Elamur Research⌕
☰
home / papers / 2025 / StrucSum: Graph-Structured Reasoning for Long Document Extractive Summarization with LLMs

StrucSum: Graph-Structured Reasoning for Long Document Extractive Summarization with LLMs

Paper · 2025 · Haohan Yuan, Sukhwa Hong and Haopeng Zhang

Summary

Large language models can choose summary sentences without training, but on long texts they often miss how the document is organized and which content is central. StrucSum targets this problem in zero-shot extractive summarization. It converts each document into a Text-Attributed Graph. Sentences are nodes embedded with Sentence-BERT, and two sentences are linked when their cosine similarity exceeds a threshold θ. Signals from this graph are then written into the prompt in one of three ways:

  • Neighbor-Aware Prompting (NAP) lists each sentence's graph neighbors.
  • Centrality-Aware Prompting (CAP) adds a normalized degree centrality score.
  • Centrality-Guided Masking (CGM) removes weakly connected sentences to shorten the input.

No model weights are changed. The authors tested the method on ArXiv, PubMed and Multi-News with models such as GPT-4o and LLaMA3-70B. It outperformed plain prompting and unsupervised baselines on both reference-overlap and factuality metrics. The study shows that simple, interpretable graph cues can steer black-box LLMs on long documents. It also gives practical advice on which strategy suits which goal.

Key points

  • On ArXiv, FactCC rose by 19.2 points and SummaC by 8.0 points relative to the baselines.
  • Each strategy has its own strength. CGM gives the best overlap with reference summaries, NAP preserves factual consistency best, and CAP balances quality against input length. Combining strategies brought no clear additional gain.
  • Prompts containing centrality (CAP or combined setups) produced significant positive Spearman correlations (p < 0.05) between sentence centrality and selection rank. NAP and CGM alone did not.
  • In a graph-only test with GPT4o-mini, sentence text was removed from the prompt. Textual neighbor lists did better than numeric or binary adjacency matrices, but still fell short of NAP with full text (ArXiv ROUGE-2: 11.87 vs 14.59). Structure therefore cannot replace content.
  • Limitations: evaluation used 200 test documents per dataset and English only, with no comparison to supervised models. Code is available at github.com/HaohanYuan01/StrucSum.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓University of Hawaiʻi at MānoaHonolulu · United States4 works · 20 citations🎓University of Hawaii at HiloHilo · United States2 works · 9 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 2Research subjects and languages the work is about.Reasoning · 1,426 worksSummarization · 232 works

Cite

Cite as Haohan Yuan, Sukhwa Hong and Haopeng Zhang. 2025. StrucSum: Graph-Structured Reasoning for Long Document Extractive Summarization with LLMs. arXiv:2505.22950.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2505.22950 · arXiv 2505.22950

@misc{yuan2025strucsum,
  title = {{StrucSum: Graph-Structured Reasoning for Long Document Extractive Summarization with LLMs}},
  author = {Haohan Yuan, Sukhwa Hong and Haopeng Zhang},
  year = {2025},
  doi = {10.48550/arxiv.2505.22950},
  eprint = {2505.22950},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2505.22950}
}

