics.

desk verdict A useful empirical recipe undercut by a missing control: the multi-LLM pipeline adds a two-stage aggregation pass the single-LLM baseline does not have, so the headline gains are not cleanly attributable to multiple models. read the letter →

arxiv 2412.15487 v2 pith:YEPZBNMT submitted 2024-12-20 cs.CL

Jiangnan Fang , Cheng-Tse Liu , Jieun Kim , Yash Bhedaru , Ethan Liu , Nikhil Singh , Nedim Lipka , Puneet Mathur
show 4 more authors
Nesreen K. Ahmed Franck Dernoncourt Ryan A. Rossi Hanieh Deilamsalehy
This is my paper · ORCID
classification cs.CL
keywords textsummarizationmulti-LLMcentralizeddecentralizedlongdocumentLLMevaluationROUGEBLEU
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 tries to establish that document summarization improves when several different LLMs each draft a summary and then one central LLM (centralized) or all of them (decentralized) select the best candidate, rather than having a single LLM summarize the text alone. On the ArXiv and GovReport datasets, the multi-LLM pipelines score higher on ROUGE-1, ROUGE-L, BLEU-1, and BLEU-4 than all four single-LLM baselines, with individual improvements up to three times the baseline values. The authors further claim that the summaries their framework selects align with human preferences for coherence, conciseness, and fluency. If true, this is evidence that diversity in generation plus collaborative selection can substitute for a single more powerful model or elaborate prompt engineering.

What carries the argument

The mechanism is a two-stage chunk-and-resummarize pipeline plus a selection round. First the long document is chunked and each chunk is summarized; then the chunk summaries are concatenated, re-chunked, and summarized again. At each stage, k different LLMs independently generate candidate summaries of the same input using a shared prompt, producing diversity; an evaluation phase then either has one central LLM pick the best candidate (with a confidence score that can trigger another round) or has every LLM vote for the best candidate, requiring a majority. The selection step is the load-bearing part: it converts model diversity into a single chosen summary. The number of participating models, the choice of evaluator model, and the maximum number of rounds are the knobs the experiments vary.

What would settle it

Run the strongest single LLM through the same two-stage chunk-then-resummarize pipeline used by the multi-LLM framework, and compare its ROUGE and BLEU scores with the multi-LLM results in Table 2; if the single LLM with hierarchical aggregation matches or exceeds the multi-LLM scores, the claimed multi-LLM advantage is not attributable to collaboration.

Watch

Extended reading notes

Core claim

The central claim is that a generate-then-evaluate loop over k LLMs produces better summaries of long documents than any single LLM used in one pass. In the centralized variant, one central LLM reads all candidate summaries and picks one; in the decentralized variant, every participating LLM votes for the best candidate and the majority winner is chosen, with a tie-breaker model as fallback. Across both variants, with two LLMs as generators and one LLM as evaluator and tie-breaker, the framework outperforms the single-LLM baselines on every reported metric, with average improvements around 70 percent and individual gains up to three times. The one-round version matches or beats the multi-round version, so the authors conclude that even the cheapest configuration delivers the gains. A small human study of ten summary pairs finds that the model-picked winner is often the human-preferred winner, with conciseness agreeing at a kappa of 0.6.

Load-bearing premise

The comparison assumes the single-LLM baselines are the right control; those baselines summarize each chunk once and concatenate the results, while the multi-LLM pipeline adds a second chunk-and-resummarize stage, so some of the gain could come from that extra aggregation pass rather than from having multiple models.

Editorial extensions

If this is right

  • Even the cheapest configuration, two LLMs and a single round of generation and evaluation, delivers the reported gains, so the method does not require many models or many rounds.
  • Conversational rounds beyond the first do not improve scores, which implies that the framework's value lies in selection rather than iterative refinement.
  • Changing the evaluator and tie-breaker model, or adding a third LLM, keeps the multi-LLM scores above the single-LLM baselines, so the effect is not tied to one particular evaluator.
  • Specialized prompts aimed at coherence or precision can feed the same framework and still outperform the single-LLM baselines, indicating the approach is not bound to a single prompt formula.
  • Short-text summarization also benefits, with the centralized one-round configuration showing the largest short-text gains on ArXiv introductions.

