[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-293661-105":59,"doc-detail-293661-en":131},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","big5-chat-shaping-llm-personalities-through-training-on-human-grounded-data","BIG5-CHAT - Shaping LLM Personalities Through Training on Human-Grounded Data","","This work embeds realistic human personality traits into large language models by addressing shortcomings of prompt-based methods that rely on behavior descriptions with realism and validity issues. It introduces BIG5-CHAT, a 100,000-dialogue dataset grounded in how humans express personality in language. Using Supervised Fine-Tuning and Direct Preference Optimization, training-based alignment matches human trait patterns more closely than prompting on BFI and IPIP-NEO, and improves reasoning with trait-dependent effects.",{"@graph":69,"@context":123},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/big5-chat-shaping-llm-personalities-through-training-on-human-grounded-data/293661/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/big5-chat-shaping-llm-personalities-through-training-on-human-grounded-data/293661.png","ImageObject",300,407,{"name":92,"@type":93},"Ben ","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-12","2026-09-17",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",18,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What problem does BIG5-CHAT address in shaping LLM personalities?","Question",{"text":113,"@type":114},"It targets the challenge that prompt-based personality methods often lack realism and validity, because they depend on questionnaire-linked behavior descriptions that may not ground how personality is expressed in language.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"What is BIG5-CHAT and how is it constructed?",{"text":118,"@type":114},"BIG5-CHAT is a large-scale dialogue dataset of 100,000 dialogues grounded in real human personality expressions in text, combining PsychGenerator and SODA through a text generation framework to induce Big Five traits within social scenarios.",{"name":120,"@type":111,"acceptedAnswer":121},"How do training-based methods compare to prompting for personality and reasoning performance?",{"text":122,"@type":114},"Supervised Fine-Tuning and Direct Preference Optimization outperform prompting on personality assessments such as BFI and IPIP-NEO, and the trained trait configurations are linked to improved performance on reasoning tasks.","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},293661,1789667300,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":140,"language":141,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":67,"update_tm":130,"read_time":145,"preview_page_count":4},2336478951081,"https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c","BIG5-CHAT: Shaping LLM Personalities  \nThrough Training on Human-Grounded Data  \nWenkai Li* Jiarui Liu* Andy Liu Xuhui Zhou Mona Diab Maarten Sap  \nLanguage Technologies Institute, Carnegie Mellon University {wenkail, jiaruil5, [msap2}@andrew.cmu.edu](msap2}@andrew.cmu.edu)  \narXiv :2410 . 1649 1v 3 [ cs .CL] 29 Jul 2025  \nAbstract  \nIn this work, we tackle the challenge of embedding realistic human personality traits into LLMs. Previous approaches have primarily focused on prompt-based methods that describe the behavior associated with the desired personality traits, suffering from realism and validity issues. To address these limitations, we introduce BIG5-CHAT, a large-scale dataset containing 100,000 dialogues designed to ground models in how humans express their personality in language. Leveraging this dataset, we explore Supervised Fine-Tuning and Direct Preference Optimization as training-based methods to align LLMs more naturally with human personality patterns. Our methods outperform prompting on personality assessments such as BFI and IPIP-NEO, with trait correlations more closely matching human data. Furthermore, our experiments reveal that models trained to exhibit higher conscientiousness, higher agreeableness, lower extraversion, and lower neuroticism display better performance on reasoning tasks, aligning with psychological findings on how these traits impact human cognitive performance. To our knowledge, this work is the first comprehensive study to demonstrate how training-based methods can shape LLM personalities through learning from real human behaviors.  \n1 Introduction  \nRealistically simulating human personality and its impact on text generation is a challenging yet crucial problem (Elster, 2015 ; Park et al., 2023 ; Serapio-García et al., 2023 ; Li et al., 2024 ; Frisch and Giulianelli, 2024) . Embedding personality traits into LLMs can greatly enhance their authenticity across a wide range of applications, from conversational agents (Pradhan and Lazar, 2021) to educational tools (Kanero et al., 2022) and mental health platforms (Tudor Car et al., 2020 ; Ahmad  \n*Equal contribution.  \net al., 2022) . By creating more human-like interactions, LLMs can better simulate diverse personasand adapt more reliably to different contexts (Gao et al., 2024a) .  \nHowever, existing methods primarily rely on prompting models with descriptions of behaviors associated with personality traits (e.g.,“You are the life of the party”; Mao et al., 2023 ; Chen et al., 2024b, 2022 ; Tu et al., 2024) . These behavior descriptions are often drawn from the same psychological questionnaires used to test their personality, raising evaluation validity concerns. More importantly, these behavioral descriptions are nonsensical for text-based LLMs (LLMs do not attend parties), failing to ground their personality in realistic patterns of how humans’ personality is expressed in their language (Vu et al., 2024) . Yet, the scarcity of large-scale, human-generated datasets annotated with personality traits has hindered the exploration of training-based approaches, limiting most prior research to prompting-based methods.  \nIn this work, we address the challenge of inducing realistic human personality traits in LLMs by constructing a large-scale dialogue dataset, BIG5-CHAT, which is grounded in real human personality expressions in text. The overview of our work is illustrated in Figure 1. We choose the well-known Big Five personality traits framework to study this (McCrae and John, 1992 ; Pittenger, 1993), due to its reliability and validity as shown from psychological research. While previous datasets typically include only persona descriptions, our dataset bridges the gap between narrowdomain personality data and general-domain social interactions, ensuring both authenticity and scenario diversity. To achieve this, we combine two primary data sources—PsychGenerator (Vu et al., 2024), a collection of 850K Facebook posts annotated with Big Fiv","cbCaiakOJyKlQg2N","https://ap.wps.com/l/cbCaiakOJyKlQg2N","pdf",838085,38,"English","# Abstract\n# 1 Introduction","[{\"question\":\"What problem does BIG5-CHAT address in shaping LLM personalities?\",\"answer\":\"It targets the challenge that prompt-based personality methods often lack realism and validity, because they depend on questionnaire-linked behavior descriptions that may not ground how personality is expressed in language.\"},{\"question\":\"What is BIG5-CHAT and how is it constructed?\",\"answer\":\"BIG5-CHAT is a large-scale dialogue dataset of 100,000 dialogues grounded in real human personality expressions in text, combining PsychGenerator and SODA through a text generation framework to induce Big Five traits within social scenarios.\"},{\"question\":\"How do training-based methods compare to prompting for personality and reasoning performance?\",\"answer\":\"Supervised Fine-Tuning and Direct Preference Optimization outperform prompting on personality assessments such as BFI and IPIP-NEO, and the trained trait configurations are linked to improved performance on reasoning tasks.\"}]","BIG5-CHAT - Shaping LLM Personalities Through Training on Human-Grounded Data | PDF",96]