Pith. sign in

REVIEW 4 major objections 7 minor 41 references

Leveraging Generative Models for Real-Time Query-Driven Text Summarization in Large-Scale Web Search

T0 review · 4 major / 7 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A 0.1-billion-parameter generative model can replace a multi-stage extractive summarization pipeline in real-time web search.

desk verdict A deployable distil-and-align recipe for low-latency query summarization with real online gains, but the SOTA claim is a statistical tie and the extractive guarantee is asserted, not enforced. read the letter →

arxiv 2508.20559 v1 pith:6O27VL4H submitted 2025-08-28 cs.CL cs.IR

Zeyu Xiong , Yixuan Nan , Li Gao , Hengzhu Tang , Shuaiqiang Wang , Junfeng Wang , Dawei Yin This is my paper · ORCID
classification cs.CLcs.IR
keywords generativesummarizationquery-driventextmodeldistillationdirectpreferenceoptimizationlookaheaddecodingFP8quantizationlow-latencyinferencewebsearch
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 claims that a 0.1-billion-parameter generative model, trained by distilling a 10-billion-parameter teacher and then refined with human annotations and online preference signals, can replace the multi-stage extractive summary pipeline of a large-scale web-search engine. On a production test set of 1,021 query-document pairs, the small model reports the highest ROUGE-2 score (51.33), edges out a 10-billion-parameter nearline generator on that metric, and is judged by human annotators to be substantially better than the deployed extractive baseline ($\Delta$GSB +20.68%). It also meets real-time serving constraints: with FP8 quantization and lookahead decoding, about 334 GPUs handle roughly 50,000 queries per second at an average latency of 55 ms. If true, this gives search engines a cheaper, lower-latency path to query-relevant summaries without the cumulative information loss of funnel-style extractive pipelines.

What carries the argument

The load-bearing object is QDGenSumRT, a four-stage training-and-serving pipeline wrapped around the 0.1B-parameter student model, a 12-layer, 768-hidden-dimension decoder-only transformer. The machinery that carries the argument is the combination of (1) distillation from a 10B teacher, which transfers the ability to select query-relevant sentences; (2) full-parameter supervised fine-tuning on 6,160 curated human summaries, which imposes the roughly 80-token, three-to-five-point structured format and filters out ambiguous supervision; (3) direct preference optimization on click-derived preferences, which nudges output structure toward what users actually choose; and (4) FP8 quantization plus lookahead speculative decoding, which removes the latency that ordinarily makes 80-token generative output too slow for live search. The prompt is deliberately kept minimal to match the student's weaker instruction-following, and the output constraint $|S|\leq 80$ tokens is what ties the generative model to the display budget of a search result page.

What would settle it

Conduct a blind head-to-head evaluation on the 1,021 query-document test pairs in which summaries are produced by both systems but judged by a completely independent annotation team that has never seen the training data conventions; if the +20.68% $\Delta$GSB advantage over the production baseline does not survive independent judging, or the ROUGE-2 lead over the 10B nearline generator reverses, the central claim of state-of-the-art quality is not supportable.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that a task-specialized 0.1B decoder-only transformer, trained in four stages, reaches or exceeds the summary quality of a 10B generative model while running in real time. Stage one distills the teacher into a lightweight student on a synthetic corpus of 14 million query-title-content-summary samples (about 130B tokens). Stage two fine-tunes the student on 6,160 human-annotated examples that enforce a consistent three-to-five-point, roughly 80-token output format. Stage three applies direct preference optimization using summaries ranked by real user click behavior, interleaved with periodic SFT refreshes. Stage four deploys the model with FP8 weight-activation quantization and a lookahead speculative-decoding strategy. In offline tests the model scores 59.21 ROUGE-1, 51.33 ROUGE-2, and 57.12 ROUGE-L; human judges prefer it to the production extractive baseline by +20.68% $\Delta$GSB (p = 1.3e-6) and find it statistically indistinguishable from the 10B nearline generator ($\Delta$GSB -0.12%, p = 0.88); and a live A/B test shows gains of +0.81% CTR, +0.33% search engagement, +0.62% satisfactory consumption, and +0.54% distribution ratio.

Load-bearing premise

The load-bearing premise is that the human-annotated summaries and preference judgments used for both training and evaluation measure query relevance and summary quality without systematic bias; if those annotations share stylistic or content biases with the trained model, the reported gains over the production baseline could shrink or vanish in genuinely independent evaluation.

