GRASP: Graph-Reasoning Aided Survey Planning for High-Fidelity Related Work Generation
Writing a literature review requires a deep understanding of the relationships among cited papers: how they build on, challenge, or offer alternative perspectives to one another. We present Graph-Reasoning Aided Survey Planning (GRASP), a framework combining LLM planning for related work generation...
Збережено в:
| Автори: | , |
|---|---|
| Формат: | Стаття |
| Опубліковано: |
2026
|
| Предмети: | |
| Онлайн доступ: | https://arxiv.org/abs/2607.03709 http://arxiv.org/abs/2607.03709 |
| Теги: |
Немає тегів, Будьте першим, хто поставить тег для цього запису!
|
| _version_ | 1879179616128598016 |
|---|---|
| author | Li, Haoming Ouyang, Jessica |
| author_browse | Li, Haoming Ouyang, Jessica |
| author_facet | Li, Haoming Ouyang, Jessica |
| author_sort | Li, Haoming |
| building | arXiv |
| collection | Acceso abierto |
| description | Writing a literature review requires a deep understanding of the relationships among cited papers: how they build on, challenge, or offer alternative perspectives to one another. We present Graph-Reasoning Aided Survey Planning (GRASP), a framework combining LLM planning for related work generation with graph algorithms to extract key relationships among cited papers. Our two-layer graph structure consists of a Graph of Thoughts and an Argument-Counterargument Planning Network, representing the cited papers at different levels of granularity, and we apply topology-aware pruning via a Steiner tree to identify the core inter-paper relationships captured in our graph. Our citation analysis-based evaluation shows that GRASP generates related work sections (RWS) that closely match human-written targets in terms of the discourse roles, intents, and grouping of citations. |
| doi_str_mv | 10.18653/v1/2026.findings-acl.1815 |
| first_indexed | 2026-10-16T01:35:27Z |
| format | Article |
| fullrecord | <oai_dc:dc xmlns:oai_dc="http://www.openarchives.org/OAI/2.0/oai_dc/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.openarchives.org/OAI/2.0/oai_dc/ https://www.openarchives.org/OAI/2.0/oai_dc.xsd"><identifier>arxiv-2607.03709</identifier><datestamp>2026-07-07</datestamp>
<dc:title>GRASP: Graph-Reasoning Aided Survey Planning for High-Fidelity Related Work Generation</dc:title>
<dc:creator>Li, Haoming</dc:creator>
<dc:creator>Ouyang, Jessica</dc:creator>
<dc:subject>Computation and Language</dc:subject>
<dc:subject>I.2.7; I.2.6</dc:subject>
<dc:description>Writing a literature review requires a deep understanding of the relationships among cited papers: how they build on, challenge, or offer alternative perspectives to one another. We present Graph-Reasoning Aided Survey Planning (GRASP), a framework combining LLM planning for related work generation with graph algorithms to extract key relationships among cited papers. Our two-layer graph structure consists of a Graph of Thoughts and an Argument-Counterargument Planning Network, representing the cited papers at different levels of granularity, and we apply topology-aware pruning via a Steiner tree to identify the core inter-paper relationships captured in our graph. Our citation analysis-based evaluation shows that GRASP generates related work sections (RWS) that closely match human-written targets in terms of the discourse roles, intents, and grouping of citations. </dc:description>
<dc:description>23 pages, 3 figures. Published in Findings of the Association for Computational Linguistics: ACL 2026</dc:description>
<dc:date>2026-07-04</dc:date>
<dc:type>text</dc:type>
<dc:identifier>https://arxiv.org/abs/2607.03709</dc:identifier>
<dc:identifier>http://arxiv.org/abs/2607.03709</dc:identifier>
<dc:identifier>Findings of the Association for Computational Linguistics: ACL 2026, pages 36427-36449, San Diego, California, United States. Association for Computational Linguistics</dc:identifier>
<dc:identifier>doi:10.18653/v1/2026.findings-acl.1815</dc:identifier>
</oai_dc:dc> |
| id | arxiv-2607.03709 |
| institution | arXiv |
| last_indexed | 2026-10-16T01:35:27Z |
| publishDate | 2026 |
| publishDateRange | 2026 |
| publishDateSort | 2026 |
| record_format | oai_dc |
| spelling | arxiv-2607.037092026-07-07 GRASP: Graph-Reasoning Aided Survey Planning for High-Fidelity Related Work Generation Li, Haoming Ouyang, Jessica Computation and Language I.2.7; I.2.6 Writing a literature review requires a deep understanding of the relationships among cited papers: how they build on, challenge, or offer alternative perspectives to one another. We present Graph-Reasoning Aided Survey Planning (GRASP), a framework combining LLM planning for related work generation with graph algorithms to extract key relationships among cited papers. Our two-layer graph structure consists of a Graph of Thoughts and an Argument-Counterargument Planning Network, representing the cited papers at different levels of granularity, and we apply topology-aware pruning via a Steiner tree to identify the core inter-paper relationships captured in our graph. Our citation analysis-based evaluation shows that GRASP generates related work sections (RWS) that closely match human-written targets in terms of the discourse roles, intents, and grouping of citations. 23 pages, 3 figures. Published in Findings of the Association for Computational Linguistics: ACL 2026 2026-07-04 text https://arxiv.org/abs/2607.03709 http://arxiv.org/abs/2607.03709 Findings of the Association for Computational Linguistics: ACL 2026, pages 36427-36449, San Diego, California, United States. Association for Computational Linguistics doi:10.18653/v1/2026.findings-acl.1815 |
| spellingShingle | Computation and Language I.2.7; I.2.6 Li, Haoming Ouyang, Jessica GRASP: Graph-Reasoning Aided Survey Planning for High-Fidelity Related Work Generation |
| title | GRASP: Graph-Reasoning Aided Survey Planning for High-Fidelity Related Work Generation |
| title_full | GRASP: Graph-Reasoning Aided Survey Planning for High-Fidelity Related Work Generation |
| title_fullStr | GRASP: Graph-Reasoning Aided Survey Planning for High-Fidelity Related Work Generation |
| title_full_unstemmed | GRASP: Graph-Reasoning Aided Survey Planning for High-Fidelity Related Work Generation |
| title_short | GRASP: Graph-Reasoning Aided Survey Planning for High-Fidelity Related Work Generation |
| title_sort | grasp graph reasoning aided survey planning for high fidelity related work generation |
| topic | Computation and Language I.2.7; I.2.6 |
| url | https://arxiv.org/abs/2607.03709 http://arxiv.org/abs/2607.03709 |
| work_keys_str_mv | AT lihaoming graspgraphreasoningaidedsurveyplanningforhighfidelityrelatedworkgeneration AT ouyangjessica graspgraphreasoningaidedsurveyplanningforhighfidelityrelatedworkgeneration |