[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-293555-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-293555-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","from-text-to-emoji-how-peft-driven-personality-manipulation-unleashes-the-emoji-potential-in-llms","From Text to Emoji - How PEFT-Driven Personality Manipulation Unleashes the Emoji Potential in LLMs","","The manipulation of large language model (LLM) personality traits is a growing research focus, yet existing methods such as prompt-based in-context knowledge editing and gradient-based model editor networks can produce irregular, variable, or gibberish outputs. This paper uses opinion QA based parameter-efficient fine-tuning with QLoRA to control Big Five traits and observes latent emoji generation after PEFT. Explainability analyses confirm intentional emoji usage linked to activated or amplified neurons, and benchmark metrics. Findings highlight more consistent personality expression than IKE.",{"@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/from-text-to-emoji-how-peft-driven-personality-manipulation-unleashes-the-emoji-potential-in-llms/293555/",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/from-text-to-emoji-how-peft-driven-personality-manipulation-unleashes-the-emoji-potential-in-llms/293555.png","ImageObject",300,407,{"name":42,"@type":43},"Ava Thompson","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-29","2026-09-17",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",11,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What problem does the paper address in LLM personality manipulation?","Question",{"text":63,"@type":64},"It addresses irregular and inconsistent outputs from existing approaches like in-context knowledge editing and model editor networks, which are sensitive to prompts and can yield unreliable behavior.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"How does the proposed method manipulate personality traits?",{"text":68,"@type":64},"It applies opinion QA based parameter-efficient fine-tuning using QLoRA to adjust Big Five traits: Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism.",{"name":70,"@type":61,"acceptedAnswer":71},"What new behavior appears after PEFT, and how is it validated?",{"text":72,"@type":64},"After PEFT, models generate emojis related to certain traits even though emojis were absent from the PEFT data. ICL explainability verifies the emojis are not random artifacts, and mechanistic interpretability links the behavior to specific activated or amplified neurons.","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},293555,1789665516,{"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},1649267921044,"https://us-avatar.wpscdn.com/avatar/1800007509477c92dfb?_k=1786009248482753345","From Text to Emoji: How PEFT-Driven Personality Manipulation Unleashes the Emoji Potential in LLMs  \nNavya Jain1,2 , Zekun Wu1,2 * , Cristian Munoz1 , Airlie Hilliard1 , Xin Guan1  \nPhilip Treleaven2 , Emre Kazim1 , Adriano Koshiyama1  \n1Holistic AI, 2University College London  \nAbstract  \nThe manipulation of the personality traits of large language models (LLMs) has emerged as a key area of research. Methods like prompt-based In-Context Knowledge Editing (IKE) and gradient-based Model Editor Networks (MEND) have been explored but show irregularity and variability; IKE depends on the prompt, leading to variability and sensitivity, while MEND yields inconsistent and gibberish outputs. To address this, we employed Opinion QA Based Parameter-Efficient Fine-Tuning (PEFT), specifically Quantized Low-Rank Adaptation (QLoRA), to manipulate the Big Five personality traits: Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism. After PEFT, models such as Mistral-7B-Instruct and LLaMA- 2-7B-chat showed a latent behaviour by generating emojis for certain traits, despite no emojis being present in the PEFT data. For instance, LLaMA-2-7B-chat generated emojis in 99.5% of extraversion-related test instances, while Mistral-7B-Instruct did so in 92.5% of openness-related test instances. ICL Explainability analysis indicated that the LLMs used emojis intentionally to express these traits. Mechanistic Interpretability analysis showed that this latent behaviour of LLMs could be traced to specific neurons that became activated or amplified after PEFT. This paper provides a number of novel contributions. First, introducing an Opinion QA dataset for PEFT-driven personality manipulation; second, developing metric models to benchmark LLM personality traits; third, demonstrating PEFT’s superiority over IKE in personality manipulation; and finally, analysing and validating emoji usage through explainability methods such as Mechanistic Interpretability and In-context learning Explainability methods.  \n* Corresponding author: [zekun.wu@holisticai.com](zekun.wu@holisticai.com)  \n1 Introduction  \nAs Large Language Models (LLMs) become more integral across various industries, there is increasing interest in enhancing not only their capabilities like reasoning, planning, comprehension, but also their ability to exhibit personality traits (Hilliardet al., 2024 ; Hu et al., 2024 ; Serapio-García et al., 2023 ; Dan et al., 2024) . Psychological research has shown that personality traits significantly influence human communication, including tone and verbosity (Liu and Sun, 2020 ; Kennison et al., 2024), raising the question of whether LLMs can be manipulated to exhibit similar expressions to make their communication more nuanced and adaptable. Several frameworks, such as the Dark Triad (Jonason and Webster, 2010), the 16 Personality Factors (Cattell and Mead, 2008), and the Big Five Personality Model (Gosling et al., 2003), have been used to analyse LLMs’ personality. Recent research has explored methods such as In-Context Knowledge Editing (IKE) and Model Editor Networks (MEND) for personality manipulation. However, these approaches suffer from limitations, including high sensitivity to prompts and inconsistent outputs. This paper addresses the challenges of personality manipulation in LLMs by introducing a novel approach grounded in the Big Five personality model. We present a new Opinion QA dataset and methodologies for systematically adjusting personality traits in LLMs. Utilising Quantized Low-Rank Adaptation (QLoRA), a method within Parameter-Efficient Fine-Tuning (PEFT) (Dettmerset al., 2023a), we demonstrate that LLMs can achieve more consistent and enduring personality expressions.  \nThis approach enhances model adaptability in interactions and reveals new behaviours, such as spontaneous emoji generation in Mistral-7BInstruct and LLaMA-7B-Chat, absent in the original models, following PEFT-based manipulations.  \n4702  \nFindings of t","cbCaidEMioq1vhWR","https://ap.wps.com/l/cbCaidEMioq1vhWR","pdf",7826132,37,"English","# Abstract\n# Introduction\n# Related Work\n## Parameter-Efficient Fine-Tuning: LoRA and QLoRA","[{\"question\":\"What problem does the paper address in LLM personality manipulation?\",\"answer\":\"It addresses irregular and inconsistent outputs from existing approaches like in-context knowledge editing and model editor networks, which are sensitive to prompts and can yield unreliable behavior.\"},{\"question\":\"How does the proposed method manipulate personality traits?\",\"answer\":\"It applies opinion QA based parameter-efficient fine-tuning using QLoRA to adjust Big Five traits: Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism.\"},{\"question\":\"What new behavior appears after PEFT, and how is it validated?\",\"answer\":\"After PEFT, models generate emojis related to certain traits even though emojis were absent from the PEFT data. ICL explainability verifies the emojis are not random artifacts, and mechanistic interpretability links the behavior to specific activated or amplified neurons.\"}]","From Text to Emoji - How PEFT-Driven Personality Manipulation Unleashes the Emoji Potential in LLMs | PDF",93]