Reading between the lines

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

  • Beyond the paper: a fair test would give a single LLM the same two-stage chunk-summarize-resummarize pipeline used by the multi-LLM framework; until such a baseline is run, part of the gain could come from the extra aggregation pass rather than from multi-LLM collaboration.
  • Beyond the paper: with exactly two LLMs, the decentralized majority rule is equivalent to unanimity, so the reported consensus behavior is really a single agreeing vote plus the tie-breaker, not a broad consensus.
  • Beyond the paper: one testable extension is whether the same generate-and-select loop improves a single LLM's output when the candidates are produced by different prompts or temperatures, which would isolate the selection step as the active ingredient.
  • Beyond the paper: the human-alignment claim rests on ten summary pairs, so agreement estimates such as kappa equal to 0.6 are suggestive rather than definitive.
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

4 major / 6 minor

Summary. The paper proposes a multi-LLM summarization framework in which k LLMs generate candidate summaries and either a single central LLM (centralized) or all k LLMs (decentralized) evaluate and iteratively refine them. Long documents are handled with a hierarchical two-stage procedure: the document is chunked, each chunk is summarized, and the concatenated intermediate summaries undergo a second round of chunking and summarization. Experiments on ArXiv and GovReport report ROUGE and BLEU gains over single-LLM baselines by up to 3x, together with ablations over model combinations, number of LLMs, prompting variants, and text length. A small human evaluation is used to claim alignment between the framework's preferences and human judgments.

Significance. If the attribution were established, this would be a useful empirical contribution: the paper includes detailed algorithm listings and prompts, provides a token-level cost analysis (Sections 4.3, 5.3, B.1, B.2), and explores several systematic ablations. The central comparison, however, is confounded by an architectural difference between the multi-LLM and single-LLM conditions, and the human-alignment claim rests on weak agreement in a small sample. The abstract's principal claim therefore needs substantial additional evidence before it is supported.

major comments (4)
  1. [§3 vs. §6.1] The single-LLM baselines are not adequate controls for the central claim. Section 6.1 states that baselines use the same chunking and that "the final summary represents a concatenation of the generated summaries," whereas Section 3 defines the multi-LLM framework as a "two stage process" that adds "a second round of chunking and summarization on the concatenated intermediate results." Every multi-LLM condition inherits this second aggregation stage, so the large gains in Tables 2, 3, 5, 6, 7, and 8 are also consistent with the possibility that a second hierarchical pass over the concatenated chunk summaries—rather than the participation of multiple LLMs—causes the improvement. The ablations in Section 6.3 cannot isolate model multiplicity because they retain the same two-stage structure. A required control is a single-LLM pipeline using the identical two-stage chunk-then-resummarize procedure, with the same prompts at each stage and a comparable number of inference calls; without it, the attribution in the abstract is unsupported.
  2. [§6.5, Tables 9–10] The contribution that "the best performing method in the framework aligns with human judgments" is not supported by the presented evidence. The human raters evaluate only the first 10 pairs of summaries generated before the final evaluation step of one pipeline (decentralized, one-round maximum), not the final outputs of the best-performing method. The reported agreement between human and machine choices is κ = 0.2, 0.6, 0.1, and 0.2 across the four comparisons, and the machine choices in Table 10 are split 5/5 between GPT-3.5 and GPT-4o mini, making the statement that humans "generally prefer" GPT-4o mini and that this aligns with the framework difficult to sustain. This claim should be removed, substantially weakened, or re-tested on final summaries from the full framework with a larger sample and a pre-specified analysis.
  3. [Appendix C.1, Tables 2, 5] The main results select GPT-3.5 as the evaluator and tie-breaker after comparing GPT-4o mini, GPT-3.5, and GPT-4o on the same test sets and observing that GPT-3.5 "emerged as the best-scoring evaluator" (Appendix C.1). This is a post-hoc selection of the evaluation configuration on the test data, which inflates the headline scores. The evaluator and tie-breaking model should be chosen on a held-out validation split, or the paper should report all evaluator configurations without designating one as best based on test-set performance.
  4. [§6.2, Tables 2, 3, 5, 6, 8] The abstract and Section 6.2 use "significantly outperform" without any significance test, confidence interval, or per-document paired comparison. All tables report single point estimates, and some "up to 3x" improvements correspond to small absolute differences (e.g., ArXiv BLEU-4 in Table 2: 0.021 vs. 0.043). The authors should report paired tests across documents (for example, bootstrap or signed-rank tests) and effect sizes for the central multi-LLM versus single-LLM comparisons.
minor comments (6)
  1. [§6.2] There is a typo in the second paragraph: "under decentralized and decentralized frameworks" should presumably read "under decentralized and centralized frameworks."
  2. [§6.1, Figures 2 and 6] Section 6.1 says the baseline generation prompt is identical to that in the first round and cites Figure 6, but Figure 6 is the prompt used for the second-stage concatenated chunks; the figure reference should be aligned with the described stage to avoid confusion.
  3. [Algorithms 1 and 2] The algorithms use S both for an ordered set of summaries and for the input text, and line 1 calls CREATE SUMMARY (S) without defining this function; the notation should be disambiguated.
  4. [§5.1.2] The displayed formula for the final summary S* uses the notation "r ∈ 1, . . . , kk" and would be much easier to parse with explicit vote-vector indexing; the current typesetting makes the tie-break rule hard to read.
  5. [§6.3, Table 3] The claim that Table 3 shows improvements "across the board" cannot be checked directly because Table 3 reports only multi-LLM scores; adding baseline rows or deltas would make the comparison explicit.
  6. [Appendix D] The human raters are disclosed to be authors and close associates; even with anonymized and randomized summaries, this is a potential source of bias and should be discussed as a limitation or addressed with independent raters.