References

39 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901.
  2. Xiuying Chen, Mingzhe Li, Shen Gao, Rui Yan, Xin Gao, and Xiangliang Zhang. 2022. Scientific paper extractive summarization enhanced by citation graphs. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 4053–4062, Abu Dhabi, United Arab Emirates. Associ… find
  3. Anshuman Chhabra, Hadi Askari, and Prasant Mohapatra. 2024. Revisiting zero-shot abstractive summarization in the era of large language models from the perspective of position bias. Preprint, arXiv:2401.01989. find
  4. Arman Cohan, Franck Dernoncourt, Doo Soon Kim, Trung Bui, Seokhwan Kim, Walter Chang, and Nazli Goharian. 2018. A discourse-aware attention model for abstractive summarization of long documents. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational…
  5. Peng Cui, Le Hu, and Yuanchao Liu. 2020. Enhancing extractive text summarization with topic-aware graph neural networks. In Proceedings of the 28th International Conference on Computational Linguistics, pages 5360–5371, Barcelona, Spain (Online). International Committee on Computational Linguistics. find
  6. Günes Erkan and Dragomir R. Radev. 2004. LexRank: Graph-based lexical centrality as salience in text summarization. Journal of Artificial Intelligence Research. find
  7. Alexander Fabbri, Irene Li, Tianwei She, Suyi Li, and Dragomir Radev. 2019. Multi-news: A large-scale multi-document summarization dataset and abstractive hierarchical model. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 1074–1084, Florence, Italy.…
  8. Tanya Goyal, Junyi Jessy Li, and Greg Durrett. 2023. News summarization and evaluation in the era of gpt-3. Preprint, arXiv:2209.12356. find
  9. Hanqi Jin, Tianming Wang, and Xiaojun Wan. 2020. Multi-granularity interaction network for extractive and abstractive multi-document summarization. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6244– 6254, Online. Association for Computational Ling… find
  10. Huan Yee Koh, Jiaxin Ju, Ming Liu, and Shirui Pan. 2022. An empirical survey on long document summarization: Datasets, models, and metrics. ACM Comput. Surv., 55(8). find
  11. Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. 2023. Large language models are zero-shot reasoners. Preprint, arXiv:2205.11916.
  12. Wojciech Kryściński, Bryan McCann, Caiming Xiong, and Richard Socher. 2019. Evaluating the factual consistency of abstractive text summarization. Preprint, arXiv:1910.12840.
  13. Philippe Laban, Tobias Schnabel, Paul N. Bennett, and Marti A. Hearst. 2022. SummaC: Re-visiting NLI-based models for inconsistency detection in summarization. Transactions of the Association for Computational Linguistics, 10:163–177. find
  14. Miao Li, Jianzhong Qi, and Jey Han Lau. 2023. Compressed heterogeneous graph for abstractive multi-document summarization. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 13085–13093. find
  15. Wei Li, Xinyan Xiao, Jiachen Liu, Hua Wu, Haifeng Wang, and Junping Du. 2020. Leveraging graph to improve abstractive multi-document summarization. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6232–6243, Online. Association for Computational Lingu… find
  16. Chin-Yew Lin. 2004. ROUGE: A package for automatic evaluation of summaries. In Text Summarization Branches Out, pages 74–81, Barcelona, Spain. Association for Computational Linguistics.
  17. Yang Liu, Dan Iter, Yichong Xu, Shuohang Wang, Ruochen Xu, and Chenguang Zhu. 2023. G-eval: Nlg evaluation using gpt-4 with better human alignment. Preprint, arXiv:2303.16634.
  18. Qianren Mao, Hongdong Zhu, Junnan Liu, Cheng Ji, Hao Peng, Jianxin Li, Lihong Wang, and Zheng Wang. 2022. Muchsum: Multi-channel graph neural network for extractive summarization. In Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGI… find
  19. Rada Mihalcea and Paul Tarau. 2004. Textrank: Bringing order into text. In Proceedings of the 2004 conference on empirical methods in natural language processing, pages 404–411. find
  20. Gianluca Moro, Luca Ragazzi, and Lorenzo Valgimigli. 2023. Graph-based abstractive summarization of extracted essential knowledge for low-resource scenarios. In ECAI 2023, pages 1747–1754. IOS Press. find
  21. Mark Newman. 2010. Networks: An Introduction. Oxford University Press. find
  22. Xiao Pu, Mingqi Gao, and Xiaojun Wan. 2023. Summarization is (almost) dead. Preprint, arXiv:2309.09558. — [arXiv:2309.09558](https://arxiv.org/abs/2309.09558) find
  23. Haohan Yuan and Haopeng Zhang. 2024. Domainsum: A hierarchical benchmark for fine-grained domain shift in abstractive text summarization. Preprint, arXiv:2410.15687. find
  24. Haopeng Zhang, Xiao Liu, and Jiawei Zhang. 2022. HEGEL: Hypergraph transformer for long document summarization. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 10167–10176, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics. find
  25. Haopeng Zhang, Xiao Liu, and Jiawei Zhang. 2023a. Extractive summarization via chatgpt for faithful summary generation. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 3270–3278. find
  26. Haopeng Zhang, Philip S. Yu, and Jiawei Zhang. 2024a. A systematic survey of text summarization: From statistical methods to large language models. Preprint, arXiv:2406.11289. find
  27. Kristin Qi, Youxiang Zhu, and Xiaohui Liang. 2025. Umb@peranssumm 2025: Enhancing perspective-aware summarization with prompt optimization and supervised fine-tuning. Preprint, arXiv:2503.11118. find
  28. Shiyue Zhang, David Wan, and Mohit Bansal. 2023b. Extractive is not faithful: An investigation of broad unfaithfulness problems in extractive summarization. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2153–2174, Toronto, C… find
  29. Qian Ruan, Malte Ostendorff, and Georg Rehm. 2022. Histruct+: Improving extractive text summarization with hierarchical structure information. Preprint, arXiv:2203.09629. find
  30. Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q. Weinberger, and Yoav Artzi. 2020. Bertscore: Evaluating text generation with bert. Preprint, arXiv:1904.09675. find
  31. Pranab Sahoo, Ayush Kumar Singh, Sriparna Saha, Vinija Jain, Samrat Mondal, and Aman Chadha. 2025. A systematic survey of prompt engineering in large language models: Techniques and applications. Preprint, arXiv:2402.07927.
  32. Tianyi Zhang, Faisal Ladhak, Esin Durmus, Percy Liang, Kathleen McKeown, and Tatsunori B. Hashimoto. 2024b. Benchmarking large language models for news summarization. Transactions of the Association for Computational Linguistics, 12:39–57.
  33. Neelabh Sinha. 2025. Qa-prompting: Improving summarization with large language models using question-answering. Preprint, arXiv:2505.14347. find
  34. Chenlong Zhao, Xiwen Zhou, Xiaopeng Xie, and Yong Zhang. 2024. Hierarchical attention graph for scientific document summarization in global and local level. In Findings of the Association for Computational Linguistics: NAACL 2024, pages 714–726. find
  35. Hwanjun Song, Hang Su, Igor Shalyminov, Jason Cai, and Saab Mansour. 2024. Finesure: Fine-grained summarization evaluation using llms. Preprint, arXiv:2407.00908.
  36. Danqing Wang, Pengfei Liu, Yining Zheng, Xipeng Qiu, and Xuanjing Huang. 2020. Heterogeneous graph neural networks for extractive document summarization. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6209–6219, Online. Association for Computational…
  37. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc Le, and Denny Zhou. 2023. Chain-of-thought prompting elicits reasoning in large language models. Preprint, arXiv:2201.11903.
  38. Hao Zheng and Mirella Lapata. 2019. Sentence centrality revisited for unsupervised summarization. arXiv preprint arXiv:1906.03508.
  39. Yang Zhong and Diane Litman. 2025. Discourse-driven evaluation: Unveiling factual inconsistency in long document summarization. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long P…

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yuan-2025-strucsum-graph-structured-reasoning-long/