[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-293168-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-293168-en":131},{"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","charactergpt-a-persona-reconstruction-framework-for-role-playing-agents","CharacterGPT - A Persona Reconstruction Framework for Role-Playing Agents","","The introduction of the Assistants API enables large language models to power role-playing agents, yet maintaining consistent character personas remains difficult because information extraction often omits critical elements like backstory and interpersonal relationships. CharacterGPT proposes Character Persona Training (CPT), which incrementally reconstructs personas by extracting eight traits from chapter-wise novel summaries. Big Five evaluations and creative narrative tasks demonstrate improved persona consistency, controllability, and role-specific knowledge.",{"@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/research-report/","Research & Report",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/charactergpt-a-persona-reconstruction-framework-for-role-playing-agents/293168/",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/charactergpt-a-persona-reconstruction-framework-for-role-playing-agents/293168.png","ImageObject",300,407,{"name":42,"@type":43},"Jiven","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-28","2026-09-17",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",14,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What problem does CharacterGPT address in role-playing agents?","Question",{"text":63,"@type":64},"CharacterGPT targets persona inconsistency caused by unstructured or incomplete information extraction that often drops key traits such as backstory or relationships.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"How does CharacterGPT reconstruct character personas?",{"text":68,"@type":64},"It uses Character Persona Training (CPT) to incrementally extract eight essential traits from chapter-wise novel summaries and append them to a chronologically evolving persona document.",{"name":70,"@type":61,"acceptedAnswer":71},"How is CharacterGPT evaluated in the paper?",{"text":72,"@type":64},"The framework is assessed with Big Five Inventory (BFI) personality consistency tests and creative tasks where crowd-workers rate short narratives across multiple metrics.","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},293168,1789660563,{"code":4,"msg":82,"data":83},"success",[84,88,92,96,101,106,111,115,120,123,127],{"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":103,"show_sort_weight":104,"slug":105},6,"Technology",50,"technology",{"id":107,"doc_module":4,"doc_module_name":25,"category_name":108,"show_sort_weight":109,"slug":110},7,"Healthcare",40,"healthcare",{"id":112,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":113,"slug":114},8,30,"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":124,"doc_module":4,"doc_module_name":25,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":25,"category_name":129,"show_sort_weight":97,"slug":130},19,"General","general",{"code":4,"msg":82,"data":132},{"doc_id":79,"user_id":133,"nickname":42,"user_avatar":134,"doc_module":4,"category_id":112,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":140,"language":141,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":12,"update_tm":80,"read_time":145,"preview_page_count":4},1099513958607,"https://ap-avatar.wpscdn.com/avatar/100002390cf8733938c?x-image-process=image/resize,m_fixed,w_180,h_180&k=1778829742770036399","CharacterGPT: A Persona Reconstruction Framework for Role-Playing  \nAgents  \nJeiyoon Park 1 , Chanjun Park2†, Heuiseok Lim2†  \n1 SOOP, 2 Korea University  \n[naruto@sooplive.com](naruto@sooplive.com)  \n{bcj1210, [limhseok}@korea.ac.kr](limhseok}@korea.ac.kr)  \narXiv :2405 . 19778v5 [ cs .CL] 23 Feb 2025  \nAbstract  \nThe recent introduction of the Assistants API highlights its potential for large language models (LLMs) in role-playing agents (RPA) . However, maintaining consistent character personas remains a significant challenge due to variability in information extraction, which frequently omits critical elements such as backstory or interpersonal relationships. To address this limitation, we introduce CharacterGPT, a framework designed to dynamically reconstruct character personas through Character Persona Training (CPT) . This approach incrementally updates personas by extracting traits from chapter-wise novel summaries, reflecting the progression of the narrative. Our framework is evaluated through Big Five personality evaluations and creative tasks, in which characters generate original narratives, demonstrating the efficacy of CharacterGPT in preserving persona consistency. The code and results are available at [https://github](https://github) . com/Jeiyoon/charactergpt  \n1 Introduction  \nThe rapid advancements in large language models (LLMs) have positioned them as the core module of various AI systems (OpenAI, 2023a,c; Anthropic, 2023 ; Google, 2024 ; DeepSeek-AI et al., 2024, 2025), enabling a wide range of applications. Building on this progress, the recent introduction of the Assistants API (OpenAI, 2023b), a tool designed for document-based information retrieval, demonstrates the potential of LLM in multiple domains, especially in role-playing agents (RPA) (Kim et al., 2019 ; Yu et al., 2023 ; Jiang et al., 2023 ; Park et al., 2023 ; Wang et al., 2023b ; Zhang et al., 2024 ; Kong et al., 2024 ; Wang et al., 2025) . However, RPAs that rely solely on documents as input often face problems of inconsistent information extraction,  \n† Corresponding Author  \nFigure 1: Comparison of response accuracy between persona-based GPT-4 assistants utilizing unstructured versus structured character traits as input. When provided with unstructured traits, the assistant demonstrates limited success in generating accurate responses. In contrast, the use of structured traits significantly improves the correctness of the assistant’s responses.  \nwhere key personality traits or background knowledge are omitted, leading to degraded persona coherence (Sadeq et al., 2024) . For example, as illustrated in Figure 1, when the Assistants API is provided with an unstructured Wiki document about the novel God’s Blessing on This Wonderful World!, it often fails to provide accurate responses, while  \nstructured character traits produce more reliable, role-specific answers.  \nIn this paper, we propose a novel framework called CharacterGPT, which addresses this challenge through a structured persona reconstruction process. Drawing inspiration from cognitive memory models, we introduce Character Persona Training (CPT), a method that incrementally updates character personas by extracting traits from chapterwise summaries of novels. This approach mirrors how human memory consolidates information into schemas over time (van Kesteren and Meeter, 2020), enabling more consistent and contextually appropriate responses from RPAs.  \nCPT operates by identifying eight essential traits—personality, physical description, motivations, backstory, emotions, relationships, growth and change, and conflict—based on character analysis literature (Forster, 1927 ; Reams, 2015) . For each chapter of a novel, these traits are extracted from summaries and appended to a character’s evolving persona, forming a document that reflects the character’s development in chronological order. Note that the extracted traits are updated separately to ensure they remain distinct and are no","cbCaipkuO8EPE8Dy","https://ap.wps.com/l/cbCaipkuO8EPE8Dy","pdf",1369437,17,"English","# Abstract\n# Introduction\n# Proposed Method\n## Preliminaries","[{\"question\":\"What problem does CharacterGPT address in role-playing agents?\",\"answer\":\"CharacterGPT targets persona inconsistency caused by unstructured or incomplete information extraction that often drops key traits such as backstory or relationships.\"},{\"question\":\"How does CharacterGPT reconstruct character personas?\",\"answer\":\"It uses Character Persona Training (CPT) to incrementally extract eight essential traits from chapter-wise novel summaries and append them to a chronologically evolving persona document.\"},{\"question\":\"How is CharacterGPT evaluated in the paper?\",\"answer\":\"The framework is assessed with Big Five Inventory (BFI) personality consistency tests and creative tasks where crowd-workers rate short narratives across multiple metrics.\"}]","CharacterGPT - A Persona Reconstruction Framework for Role-Playing Agents | PDF",43]