Editorial extensions

If this is right

  • A 0.1B generative model can match a 10B teacher on a narrow, well-specified summarization task, so task-specific distillation plus preference alignment can substitute for raw parameter count.
  • The funnel-style extractive pipeline can be replaced by one end-to-end model, removing the cumulative information loss and the weakest-stage bottleneck.
  • FP8 quantization and lookahead decoding make generative 80-token output viable under industrial latency budgets: about 334 GPUs sustain roughly 50,000 queries per second at 55 ms average latency.
  • Online engagement metrics (CTR, search engagement, satisfactory consumption, distribution ratio) improve with the generative summaries at p < 0.05, implying users find the new summaries more useful, not just higher-scoring on offline metrics.
  • Because the summaries stay extractive, users can still click through to find the exact source sentence, so the generative approach does not compromise verifyability against the landing page.

Reading between the lines

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

  • This suggests the same distillation-SFT-DPO recipe could transfer to other constrained-output generation tasks in search, such as query-biased title rewriting, answer snippets, or structured product highlights, where extractive pipelines currently dominate.
  • A testable extension would be to measure whether the click-derived preference signal alone, without the SFT anchor, is stable over time; the paper interleaves DPO with periodic SFT refreshes, implying that pure preference drift may degrade format consistency.
  • One could also probe generalization: since the student is distilled on a 10B teacher and 14M queries, holding out an entire query category would reveal whether the apparent state of the art comes from task specialization or from memorizing the teacher's preferences.
  • The reported $\Delta$GSB near-parity with the 10B nearline generator suggests that, at equal quality, serving cost becomes the deciding factor; the paper's own numbers imply the 0.1B model is the cheaper deployment choice by a large margin.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The manuscript presents QDGenSumRT, a four-stage framework for query-driven text summarization in Baidu Search: distillation from a 10B-parameter ERNIE-Lite-8K teacher into a 0.1B-parameter GPT-2-style student, full-parameter SFT on 6,160 human-annotated examples, DPO on click-based preference data, and deployment with FP8 quantization, TensorRT-LLM, and lookahead decoding. Offline evaluation reports ROUGE on a 1,021-pair held-out set, human evaluation using the ΔGSB metric against two systems, online A/B metrics, efficiency measurements, and ablations. The central claim is that this 0.1B model outperforms the production extractive baseline, matches a 10B nearline model, and achieves a new state of the art at much lower latency.

Significance. The production-scale evidence is the main strength. The +20.68% ΔGSB gain over Base (p=1.3e-6), the uniformly positive online A/B metrics, and the 149.7 QPS with 55 ms average inference time on an L20 GPU make a credible case that a small generative model can replace a multi-stage extractive pipeline in a real-time search setting. The ablation isolating distillation, SFT, and DPO is clear and informative. However, the paper's 'new state of the art' claim is not supported by its own Table 2, and the formal extractive definition is not enforced by any component of the method. These issues are fixable by reframing the claims and adding a faithfulness or extraction-rate evaluation.

