[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-135757-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-135757-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","twinvoice-a-multi-dimensional-benchmark-towards-digital-twins-via-llm-persona-simulation","TwinVoice - A Multi-dimensional Benchmark Towards Digital Twins via LLM Persona Simulation","","Large Language Models (LLMs) are increasingly capable of generating human-like communication patterns, motivating their use for simulating individual behaviors and personality traits. Existing evaluations of LLM-based persona simulation are still narrow, often relying on synthetic dialogues and lacking fine-grained capability analysis. TwinVoice introduces a benchmark spanning three real-world persona dimensions—Social, Interpersonal, and Narrative—and assesses six capabilities: memory recall, logical reasoning, opinion consistency, lexical fidelity, persona tone, and syntactic style.",{"@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/twinvoice-a-multi-dimensional-benchmark-towards-digital-twins-via-llm-persona-simulation/135757/",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/twinvoice-a-multi-dimensional-benchmark-towards-digital-twins-via-llm-persona-simulation/135757.png","ImageObject",300,407,{"name":42,"@type":43},"Bill Black","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-17","2026-08-21",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",18,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What does TwinVoice benchmark for LLM-based persona simulation?","Question",{"text":63,"@type":64},"TwinVoice benchmarks persona simulation across three dimensions—Social Persona, Interpersonal Persona, and Narrative Persona—and evaluates six capabilities including memory recall, logical reasoning, opinion consistency, lexical fidelity, persona tone, and syntactic style.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"Why is persona consistency important for digital twins?",{"text":68,"@type":64},"A complete digital twin requires stable and authentic identity across diverse real-world interactions, which depends on the model’s ability to maintain consistent persona over time.",{"name":70,"@type":61,"acceptedAnswer":71},"What capability gaps do advanced models still show in TwinVoice results?",{"text":72,"@type":64},"Even though advanced models achieve moderate accuracy, they still fall short in capabilities such as syntactic style and memory recall.","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},135757,1787319346,{"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},24189269381491,"https://ap-avatar.wpscdn.com/avatar/160000cf11732dd8392?x-image-process=image/resize,m_fixed,w_180,h_180&k=1788146458752108895","TwinVoice: A Multi-dimensional Benchmark Towards Digital Twins via  \nLLM Persona Simulation  \nBangde Du 1 * Minghao Guo2 * Songming He3 Ziyi Ye3 Xi Zhu2 Weihang Su 1  \nShuqi Zhu 1 Yujia Zhou 1 Yongfeng Zhang2 Qingyao Ai 1† Yiqun Liu 1  \n1Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China  \n2Rutgers University, New Brunswick, NJ, USA  \n3Institute of Trustworthy Embodied AI, Fudan University, Shanghai, China  \nFigure 1: The conceptual framework of TwinVoice: (Left) The evaluation is structured across three dimensions that represent different aspects of a persona: (1) Social Persona that reflects its behavior on social platforms; Interpersonal Persona that reflects its private interaction, and Narrative Persona that corresponds to a fictional scenario. The LLMs are prompted with a person’s historical context to simulate their behavior and evaluated by six fundamental capabilities: memory recall, logical reasoning, opinion consistency, lexical fidelity, persona tone, and syntactic style. (Right) Experimental results on modern LLMs averaged over three dimensions.  \nAbstract  \nLarge Language Models (LLMs) are exhibiting emergent human-like abilities and are envisioned as the tool for simulating an individual’s communication patterns, behaviors, and personality traits. However, current evaluations of LLM-based persona simulation remain limited: most rely on synthetic dialogues and lack finegrained analysis of the capability for persona simulation. To address these limitations, we introduce TwinVoice, a comprehensive benchmark for assessing persona simulation across diverse real-world contexts. TwinVoice encompasses three dimensions: Social Persona (public social interactions), Interpersonal Persona (private dialogues), and Narrative Persona (rolebased expression) . It further decomposes the evaluation into six fundamental capabilities, including opinion consistency, memory recall, logical reasoning, lexical fidelity, persona tone, and syntactic style. Experimental results reveal that while advanced models achieve moderate  \n*Equal contribution.  \n†Corresponding author: [aiqy@tsinghua.edu.cn](aiqy@tsinghua.edu.cn)  \naccuracy in persona simulation, they still fall short of capabilities such as syntactic style and memory recall. Our data, code, and evaluation results are available 1.  \n1 Introduction  \nLarge Language Models (LLMs) are rapidly evolving from text generation tools into human-like agents (Bubeck et al., 2023 ; Wei et al., 2022 ; Changet al., 2024) . Existing studies have shown that the most advanced LLMs are capable of producing text indistinguishable from human writing (Jones and Bergen, 2025 ; Jones et al., 2025 ; Jones and Bergen, 2024) . Consequently, the research focus is shifting toward the ambitious vision of constructing “digital twins”, AI agents capable of replicating the unique communication patterns and personality traits of specific individuals. However, the realization of a complete digital twin hinges upon a critical and non-negotiable prerequisite: persona consistency. Before complex longitudinal behav-  \n1 [https://github.com/TwinVoice/TwinBench](https://github.com/TwinVoice/TwinBench)  \n19604  \nFindings of the Association for Computational Linguistics: ACL 2026 , pages 19604–19628 July 2-7, 2026 ©2026 Association for Computational Linguistics  \niors can be simulated, a model must first demonstrate the foundational ability to maintain a stable and authentic identity across diverse, real-world interactions. In this paper, we focus on evaluating LLM-based persona simulation (Park et al., 2023) as the fundamental building block toward this vision. Distinct from generic role-playing (Shanahan et al., 2023), which often relies on static, descriptive prompts (e.g.,“acting like a pirate”), or personalization, which primarily focuses on adapting to user preferences, persona consistency emphasizes deep behavioral mimicry. It requires the model to replicate the specific, longitu","cbCaivr930VLLNTn","https://ap.wps.com/l/cbCaivr930VLLNTn","pdf",1641891,25,"English","# Abstract\n# Introduction\n## Persona consistency as a prerequisite for digital twins\n## LLM-based persona simulation and core evaluation gap\n## Related benchmarks and directions","[{\"question\":\"What does TwinVoice benchmark for LLM-based persona simulation?\",\"answer\":\"TwinVoice benchmarks persona simulation across three dimensions—Social Persona, Interpersonal Persona, and Narrative Persona—and evaluates six capabilities including memory recall, logical reasoning, opinion consistency, lexical fidelity, persona tone, and syntactic style.\"},{\"question\":\"Why is persona consistency important for digital twins?\",\"answer\":\"A complete digital twin requires stable and authentic identity across diverse real-world interactions, which depends on the model’s ability to maintain consistent persona over time.\"},{\"question\":\"What capability gaps do advanced models still show in TwinVoice results?\",\"answer\":\"Even though advanced models achieve moderate accuracy, they still fall short in capabilities such as syntactic style and memory recall.\"}]","TwinVoice - A Multi-dimensional Benchmark Towards Digital Twins via LLM Persona Simulation | PDF",63]