Circularity Check

1 steps flagged · score 2.0 of 10

The main ROUGE/BLEU comparison is independent and externally benchmarked; the only exhibited circularity is the human-evaluation tie-break, which defines some 'human' choices using the system's own default.

  1. self definitional [Appendix D, Table 10 note (human evaluation tie-break)]
    "When two summaries in a particular criterion have the same average score, we fallback on the default choice GPT-3.5, consistent with the evaluation step in our multi-LLM framework."

    Human-preference labels are not fully rater-derived: whenever averaged scores tie, the paper assigns the label GPT-3.5 because that is also the multi-LLM framework's tie-breaker. The reported Cohen's kappa between human and machine choices therefore counts tie-resolved agreements that are forced by sharing the same default. For example, in Table 9 Summary 1 is tied at 3.57 for Coherence, so the 'human' choice becomes GPT-3.5 by rule, and the framework also chooses GPT-3.5; this agreement contributes to the reported kappa by construction rather than as evidence of alignment. The main ROUGE/BLEU results are unaffected.

full rationale

The central claim—multi-LLM summarization outperforms single-LLM baselines on ROUGE/BLEU on ArXiv and GovReport—is an empirical pipeline comparison against external reference summaries, with no fitted equation and no self-authored uniqueness theorem that forces the result. The only exhibited circular step is the human-evaluation tie-break described above; it is local to the secondary 'aligns with human judgments' claim and does not transfer to the ROUGE/BLEU tables. Three other worries are real but are not circularity under the rules: the single-LLM baselines in Section 6.1 are one-pass concatenations while every multi-LLM condition in Section 3 includes a second chunk-and-summarize stage, so the attribution of gains to 'multiple LLMs' is confounded; the GPT-3.5 evaluator and tie-breaker was selected after inspecting the same test sets in Appendix C.1, giving selection bias; and the human raters are authors and close associates in Appendix D, limiting independence. These affect interpretation and generalizability of the empirical claims, but none is a definitional reduction of a reported score to its input.