major comments (4)
  1. [Section 3.1, Sections 3.2–3.4, Eqs. (1)–(5)] The paper defines QDTS as selecting a subset S of D and motivates in Section 1 that summaries should be extractive so that users are not shown 'entirely rephrased or hallucinated content.' However, the student is an autoregressive GPT-2-style model trained with the MLE objective in Eq. (3) and the DPO objective in Eq. (5); no mechanism such as constrained decoding, sentence copying, or a selection loss enforces S ⊆ D, and no faithfulness or hallucination metric is reported. The ROUGE and ΔGSB metrics cannot distinguish extractive sentence selection from fluent abstractive generation, so the production-safety property asserted in Section 1 is unverified. Please either add a constraint or enforcement mechanism and report an extraction-rate or faithfulness metric, or explicitly reframe the system as abstractive with an accompanying hallucination and factuality evaluation.
  2. [Abstract and Section 4.5, Table 2] The claim that the model 'achieves a new state of the art' is contradicted by the paper's own comparison against NGS: ΔGSB is -0.12% with p=0.88, and ROUGE-1 and ROUGE-L are lower than NGS (59.21 vs. 60.48 and 57.12 vs. 58.31, respectively). Only ROUGE-2 is numerically higher (51.33 vs. 50.25). The claim should be weakened to 'competitive with the 10B nearline model' unless a statistically significant advantage over NGS is demonstrated.
  3. [Sections 4.1–4.2] Both the SFT training set (6,160 triplets) and the test set (1,021 triplets) are produced by Baidu's in-house annotation platform under the same stated criteria, which creates a distributional-circularity risk for the reported gains. The paper does not report inter-annotator agreement, and it does not specify how 'factual consistency, relevance, and centrality' were operationalized. Please add agreement statistics and, if possible, an independently annotated evaluation set that was not produced under the same conventions as the SFT data.
  4. [Sections 4.4 and 4.5, Table 3] The production baseline Base is the anchor for both the headline ΔGSB gain and the online A/B results, but its design is concealed with the statement 'Specific details are omitted due to confidentiality constraints.' The online metrics are reported only as relative percentages with a blanket 'p-values for these experimental metrics are less than 0.05'; exact p-values and confidence intervals are missing, and with four correlated metrics the multiple-comparison issue is not addressed. Please provide a high-level description of Base, exact p-values and confidence intervals for Table 3, and state how many comparisons were tested.
minor comments (7)
  1. [Section 4.2 heading] The heading 'Evaluation Metircs' contains a typo; it should be 'Evaluation Metrics.'
  2. [Section 4.2, Eq. (7)] The ROUGE-L formula is not the standard F-measure definition and does not define L(X, Y) or state how recall is combined with precision; please use the standard ROUGE-L F-score and define all symbols.
  3. [Section 3.2] The statement that 14 million samples yield 130B tokens implies roughly 9,300 tokens per sample; please clarify whether this count includes the source document content and whether near-duplicate or low-quality teacher generations were filtered before training.
  4. [Section 3.4] Click-based preference collection at the same display position is a reasonable weak signal, but the paper should specify how display order was randomized and how position bias or novelty effects were controlled when constructing D_dpo.
  5. [Table 4] The column heading 'Train Time (per epoch)' is inconsistent with the text, which compares total training time ('reduces training time by more than 35x'); please clarify whether the reported values are per-epoch or total and define what 'E2E Latency' includes.
  6. [Abstract and Section 4.5] The abstract's figure of 334 NVIDIA L20 GPUs for ~50,000 queries per second is never derived in the body; Table 5 reports 149.7 QPS per GPU, which implies about 334 GPUs, but the calculation should be stated explicitly in the experiments.
  7. [Section 4.5] The phrase 'significantly outperforms most existing methods' overstates the comparison, since DeepExtract and BART are offline, out-of-domain baselines and the only statistically significant comparison in the paper is against Base.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the empirical distillation/SFT/DPO results are independently evaluated and do not reduce by construction to their inputs.

full rationale

The paper's load-bearing claims are empirical measurements, not formal derivations in which an output is defined in terms of an input. QDGenSumRT is built by distilling ERNIE-Lite-8K outputs into a 0.1B GPT-2-style student, then applying SFT on 6,160 human-annotated triples, DPO on click-derived preferences, FP8 quantization, and lookahead decoding (Sections 3.2-3.5). The reported results (ROUGE-2 51.33 in Table 1; ΔGSB +20.68% vs. Base in Table 2; online CTR/SE/SCR/DR lifts in Table 3) are evaluated on a held-out set of 1,021 test triples and live A/B traffic; no fitted parameter is renamed as a prediction. The annotation platform is shared between SFT and test data, which is a potential validity threat but not a circularity: no equation or selection rule makes the test labels a deterministic function of the training examples. DPO optimizes click-derived preferences while the online A/B test also measures CTR, but the paper reports the measured lift rather than deriving it from the DPO loss by construction, so this is a feedback-loop concern rather than a circular step. The extractive-definition mismatch flagged by the reviewer (S ⊆ D in Section 3.1 is not enforced by the autoregressive objective) is a correctness/faithfulness concern, not a circular step. The only same-group citation found, reference [17], supports a general remark about deployment difficulty and is not load-bearing. Accordingly no circular step can be exhibited, and the score is 0.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical or mathematical entities; its framework is a trained system. The central claims rest on domain assumptions about data quality, annotation validity, and user-behavior signals. The only independent check outside the training pipeline is the online A/B test, but its setup is summarized only briefly and the exact configuration is not public.

