[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81822-en":3,"doc-seo-81822-105":31,"detail-sidebar-cat-0-en-105":93},{"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":28,"seo_description":14,"update_tm":29,"read_time":30},81822,4398048950312,"Violet","https://ap-avatar.wpscdn.com/avatar/400002538284de19e3c?_k=1778320343897328908",8,"Research & Report","CoPersona Collaborative Persona Graphs for Robust LLM Personalization","Real-world LLM personalization often fails under sparse, skewed user histories where key persona attributes remain weakly observed and incomplete. CoPersona introduces a collaborative persona completion framework that augments a target user by borrowing signals from behaviorally similar peers, while correcting facet coverage bias. It decomposes interactions into multiple facet-level representations and aligns peers via a multiplex persona graph. A dual-branch design combines non-parametric peer retrieval with parametric graph reasoning, improving robustness across domains and model scales.","CoPersona: Collaborative Persona Graphs for Robust LLM Personalization  \nYangtian Zhang∗ [yangtian.zhang@yale.edu](yangtian.zhang@yale.edu)[ ](yangtian.zhang@yale.edu)Yale University  \nNew Haven, Connecticut, USA  \nLeyao Wang∗ [leyao.wang.lw855@yale.edu](leyao.wang.lw855@yale.edu)  \nYale University New Haven, Connecticut, USA  \nHiren Madhu [hiren.madhu@yale.edu](hiren.madhu@yale.edu)  \nYale University New Haven, Connecticut, USA  \nNgoc Bui[ngocbh.pt@gmail.com](ngocbh.pt@gmail.com)  \nYale University New Haven, Connecticut, USA  \nWalter Roznyatovskiy [vlad.r@samsung.com](vlad.r@samsung.com)  \nSamsung Mountain View, California, USA  \nRex Ying [rex.ying@yale.edu](rex.ying@yale.edu)  \nYale University New Haven, Connecticut, USA  \narXiv :2607 .0 1485v 1 [ cs .IR] 1 Jul 2026  \nAbstract  \nReal-world LLM personalization is often constrained by sparse and skewed user histories: most users provide only a handful of interactions, while even frequent users’ logs capture an incomplete and biased view of their preferences. As a result, weakly observed user attributes are difficult to infer, leading to brittle personalization when test-time requests shift toward under-supported facets.  \nMotivated by this limitation, we present CoPersona, a graphbased collaborative personalization framework that completes sparse user profiles by borrowing signals from behaviorally similar peers. However, directly transferring signals is difficult because uneven facet coverage introduces bias into interaction histories, obscuring user similarity in the unstructured global space. To address this issue, CoPersona decomposes interaction histories into multiple facet-level representations and explicitly models peer-to-peer, facetlevel alignment through a multiplex persona graph. To effectively leverage peer information at inference time, we employ a dualbranch architecture that combines non-parametric peer retrieval with parametric graph reasoning. Experiments across multiple domains and model scales demonstrate consistent improvements over strong baselines, validating CoPersona as an effective approach for robust LLM personalization.  \nCCS Concepts  \n• Information systems → Personalization.  \nKeywords  \nLarge Language Models, Personalization, Collaborative Learning, Graph Neural Networks, Retrieval-Augmented Generation  \nACM Reference Format:  \nYangtian Zhang, Leyao Wang, Hiren Madhu, Ngoc Bui, Walter Roznyatovskiy, and Rex Ying. 2026. CoPersona: Collaborative Persona Graphs for Robust LLM Personalization. In Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD’26), August 09–13,  \n∗ Both authors contributed equally to this research.  \nThis work is licensed under a Creative Commons Attribution 4 .0 International License. KDD’26, Jeju Island, Republic of Korea  \n© 2026 Copyright held by the owner/author(s) .  \nACM ISBN 979-8-4007-2259-2/2026/08  \n[https://doi.org/10.1145/3770855.3817892](https://doi.org/10.1145/3770855.3817892)  \n2026, Jeju Island, Republic of Korea. ACM, New York, NY, USA, 12 pages. [https://doi.org/10.1145/3770855.3817892](https://doi.org/10.1145/3770855.3817892)  \n1 Introduction  \nLarge language models (LLMs) have rapidly expanded in capability, powering applications from conversational assistants and writing tools to decision support and recommendation. Despite these advances, most LLMs remain largely preference-agnostic: they are optimized for population-level behavior and often overlook the goals, values, and writing style of a particular individual [7, 9, 50] . In many practical settings, however, the same request can warrant different responses for different users, motivating LLM personalization—adapting generation using user-specific information to improve usefulness and user-level alignment [2, 32, 62] .  \nA dominant line of work [21, 24, 34, 45, 49] instantiates personalization through a memory–retrieval pipeline: given a user’s current query, the system retrieves relevant snippets from the user’","cbCaigPAxo7otgpC","https://ap.wps.com/l/cbCaigPAxo7otgpC","pdf",2753049,6,1,12,"English","en",105,"# Introduction\n## Motivation: Facet-level cold-start and bias\n## Collaborative remedy via peer signals\n## Overview of CoPersona approach","[{\"question\":\"Why does LLM personalization become brittle in real-world settings?\",\"answer\":\"Because user histories are sparse and skewed, missing persona facets lead to weak inference and cause failures when test requests rely on under-supported attributes.\"},{\"question\":\"What problem does CoPersona address with peer collaboration?\",\"answer\":\"Direct peer signal transfer is hindered by uneven facet coverage, which introduces bias and obscures which facets truly match between users.\"},{\"question\":\"How does CoPersona use graphs and retrieval to improve personalization?\",\"answer\":\"It builds facet-level representations and uses a multiplex persona graph to align peers, then applies a dual-branch architecture that combines peer retrieval with graph reasoning during inference.\"}]","CoPersona Collaborative Persona Graphs for Robust LLM Personalization | PDF",1784176377,30,{"code":4,"msg":32,"data":33},"ok",{"site_id":25,"language":24,"slug":34,"title":13,"keywords":35,"description":14,"schema_data":36,"social_meta":88,"head_meta":90,"extra_data":92,"updated_unix":29},"copersona-collaborative-persona-graphs-for-robust-llm-personalization","",{"@graph":37,"@context":87},[38,55,70],{"@type":39,"itemListElement":40},"BreadcrumbList",[41,45,49,52],{"item":42,"name":43,"@type":44,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":46,"name":47,"@type":44,"position":48},"https://docshare.wps.com/document/","Document",2,{"item":50,"name":12,"@type":44,"position":51},"https://docshare.wps.com/document/research-report/",3,{"item":53,"name":13,"@type":44,"position":54},"https://docshare.wps.com/document/copersona-collaborative-persona-graphs-for-robust-llm-personalization/81822/",4,{"url":53,"name":13,"@type":56,"author":57,"headline":13,"publisher":59,"fileFormat":62,"inLanguage":24,"description":14,"dateModified":63,"datePublished":64,"encodingFormat":62,"isAccessibleForFree":65,"interactionStatistic":66},"DigitalDocument",{"name":9,"@type":58},"Person",{"url":42,"name":60,"@type":61},"DocShare","Organization","application/pdf","2026-08-03","2026-07-16",true,{"@type":67,"interactionType":68,"userInteractionCount":20},"InteractionCounter",{"@type":69},"ViewAction",{"@type":71,"mainEntity":72},"FAQPage",[73,79,83],{"name":74,"@type":75,"acceptedAnswer":76},"Why does LLM personalization become brittle in real-world settings?","Question",{"text":77,"@type":78},"Because user histories are sparse and skewed, missing persona facets lead to weak inference and cause failures when test requests rely on under-supported attributes.","Answer",{"name":80,"@type":75,"acceptedAnswer":81},"What problem does CoPersona address with peer collaboration?",{"text":82,"@type":78},"Direct peer signal transfer is hindered by uneven facet coverage, which introduces bias and obscures which facets truly match between users.",{"name":84,"@type":75,"acceptedAnswer":85},"How does CoPersona use graphs and retrieval to improve personalization?",{"text":86,"@type":78},"It builds facet-level representations and uses a multiplex persona graph to align peers, then applies a dual-branch architecture that combines peer retrieval with graph reasoning during inference.","https://schema.org",{"og:url":53,"og:type":89,"og:title":13,"og:site_name":60,"og:description":14},"article",{"robots":91,"canonical":53},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":94},[95,99,103,107,112,116,121,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":47,"category_name":96,"show_sort_weight":97,"slug":98},"Story & Novel",90,"story-novel",{"id":48,"doc_module":4,"doc_module_name":47,"category_name":100,"show_sort_weight":101,"slug":102},"Literature",80,"literature",{"id":54,"doc_module":4,"doc_module_name":47,"category_name":104,"show_sort_weight":105,"slug":106},"Exam",70,"exam",{"id":108,"doc_module":4,"doc_module_name":47,"category_name":109,"show_sort_weight":110,"slug":111},5,"Comic",60,"comic",{"id":20,"doc_module":4,"doc_module_name":47,"category_name":113,"show_sort_weight":114,"slug":115},"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":47,"category_name":118,"show_sort_weight":119,"slug":120},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":47,"category_name":12,"show_sort_weight":30,"slug":122},"research-report",{"id":124,"doc_module":4,"doc_module_name":47,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":47,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":47,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":47,"category_name":137,"show_sort_weight":108,"slug":138},19,"General","general"]