Assumptions & free parameters 7 free parameters · 3 assumptions · 0 invented entities

The framework's headline results depend on hand-chosen pipeline constants (chunk size, word limit, max rounds, k) and on a post-hoc selected evaluator and tie-breaker model. No new physical or formal entities are introduced; the centralized/decentralized topologies are organizational labels rather than invented entities.

free parameters (7)
  • chunk_size = 4K characters
    Global chunk size for both baselines and multi-LLM pipelines; chosen by hand, not swept in the main results.
  • summary_word_limit W = 160 words
    Prompt target for summary length in all models; fixed by hand.
  • evaluator model = GPT-3.5
    Central evaluator in Table 2; selected because Appendix C.1 showed it scored best on the same test sets, which is a post-hoc selection.
  • tie-breaker model = GPT-3.5
    Decentralized default when no absolute majority; same post-hoc selection as the evaluator.
  • confidence threshold = not reported
    The centralized stopping rule uses a threshold on the LLM's 0-10 confidence score, but the threshold value is never stated in the paper.
  • max rounds tmax = 1 or 3
    Two values are compared; 3 rounds never beat 1 round, so the multi-round mechanism is not shown to add benefit.
  • number of models k = 2 (and 3 in ablation)
    The main results use 2 LLMs; increasing to 3 does not improve scores, so k is a hand-chosen capacity parameter.
assumptions (3)
  • domain assumption An LLM evaluator can reliably select the best summary among candidates, either as a single central judge or through majority vote.
    The selection mechanism in Sections 4 and 5 assumes judges choose a higher-quality summary; no validation shows judges outperform random or that voting helps.
  • domain assumption The two-stage chunk-then-resummarize pipeline is comparable across conditions, or does not itself produce the observed gains.
    Section 6.1 baselines concatenate first-pass summaries while Sections 3-5 use a second re-summarization stage; the attribution of gains to multi-LLM depends on this unstated comparability.
  • domain assumption ROUGE-1, ROUGE-L, BLEU-1, and BLEU-4 are adequate proxies for summary quality on ArXiv and GovReport.
    Standard NLP assumption, but long-document abstractive summaries have known weak correlation with these lexical metrics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multi-LLM Text Summarization." pith.science (2026). https://pith.science/paper/YEPZBNMT

@misc{pith2026241215487,
  author       = {Pith},
  title        = {Pith review of: Multi-LLM Text Summarization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YEPZBNMT}},
  note         = {Machine review of arXiv:2412.15487}
}
read the original abstract

In this work, we propose a Multi-LLM summarization framework, and investigate two different multi-LLM strategies including centralized and decentralized. Our multi-LLM summarization framework has two fundamentally important steps at each round of conversation: generation and evaluation. These steps are different depending on whether our multi-LLM decentralized summarization is used or centralized. In both our multi-LLM decentralized and centralized strategies, we have k different LLMs that generate diverse summaries of the text. However, during evaluation, our multi-LLM centralized summarization approach leverages a single LLM to evaluate the summaries and select the best one whereas k LLMs are used for decentralized multi-LLM summarization. Overall, we find that our multi-LLM summarization approaches significantly outperform the baselines that leverage only a single LLM by up to 3x. These results indicate the effectiveness of multi-LLM approaches for summarization.

Figures

Figures reproduced from arXiv: 2412.15487 by the authors.

Figure 1
Figure 1. Centralized and Decentralized approaches us [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Prompt for generating the initial summary in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Generation prompt that is used after the initial [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (7 more)
Figure 6
Figure 6. Figure 6: Generation prompt for generating the final [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 4
Figure 4. Figure 4: Evaluation prompt for evaluating the sum [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Evaluation prompt for evaluating the sum [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 7
Figure 7. Figure 7: Prompt 1 for generating the initial summary [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: Prompt 2 for generating the initial summary [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Here we showcase an example of how we choose at which point an introduction ends. The total word [PITH_FULL_IMAGE:figures/full_fig_p018_9.png]
Figure 10
Figure 10. Figure 10: Screenshot of the interface (with scores already filled in) and instructions given to raters for human [PITH_FULL_IMAGE:figures/full_fig_p019_10.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

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

  1. Network-informed Prompt Engineering against Organized Astroturf Campaigns under Extreme Class Imbalance

    cs.CL 2025-01 conditional novelty 5.0 of 10

    Frozen LLMs with balanced retrieval-augmented prompting detect astroturf campaigns better than GNN baselines on a 2016 US election dataset, but the reported margins are overstated.