free parameters (5)
  • Lookahead decoding hyperparameters (w, n, v) = 4, 6, 4
    Selected from Table 5 to maximize QPS and minimize inference time while keeping ROUGE-2 near 51.3; the paper's deployment throughput claim (149.7 QPS per L20 GPU) depends on these values.
  • Summary length constraint k = 80 tokens
    Defined in Section 3.1 as the display constraint and used in training and evaluation; a product choice, not derived from data.
  • Distillation dataset size N = 14 million samples, 130B tokens
    Chosen to transfer teacher capability at scale; no analysis of diminishing returns or saturation is provided.
  • SFT training epochs and batch size = 6 epochs, batch 128
    Reported in Section 4.3 for all SFT results; no sensitivity analysis is given.
  • DPO coefficient beta
    Equation (5) includes beta, a key preference-alignment strength, but neither its value nor its sensitivity is reported, despite DPO contributing +5.46 ROUGE-2 in Table 6.
assumptions (4)
  • domain assumption Human-labeled summaries are valid ground truth for query-document relevance and summary quality.
    Section 4.1 uses annotator-produced summaries as reference labels and as the basis for GSB evaluation. If these are biased or inconsistent, the offline results are not meaningful.
  • domain assumption User click behavior in live A/B experiments is a valid implicit preference signal between two alternative summaries.
    Section 3.4 builds the DPO preference dataset from click signals; Section 4.5 uses online engagement metrics as evidence of improvement. Click behavior can be confounded by position, formatting, or salience.
  • domain assumption Teacher outputs from ERNIE-Lite-8K are suitable supervision targets for the QDTS task.
    Section 3.2 uses 14 million teacher-generated summaries as distillation training labels without independent verification of their correctness or coverage.
  • domain assumption The private test set and online traffic are representative of the production query distribution.
    Section 4.1 describes a real-world Baidu collection, but the sampling criteria for the 1,021 test triplets are not detailed, and the query distribution is not characterized.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Leveraging Generative Models for Real-Time Query-Driven Text Summarization in Large-Scale Web Search." pith.science (2026). https://pith.science/paper/6O27VL4H

@misc{pith2026250820559,
  author       = {Pith},
  title        = {Pith review of: Leveraging Generative Models for Real-Time Query-Driven Text Summarization in Large-Scale Web Search},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6O27VL4H}},
  note         = {Machine review of arXiv:2508.20559}
}
read the original abstract

In the dynamic landscape of large-scale web search, Query-Driven Text Summarization (QDTS) aims to generate concise and informative summaries from textual documents based on a given query, which is essential for improving user engagement and facilitating rapid decision-making. Traditional extractive summarization models, based primarily on ranking candidate summary segments, have been the dominant approach in industrial applications. However, these approaches suffer from two key limitations: 1) The multi-stage pipeline often introduces cumulative information loss and architectural bottlenecks due to its weakest component; 2) Traditional models lack sufficient semantic understanding of both user queries and documents, particularly when dealing with complex search intents. In this study, we propose a novel framework to pioneer the application of generative models to address real-time QDTS in industrial web search. Our approach integrates large model distillation, supervised fine-tuning, direct preference optimization, and lookahead decoding to transform a lightweight model with only 0.1B parameters into a domain-specialized QDTS expert. Evaluated on multiple industry-relevant metrics, our model outperforms the production baseline and achieves a new state of the art. Furthermore, it demonstrates excellent deployment efficiency, requiring only 334 NVIDIA L20 GPUs to handle \textasciitilde50,000 queries per second under 55~ms average latency per query.

Figures

Figures reproduced from arXiv: 2508.20559 by the authors.

Figure 1
Figure 1. Differences between existing methods and ours. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Pipeline overview of QDGenSumRT. We first take user queries, along with web page titles and contents, as input, and [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 30 canonical work pages

  1. [1]

    Mahsa Afsharizadeh, Hossein Ebrahimpour-Komleh, and Ayoub Bagheri. 2018. Query-oriented text summarization using sentence extraction technique. In 2018 4th international conference on web research (ICWR) . IEEE, 128–132

    work page 2018

  2. [2]

    Hannah Bast and Marjan Celikik. 2014. Efficient index-based snippet generation. ACM Transactions on Information Systems (TOIS) 32, 2 (2014), 1–24

    work page 2014

  3. [3]

    Ronald Brandow, Karl Mitze, and Lisa F Rau. 1995. Automatic condensation of electronic publications by sentence selection. Information Processing & Manage- ment 31, 5 (1995), 675–685

    work page 1995

  4. [4]

