[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-273770-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-273770-en":130},{"code":4,"msg":5,"data":6},0,"ok",{"site_id":7,"language":8,"slug":9,"title":10,"keywords":11,"description":12,"schema_data":13,"social_meta":74,"head_meta":76,"extra_data":78,"updated_unix":80},105,"en","editppt-faithful-long-deck-slide-editing-via-structured-tool-using-multi-agent-with-dual-modal-validators","EDITPPT - Faithful Long-Deck Slide Editing via Structured Tool-Using Multi-Agent with Dual-Modal Validators","","Slide editing automation must satisfy modification accuracy, preservation fidelity, and robustness as deck length increases. Many LLM-based approaches struggle with real presentation files because they depend on idealized intermediate representations or open-ended code generation, which can trigger cascading errors across long decks. EDITPPT reformulates slide editing as a constrained tool-selection task, executing localized shape-level operations through the native PowerPoint COM interface while preserving application-resolved structure. Dual-modal validation improves assessment of both instruction fidelity and visual quality, supported by DECKEDITBENCH.",{"@graph":14,"@context":73},[15,34,56],{"@type":16,"itemListElement":17},"BreadcrumbList",[18,23,27,31],{"item":19,"name":20,"@type":21,"position":22},"https://docshare.wps.com","Home","ListItem",1,{"item":24,"name":25,"@type":21,"position":26},"https://docshare.wps.com/document/","Document",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/document/technology/","Technology",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/editppt-faithful-long-deck-slide-editing-via-structured-tool-using-multi-agent-with-dual-modal-validators/273770/",4,{"url":32,"name":10,"@type":35,"image":36,"author":41,"headline":10,"publisher":44,"fileFormat":47,"inLanguage":8,"description":12,"dateModified":48,"datePublished":49,"encodingFormat":47,"isAccessibleForFree":50,"interactionStatistic":51},"DigitalDocument",{"url":37,"@type":38,"width":39,"height":40},"https://docshare.wps.com/thumbnails/editppt-faithful-long-deck-slide-editing-via-structured-tool-using-multi-agent-with-dual-modal-validators/273770.png","ImageObject",300,407,{"name":42,"@type":43},"Seraphina","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-20","2026-09-15",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",10,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What are the main constraints slide editing automation must satisfy?","Question",{"text":63,"@type":64},"It must achieve modification accuracy, preserve unrelated content and formatting faithfully, and remain robust as slide deck length and complexity grow.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"Why do many LLM-based slide editing systems fail on real presentation files?",{"text":68,"@type":64},"They often rely on idealized intermediate representations or open-ended code generation, which can mismatch native .pptx structure and cause cascading errors in long decks.",{"name":70,"@type":61,"acceptedAnswer":71},"How does EDITPPT perform slide editing while improving reliability?",{"text":72,"@type":64},"EDITPPT treats editing as a constrained tool-selection problem, runs deterministic localized shape-level operations via the PowerPoint COM interface, and uses dual-modal validation to assess instruction fidelity and visual quality.","https://schema.org",{"og:url":32,"og:type":75,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":77,"canonical":32},"index,follow",{"doc_id":79,"site_id":7},273770,1789472755,{"code":4,"msg":82,"data":83},"success",[84,88,92,96,101,105,110,115,120,123,126],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":85,"show_sort_weight":86,"slug":87},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":89,"show_sort_weight":90,"slug":91},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":93,"show_sort_weight":94,"slug":95},"Exam",70,"exam",{"id":97,"doc_module":4,"doc_module_name":25,"category_name":98,"show_sort_weight":99,"slug":100},5,"Comic",60,"comic",{"id":102,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":103,"slug":104},6,50,"technology",{"id":106,"doc_module":4,"doc_module_name":25,"category_name":107,"show_sort_weight":108,"slug":109},7,"Healthcare",40,"healthcare",{"id":111,"doc_module":4,"doc_module_name":25,"category_name":112,"show_sort_weight":113,"slug":114},8,"Research & Report",30,"research-report",{"id":116,"doc_module":4,"doc_module_name":25,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":25,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":55,"doc_module":4,"doc_module_name":25,"category_name":124,"show_sort_weight":55,"slug":125},"Lifestyle","lifestyle",{"id":127,"doc_module":4,"doc_module_name":25,"category_name":128,"show_sort_weight":97,"slug":129},19,"General","general",{"code":4,"msg":82,"data":131},{"doc_id":79,"user_id":132,"nickname":42,"user_avatar":133,"doc_module":4,"category_id":102,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":113,"language":139,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":140,"faqs":141,"seo_title":142,"seo_description":12,"update_tm":80,"read_time":143},962075114101,"https://ap-avatar.wpscdn.com/avatar/e000253a75eb197efd?x-image-process=image/resize,m_fixed,w_180,h_180&k=1780044092746381165","EditPPT: Faithful Long-Deck Slide Editing via Structured Tool-Using Multi-Agent with Dual-Modal Validators  \nJiheon Kim1 Kyudan Jung2 Jaegul Choo2  \nKAIST AI  \n{ben8169, kyudan, [jchoo}@kaist.ac.kr](jchoo}@kaist.ac.kr)  \narXiv : 2608 .2038 1v 1 [cs .CL] 29 Jun 2026  \nAbstract  \nAutomating slide editing requires simultaneously satisfying modification accuracy, preservation fidelity, and robustness to deck length. Existing LLM-based systems often fail on realworld presentation files because they rely on idealized intermediate representations or openended code generation, which are prone to cascading errors in long decks. We introduce EDITPPT, a multi-agent framework that reformulates slide editing as a constrained toolselection problem. By executing localized shape-level operations through the native PowerPoint COM interface, EDITPPT narrows the LLM action space while preserving the application-resolved structure of user-authored decks. By separating validation across modalities, our dual-modal validation provides more robust assessment of both instruction fidelity and visual quality. We also present DECKEDITBENCH, a benchmark with 28 human-authored decks, 582 slides, and 183 editing prompts across short, medium, and long deck tiers. Experiments show that EDITPPT achieves a 99.5% execution rate, 88.7% slide-targeting F1, 82.5% instruction following, and 91.5% object preservation overall, while maintaining strong performance on long decks. Our code and benchmark are available at here.  \n1 Introduction  \nSlide decks are a ubiquitous medium for organizing and communicating information across business, academic, and professional contexts (Shi et al., 2025) . Naturally, automating slide workflows has emerged as a high-value task. Recent systems have made rapid progress in presentation generation from natural-language instructions, source documents, and user preferences (Ge et al., 2025 ; Zheng et al., 2025 ; Yang et al., 2025 ; Zeng et al., 2025 ; Zheng et al., 2026 ; Wu et al., 2026); however, reliable slide editing remains a significant challenge (Jung et al., 2026 ; Ofengenden et al., 2025 ;  \nyour-slides + PPTX Skill  \nInstruction Following (%)  \n Preserv. (%)   \nAvg. Cost  \n(×10−2 $/slide)  \nFigure 1: EDITPPT maintains strong editing across deck lengths while preserving unrequested objects.  \nJang et al., 2026) . Unlike generation, slide editing is strictly bound by three simultaneous constraints: modification accuracy, preservation fidelity, and robustness to deck length. An editing system must execute a requested modification while leaving unrelated content and formatting untouched. Crucially, this preservation behavior must remain stable asthe presentation grows in length and complexity, since unintended shifts in fonts, layouts, or shapes can render an entire deck unusable.  \nA central obstacle is the mismatch between idealized editing representations and native .pptx files. Many LLM-based pipelines operate on regularized HTML, code, or structured markup representations where layers, coordinates, and object boundaries are explicit (Ge et al., 2025 ; Zheng et al., 2025 ; Yang et al., 2025 ; Jang et al., 2026) . User-authored presentations rarely satisfy these assumptions because they include grouped objects, implicit formatting, inconsistent layouts, and artifacts whose semantic role differs from their file-level type (Wu et al., 2026 ; Pan et al., 2026) . Direct manipulation of native files through code generation or OOXML editing avoids some representational loss, but it requires the LLM to synthesize brittle stateful pro-  \n(a) (b)  \nPlanner   \nNatural Language Request  \n(f)  \nMulti-Agent Dispatcher  \n(d) Specialized sub-agents  \nText_style   \nTable  \nChart  \nShape_layout  \nSlide  \n(c)  \nCycle  \n(e) Differently defined tools per sub-agent  \nTranslate all the pages into English.  \nBefore edit  \nAfter edit  \nComplete  \nsub-plan 3  \n|  |  |  |  |  |  |  |\n| --- | --- | --- | --- | --- | --- | --- |\n|  |  |  | Tool 2 |  |  |  ","cbCaiqecHMFhubS1","https://ap.wps.com/l/cbCaiqecHMFhubS1","pdf",3094167,"English","# Introduction\n## Core challenge in slide editing\n## EDITPPT approach: constrained tool selection\n## Tool execution and modality-decoupled validation\n## DECKEDITBENCH evaluation","[{\"question\":\"What are the main constraints slide editing automation must satisfy?\",\"answer\":\"It must achieve modification accuracy, preserve unrelated content and formatting faithfully, and remain robust as slide deck length and complexity grow.\"},{\"question\":\"Why do many LLM-based slide editing systems fail on real presentation files?\",\"answer\":\"They often rely on idealized intermediate representations or open-ended code generation, which can mismatch native .pptx structure and cause cascading errors in long decks.\"},{\"question\":\"How does EDITPPT perform slide editing while improving reliability?\",\"answer\":\"EDITPPT treats editing as a constrained tool-selection problem, runs deterministic localized shape-level operations via the PowerPoint COM interface, and uses dual-modal validation to assess instruction fidelity and visual quality.\"}]","EDITPPT - Faithful Long-Deck Slide Editing via Structured Tool-Using Multi-Agent with Dual-Modal Validators | PDF",76]