  2. PolicyStory: Leveraging Large Language Models to Generate Comprehensible Summaries of Policy-News in India

    cs.CY 2025-09 conditional novelty 4.0 of 10

    PolicyStory uses Llama-3.2-1B to produce topic-wise, chronological, three-level summaries of Indian policy news, and a 22-person user study reports positive usability feedback.

  3. Toward Edge General Intelligence with Multiple-Large Language Model (Multi-LLM): Architecture, Trust, and Orchestration

    cs.NI 2025-07 conditional novelty 4.0 of 10

    A survey of multi-LLM systems in edge computing, covering architectures, enabling technologies, trust mechanisms, applications, and open datasets for edge general intelligence.

Reference graph

Works this paper leans on

46 extracted references · 15 canonical work pages · cited by 3 Pith papers

  1. [1]

    URL: " 'urlintro :=

    ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprinturl eprintpr...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Griffin Adams, Alexander Fabbri, Faisal Ladhak, Eric Lehman, and Noémie Elhadad. 2023. http://arxiv.org/abs/2309.04269 From sparse to dense: Gpt-4 summarization with chain of density prompting

    arXiv 2023

  4. [4]

    Joshua Ainslie, Santiago Ontanon, Chris Alberti, Vaclav Cvicek, Zachary Fisher, Philip Pham, Anirudh Ravula, Sumit Sanghai, Qifan Wang, and Li Yang. 2020. http://arxiv.org/abs/2004.08483 Etc: Encoding long and structured inputs in transformers

    arXiv 2020

  5. [5]

    Lochan Basyal and Mihir Sanghvi. 2023. http://arxiv.org/abs/2310.10449 Text summarization using large language models: A comparative study of mpt-7b-instruct, falcon-7b-instruct, and openai chat-gpt models

    arXiv 2023

  6. [6]

    Peters, and Arman Cohan

    Iz Beltagy, Matthew E. Peters, and Arman Cohan. 2020. http://arxiv.org/abs/2004.05150 Longformer: The long-document transformer

    arXiv 2020

  7. [7]

    Yapei Chang, Kyle Lo, Tanya Goyal, and Mohit Iyyer. 2024. http://arxiv.org/abs/2310.00785 Booookscore: A systematic exploration of book-length summarization in the era of llms

    arXiv 2024

  8. [8]

    Justin Chih-Yao Chen, Swarnadeep Saha, and Mohit Bansal. 2024. http://arxiv.org/abs/2309.13007 Reconcile: Round-table conference improves reasoning via consensus among diverse llms

    arXiv 2024

Show all 46 references
  1. [9]

    Aakanksha Chowdhery, Sharan Narang, Jacob Devlin, Maarten Bosma, Gaurav Mishra, Adam Roberts, Paul Barham, Hyung Won Chung, Charles Sutton, Sebastian Gehrmann, Parker Schuh, Kensen Shi, Sasha Tsvyashchenko, Joshua Maynez, Abhishek Rao, Parker Barnes, Yi Tay, Noam Shazeer, Vino...

    2022 arXiv

  2. [10]

    Janara Christensen, Stephen Soderland, Gagan Bansal, and Mausam. 2014. Hierarchical summarization: Scaling up multi-document summarization. In Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 902--912. Assoc...

    2014

  3. [11]

    Arman Cohan, Franck Dernoncourt, Doo Soon Kim, Trung Bui, Seokhwan Kim, Walter Chang, and Nazli Goharian. 2018. http://arxiv.org/abs/1804.05685 A discourse-aware attention model for abstractive summarization of long documents

    2018 arXiv

  4. [12]

    Conroy and Hoa Trang Dang

    John M. Conroy and Hoa Trang Dang. 2008. https://aclanthology.org/C08-1019/ Mind the gap: Dangers of divorcing evaluations of summary content from linguistic quality . In Proceedings of the 22nd International Conference on Computational Linguistics (Coling 2008), pages 145--15...

    2008

  5. [13]

    Zihang Dai, Zhilin Yang, Yiming Yang, Jaime Carbonell, Quoc Le, and Ruslan Salakhutdinov. 2019. https://doi.org/10.18653/v1/P19-1285 Transformer-xl: Attentive language models beyond a fixed-length context . In Proceedings of the 57th Annual Meeting of the Association for Compu...