    Gayathri Venu Madhuri Chandu, Amritha Premkumar, Nalini Sampath, et al. 2019. Extractive approach for query based text summarization. In 2019 International Conference on Issues and Challenges in Intelligent Computing Techniques (ICICT) , Vol. 1. IEEE, 1–5

    work page 2019

  5. [5]

    Hoa Trang Dang. 2006. DUC 2005: Evaluation of question-focused summariza- tion systems. In Proceedings of the workshop on task-focused summarization and question answering. 48–55

    work page 2006

  6. [6]

    Yang Deng, Wai Lam, Yuexiang Xie, Daoyuan Chen, Yaliang Li, Min Yang, and Ying Shen. 2020. Joint learning of answer selection and answer summary genera- tion in community question answering. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 34. 7651–7658

    work page 2020

  7. [7]

    Günes Erkan and Dragomir R Radev. 2004. Lexrank: Graph-based lexical centrality as salience in text summarization. Journal of artificial intelligence research 22 (2004), 457–479

    work page 2004

  8. [8]

    Yichao Fu, Peter Bailis, Ion Stoica, and Hao Zhang. 2024. Break the sequen- tial dependency of llm inference using lookahead decoding. arXiv preprint arXiv:2402.02057 (2024)

    arXiv 2024

Show all 41 references
  1. [9]

    Nikolaos Giarelis, Charalampos Mastrokostas, and Nikos Karacapilidis. 2023. Abstractive vs. extractive summarization: An experimental review. Applied Sciences 13, 13 (2023), 7620

    2023

  2. [10]

    Rupali Goyal, Parteek Kumar, and VP Singh. 2025. QFAS-KE: Query focused answer summarization using keyword extraction. Information Processing & Management 62, 4 (2025), 104104

    2025

  3. [11]

    Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, et al. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783 (2024)

    2024 arXiv

  4. [12]

    Shaoyao Huang, Luozheng Qin, and Ziqiang Cao. 2023. Diffusion language model with query-document relevance for query-focused summarization. In Findings of the Association for Computational Linguistics: EMNLP 2023 . 11020–11030

    2023

  5. [13]

    Akanksha Joshi, Eduardo Fidalgo, Enrique Alegre, and Laura Fernández-Robles

  6. [14]

    Khosrow Kaikhah. 2004. Automatic text summarization with neural networks. In 2004 2nd international IEEE conference on’Intelligent Systems’. Proceedings (IEEE cat. No. 04EX791), Vol. 1. IEEE, 40–44

    2004

  7. [15]

    Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. 2020. Scaling laws for neural language models. arXiv preprint arXiv:2001.08361 (2020)

    2020 arXiv

  8. [16]

    Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Veselin Stoyanov, and Luke Zettlemoyer. 2020. BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension. In Proceedings of the 58t...

    2020

  9. [17]

    Xiaoyu Li, Xiao Li, Li Gao, Yiding Liu, Xiaoyang Wang, Shuaiqiang Wang, Junfeng Wang, and Dawei Yin. 2025. Proactive Guidance of Multi-Turn Conversation in Industrial Search. arXiv preprint arXiv:2505.24251 (2025)

    2025 arXiv

  10. [18]

    Chin-Yew Lin. 2004. Rouge: A package for automatic evaluation of summaries. In Text summarization branches out. 74–81

    2004

  11. [19]

    Hans Peter Luhn. 1958. The automatic creation of literature abstracts. IBM Journal of research and development 2, 2 (1958), 159–165

    1958

  12. [20]

    Wenjuan Luo, Fuzhen Zhuang, Qing He, and Zhongzhi Shi. 2013. Exploiting rele- vance, coverage, and novelty for query-focused multi-document summarization. Knowledge-Based Systems 46 (2013), 33–42

    2013

  13. [21]

    Paulius Micikevicius, Dusan Stosic, Neil Burgess, Marius Cornea, Pradeep Dubey, Richard Grisenthwaite, Sangwon Ha, Alexander Heinecke, Patrick Judd, John Kamalu, et al. 2022. Fp8 formats for deep learning. arXiv preprint arXiv:2209.05433 (2022)

    2022 arXiv

  14. [22]

    Rada Mihalcea and Paul Tarau. 2004. Textrank: Bringing order into text. In Pro- ceedings of the 2004 conference on empirical methods in natural language processing. 404–411

    2004

  15. [23]

    Muhidin Mohamed, Mourad Oussalah, and Victor Chang. 2024. SDbQfSum: Query-focused summarization framework based on diversity and text semantic analysis. Expert Systems 41, 1 (2024), e13462

