[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85900-en":3,"doc-seo-85900-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},85900,687197207639,"Asher","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","A Stepwise Questioning Expert-Editor Multi-Agent Framework for Long-Document Summarization","Long-document summarization with large language models is hindered by input-length limits and by hallucinations that reduce faithfulness and coverage. This paper introduces an expert–editor stepwise questioning multi-agent framework in which an expert and an editor steer another agent through progressively posed questions targeting semantic aspects and surface-level errors. The approach refines summaries iteratively and improves consistency with the source text. Experiments on two scientific long-document datasets validate effectiveness using widely adopted automatic metrics.","1  \nA Stepwise Questioning Expert-Editor Multi-Agent Framework for Long-Document Summarization  \nLingyun Shen1[0000-0001-9054-6493] and Xuejia Guo1[0009-0006-6915-3963]  \n1 China South Industry Academy BJ 102209, China  \n[sheryshen20@gmail.com](sheryshen20@gmail.com)  \nAbstract. Although large language models (LLMs) have shown promising potential in news summarization tasks, their performance on long-document summarization remains challenging as their length often exceeds the input limits. As the agent investment, which provide possibility to improve the inherent capabilities of LLMs. To enhance the effectiveness of long-document summarization based on LLMs, this paper proposes an expert–editor stepwise questioning multi-agent method, in which the expert and the editor guide another agent to refine the summary by posing questions on different aspects of the content and providing targeted clues for revision. We conducted experiments on two representative long-document scientific datasets and evaluated the results through widely recognized automatic metrics. The results demonstrated the effectiveness of our method.  \nKeywords: Long-Document Summary, Multi-Agent, Agent Collaboration,  \nPrompt Engineering.  \n1 Introduction  \nLarge language models (LLMs) have achieved remarkable success across nature language understanding and generation task.(Sun et al., 2021; Chowdhery et al., 2023; Brown et al., 2020) Leveraging LLMs for summarization with Instruction has shown significant potential(Zhang et al., 2023; Liu et al., 2025; Li et al., 2025a) . However, owing to the context processing capability and hallucination issues, LLMs ability on long document summarization remains need further improved(Wan et al., ; Chan et al., 2023; Liu, 2024). Recently, methods that leverage multi-agent systems to enhance the intrinsic abilities of large language models have shown promising potential in various long-form generation and reasoning tasks(Gu, 2025; Wan et al., ; Hu, 2025; Fang et al., ; Mo and Hu, 2024). This suggests that multi-agent collaboration and planning may address the limitations in long-document summarization based on LLMs.  \nCurrently, various agent framework are work on leading LLMs particpatly subtask by prompt technologies(Mo and Hu, 2024; Fang et al., ; Wan et al., ; Hu, 2025). Prior prompt technologies studies provides lots of experience that utilize large language models to construct agents, such as role play(Schmidt et al., 2023; Wang et al., 2024c; Zheng et al., 2024)， decompose a complex problem into a series of simple  \n2  \nquestions(Press, 2023), iterative refined(Jha et al., 2023; Bakharia, 2025),etc. Based on empirical observation, we consider that stepwise questioning may stimulate LLMs torevie their previously generated output. As shown in figure1, the question “What are the key terminologies in the original article?” triggered the large language model to reflect on and revise its prior summary.  \nInspired by these prior studies and practical experiences, we propose a stepwise questioning multi-agent approach for long-document summarization. By capturing key information more accurately to grasp the core idea of the article (Shen and Le, 2023), this method can address the issues of incomplete information and unfaithful in long document summarization base on LLMs. Specifically, We design two primary roles, expert who foucuses on semantic understanding and the important clue of the main idea about article for generating consistent, fluently and faithful summary, and editor who focuses on the surface-level literal errors such as grammatical and syntactic accuracy, as well as consistency with the original text. They both guide the author agent to reflect on and revise its initial summary written previously through progressively asking questions. Both guide another Author agent through step-by-step questioning to reflect on and revise the initially written summary. This idea is inspired by the academic review process,where ","cbCaigjw2wi6o8X6","https://ap.wps.com/l/cbCaigjw2wi6o8X6","pdf",455704,4,1,12,"English","en",105,"# Introduction\n# Related works\n## Long Document Summarization with Large Language Models\n## Multi-agent systems for long-form generation and reasoning","[{\"question\":\"What problem does the proposed framework address in long-document summarization?\",\"answer\":\"It addresses the difficulty of summarizing documents whose length exceeds model input limits, along with reliability issues such as incomplete information and unfaithful summaries.\"},{\"question\":\"How do the expert and editor roles contribute to summary refinement?\",\"answer\":\"The expert focuses on semantic understanding and key clues for the main idea, while the editor targets surface-level literal errors and checks consistency with the original text.\"},{\"question\":\"What method is used to guide the summarization refinement process?\",\"answer\":\"The framework uses stepwise questioning: it progressively asks targeted questions so the author agent can reflect on and revise its previously generated summary.\"}]",1784207038,30,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"a-stepwise-questioning-expert-editor-multi-agent-framework-for-long-document-summarization","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":20},"https://docshare.wps.com/document/a-stepwise-questioning-expert-editor-multi-agent-framework-for-long-document-summarization/85900/",{"url":52,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-26","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does the proposed framework address in long-document summarization?","Question",{"text":75,"@type":76},"It addresses the difficulty of summarizing documents whose length exceeds model input limits, along with reliability issues such as incomplete information and unfaithful summaries.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How do the expert and editor roles contribute to summary refinement?",{"text":80,"@type":76},"The expert focuses on semantic understanding and key clues for the main idea, while the editor targets surface-level literal errors and checks consistency with the original text.",{"name":82,"@type":73,"acceptedAnswer":83},"What method is used to guide the summarization refinement process?",{"text":84,"@type":76},"The framework uses stepwise questioning: it progressively asks targeted questions so the author agent can reflect on and revise its previously generated summary.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,122,127,130,134],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":29,"slug":121},"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]