    2019 doi

  6. [14]

    Tenenbaum, and Igor Mordatch

    Yilun Du, Shuang Li, Antonio Torralba, Joshua B. Tenenbaum, and Igor Mordatch. 2023. http://arxiv.org/abs/2305.14325 Improving factuality and reasoning in language models through multiagent debate

    2023 arXiv

  7. [15]

    Alexios Gidiotis and Grigorios Tsoumakas. 2020. A divide-and-conquer approach to the summarization of long documents. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 28:3029--3040

    2020

  8. [16]

    Yihong Gong and Xin Liu. 2001. Generic text summarization using relevance measure and latent semantic analysis. In Proceedings of the 24th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 19--25. ACM

    2001

  9. [17]

    Tanya Goyal, Junyi Jessy Li, and Greg Durrett. 2023. http://arxiv.org/abs/2209.12356 News summarization and evaluation in the era of gpt-3

    2023 arXiv

  10. [18]

    Mandy Guo, Joshua Ainslie, David Uthus, Santiago Ontanon, Jianmo Ni, Yun-Hsuan Sung, and Yinfei Yang. 2021. http://arxiv.org/abs/2112.07916 Longt5: Efficient text-to-text transformer for long sequences

    2021 arXiv

  11. [19]

    Karl Moritz Hermann, Tomáš Kočiský, Edward Grefenstette, Lasse Espeholt, Will Kay, Mustafa Suleyman, and Phil Blunsom. 2015. http://arxiv.org/abs/1506.03340 Teaching machines to read and comprehend

    2015 arXiv

  12. [20]

    Luyang Huang, Shuyang Cao, Nikolaus Parulian, Heng Ji, and Lu Wang. 2021. https://doi.org/10.18653/v1/2021.naacl-main.112 Efficient attentions for long document summarization . In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computati...

    2021 doi

  13. [21]

    Emma J \"a rvinen. 2024. Long-input summarization using large language models

    2024

  14. [22]

    Gunjan Keswani, Wani Bisen, Hirkani Padwad, Yash Wankhedkar, Sudhanshu Pandey, and Ayushi Soni. 2024. Abstractive long text summarization using large language models. International Journal of Intelligent Systems and Applications in Engineering, 12(12s):160--168

    2024

  15. [23]

    Anastassia Kornilova and Vladimir Eidelman. 2019. https://doi.org/10.18653/v1/D19-5406 B ill S um: A corpus for automatic summarization of US legislation . In Proceedings of the 2nd Workshop on New Frontiers in Summarization, pages 48--56, Hong Kong, China. Association for Com...

    2019 doi

  16. [24]

    Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Veselin Stoyanov, and Luke Zettlemoyer. 2020. https://doi.org/10.18653/v1/2020.acl-main.703 Bart: Denoising sequence-to-sequence pre-training for natural language generation, translation...

    2020 doi

  17. [25]

    Irene Li, Aosong Feng, Dragomir Radev, and Rex Ying. 2023. https://doi.org/10.18653/v1/2023.acl-short.16 Hipool: Modeling long documents using graph neural networks . In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Pa...

    2023 doi

  18. [26]

    Yunxuan Li, Yibing Du, Jiageng Zhang, Le Hou, Peter Grabowski, Yeqing Li, and Eugene Ie. 2024. http://arxiv.org/abs/2406.11776 Improving multi-agent debate with sparse communication topology

    2024 arXiv

  19. [27]

    Tian Liang, Zhiwei He, Wenxiang Jiao, Xing Wang, Rui Wang, Yujiu Yang, Zhaopeng Tu, and Shuming Shi. 2024. http://arxiv.org/abs/2305.19118 Encouraging divergent thinking in large language models through multi-agent debate

    2024 arXiv

  20. [28]

    Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang

    Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. 2023. http://arxiv.org/abs/2307.03172 Lost in the middle: How language models use long contexts

    2023 arXiv

  21. [29]

    Yang Liu and Mirella Lapata. 2019. Text summarization with pretrained encoders. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 3730--3740

    2019

  22. [30]

    Yixin Liu, Pengfei Liu, Dragomir Radev, and Graham Neubig. 2022. http://arxiv.org/abs/2203.16804 Brio: Bringing order to abstractive summarization