    2024

  16. [24]

    Muhidin A Mohamed and Mourad Oussalah. 2015. Similarity-based query- focused multi-document summarization using crowdsourced and manually-built lexical-semantic resources. In 2015 IEEE Trustcom/BigDataSE/ISPA, Vol. 2. IEEE, 80–87

    2015

  17. [25]

    Preksha Nema, Mitesh Khapra, Anirban Laha, and Balaraman Ravindran. 2017. Diversity driven attention model for query-based abstractive summarization. arXiv preprint arXiv:1704.08300 (2017)

    2017 arXiv

  18. [26]

    NVIDIA Corporation. 2025. TensorRT-LLM: Large Language Model Inference Library by NVIDIA. https://github.com/NVIDIA/TensorRT-LLM. Accessed: 2025-04-05

    2025

  19. [27]

    Aytuğ Onan and Hesham A Alhumyani. 2024. DeepExtract: Semantic-driven extractive text summarization framework using LLMs and hierarchical positional encoding. Journal of King Saud University-Computer and Information Sciences 36, 8 (2024), 102178

    2024

  20. [28]

    Choongwon Park and Youngjoong Ko. 2022. Qsg transformer: Transformer with query-attentive semantic graph for query-focused summarization. In Proceedings of the 45th international ACM SIGIR conference on research and development in information retrieval. 2589–2594

    2022

  21. [29]

    Min Peng, Binlong Gao, Jiahui Zhu, Jiajia Huang, Mengting Yuan, and Fei Li

  22. [30]

    Muhammad Reza Qorib, Qisheng Hu, and Hwee Tou Ng. 2025. Just What You Desire: Constrained Timeline Summarization with Self-Reflection for Enhanced Relevance. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 39. 25065–25073

    2025

  23. [31]

    Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. 2019. Language models are unsupervised multitask learners. OpenAI blog 1, 8 (2019), 9

    2019

  24. [32]

    Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2023. Direct preference optimization: Your language model is secretly a reward model. Advances in Neural Information Processing Systems 36 (2023), 53728–53741

    2023

  25. [33]

    Nazreena Rahman and Bhogeswar Borah. 2015. A survey on existing extractive techniques for query-based text summarization. In 2015 international symposium on advanced computing and communication (ISACC) . IEEE, 98–102

    2015

  26. [34]

    Sajad Sotudeh and Nazli Goharian. 2023. Qontsum: On contrasting salient content for query-focused summarization. arXiv preprint arXiv:2307.07586 (2023)

    2023 arXiv

  27. [35]

    Yu Sun, Shuohuan Wang, Shikun Feng, Siyu Ding, Chao Pang, Junyuan Shang, Ji- axiang Liu, Xuyi Chen, Yanbin Zhao, Yuxiang Lu, et al. 2021. Ernie 3.0: Large-scale knowledge enhanced pre-training for language understanding and generation. arXiv preprint arXiv:2107.02137 (2021)

    2021 arXiv

  28. [36]

    Andrew Turpin, Yohannes Tsegay, David Hawking, and Hugh E Williams. 2007. Fast generation of result snippets in web search. InProceedings of the 30th annual international ACM SIGIR conference on Research and development in information retrieval. 127–134

    2007

  29. [37]

    Zhichao Xu and Daniel Cohen. 2023. A lightweight constrained generation alter- native for query-focused summarization. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval . 1745–1749

    2023

  30. [38]

    Jin-ge Yao, Xiaojun Wan, and Jianguo Xiao. 2017. Recent advances in document summarization. Knowledge and Information Systems 53 (2017), 297–336

    2017

  31. [39]

    Jiaqi Zhang, Ling Lu, Liang Zhang, Yinong Chen, and Wanping Liu. 2024. DCD- Sum: An interpretable extractive summarization framework based on contrastive learning method. Engineering Applications of Artificial Intelligence 133 (2024), 108148

    2024

  32. [2016]

    Expert Systems with applications 44 (2016), 92–101

    High quality information extraction and query-oriented summarization for automatic query-reply in social network. Expert Systems with applications 44 (2016), 92–101

    2016

  33. [2023]

    Expert Systems with Applications 211 (2023), 118442

    DeepSumm: Exploiting topic models and sequence to sequence networks for extractive text summarization. Expert Systems with Applications 211 (2023), 118442

    2023

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.