    2022 arXiv

  23. [31]

    Mallick, A

    S. Mallick, A. Ghosh, et al. 2019. A survey on extractive text summarization. Journal of Artificial Intelligence Research, 65:123--143

    2019

  24. [32]

    Rada Mihalcea and Paul Tarau. 2004. Textrank: Bringing order into texts. In Proceedings of the 2004 Conference on Empirical Methods in Natural Language Processing, pages 404--411

    2004

  25. [33]

    Ramesh Nallapati, Bowen Zhou, Cicero Nogueira dos santos, Caglar Gulcehre, and Bing Xiang. 2016. http://arxiv.org/abs/1602.06023 Abstractive text summarization using sequence-to-sequence rnns and beyond

    2016 arXiv

  26. [34]

    Cohen, and Mirella Lapata

    Shashi Narayan, Shay B. Cohen, and Mirella Lapata. 2018. http://arxiv.org/abs/1808.08745 Don't give me the details, just the summary! topic-aware convolutional neural networks for extreme summarization

    2018 arXiv

  27. [35]

    Bo Pang, Erik Nijkamp, Wojciech Kryściński, Silvio Savarese, Yingbo Zhou, and Caiming Xiong. 2022. http://arxiv.org/abs/2203.07586 Long document summarization with top-down and bottom-up inference

    2022 arXiv

  28. [36]

    Dongqi Pu, Yifan Wang, and Vera Demberg. 2023 a . http://arxiv.org/abs/2305.16784 Incorporating distributions of discourse structure for long document abstractive summarization

    2023 arXiv

  29. [37]

    Xiao Pu, Mingqi Gao, and Xiaojun Wan. 2023 b . http://arxiv.org/abs/2309.09558 Summarization is (almost) dead

    2023 arXiv

  30. [38]

    Rush, Sumit Chopra, and Jason Weston

    Alexander M. Rush, Sumit Chopra, and Jason Weston. 2015. A neural attention model for abstractive sentence summarization. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 379--389

    2015

  31. [39]

    Victor Sanh, Albert Webson, Colin Raffel, Stephen H. Bach, Lintang Sutawika, Zaid Alyafeai, Antoine Chaffin, Arnaud Stiegler, Teven Le Scao, Arun Raja, Manan Dey, M Saiful Bari, Canwen Xu, Urmish Thakker, Shanya Sharma Sharma, Eliza Szczechla, Taewoon Kim, Gunjan Chhablani, Ni...

    2022 arXiv

  32. [40]

    Liu, and Christopher D

    Abigail See, Peter J. Liu, and Christopher D. Manning. 2017. Get to the point: Summarization with pointer-generator networks. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1073--1083

    2017

  33. [41]

    Sam Shleifer. 2020. Distilbart-cnn-12-6. https://huggingface.co/sshleifer/distilbart-cnn-12-6. Accessed: 2024-05-29

    2020

  34. [42]

    Sainbayar Sukhbaatar, Edouard Grave, Piotr Bojanowski, and Armand Joulin. 2019. https://doi.org/10.18653/v1/P19-1032 Adaptive attention span in transformers . In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 331--335, Florence, ...

    2019 doi

  35. [43]

    Wen Xiao and Giuseppe Carenini. 2019. https://doi.org/10.18653/v1/D19-1298 Extractive summarization of long documents by combining global and local context . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joi...

    2019 doi

  36. [44]

    Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, and Amr Ahmed. 2021. http://arxiv.org/abs/2007.14062 Big bird: Transformers for longer sequences

    2021 arXiv

  37. [45]

    Jingqing Zhang, Yao Zhao, Mohammad Saleh, and Peter J. Liu. 2020. http://arxiv.org/abs/1912.08777 Pegasus: Pre-training with extracted gap-sentences for abstractive summarization

    2020 arXiv

  38. [46]

    Hashimoto

    Tianyi Zhang, Faisal Ladhak, Esin Durmus, Percy Liang, Kathleen McKeown, and Tatsunori B. Hashimoto. 2023. http://arxiv.org/abs/2301.13848 Benchmarking large language models for news summarization

    2023 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.