[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-501590-105":59,"doc-detail-501590-en":130},{"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":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","tuning-free-personalized-alignment-via-trial-error-explain-in-context-learning-paper","Tuning-Free Personalized Alignment via Trial-Error-Explain In-Context Learning - Paper","","Language models aligned to collective preferences tend to produce generic text that fails to reflect individual user style. The work introduces Trial-Error-Explain In-Context Learning (TICL), a tuning-free personalization method requiring fewer than 10 examples per user. TICL iteratively expands an in-context learning prompt by trial-error, adding model-generated negative samples and targeted explanations to guide stylistic alignment. Results on email, essay, and news tasks show up to 91.5% win rates versus prior state of the art, with lexical and qualitative analyses highlighting reduced bias toward formal phrases.",{"@graph":69,"@context":122},[70,84,105],{"@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/tuning-free-personalized-alignment-via-trial-error-explain-in-context-learning-paper/501590/",{"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/tuning-free-personalized-alignment-via-trial-error-explain-in-context-learning-paper/501590.png","ImageObject",300,407,{"name":92,"@type":93},"vwvw","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-17","2026-10-03",true,{"@type":102,"interactionType":103,"userInteractionCount":81},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"What problem does TICL address in language model outputs?","Question",{"text":112,"@type":113},"TICL addresses the tendency of aligned language models to generate generic responses that do not match a specific user’s writing style.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"How does Trial-Error-Explain In-Context Learning personalize a model without tuning?",{"text":117,"@type":113},"TICL expands an in-context learning prompt through trial-error, adding model-generated negative samples and explanations that highlight shortcomings relative to the user’s style.",{"name":119,"@type":110,"acceptedAnswer":120},"What evidence shows TICL improves personalized writing tasks?",{"text":121,"@type":113},"Experiments report higher pairwise win rates up to 91.5% using LLM-as-a-judge comparisons, and ablations show explanations contribute the majority of the gains over a simple ICL baseline.","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},501590,1792117871,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":81,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":139,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":144,"read_time":145},3985747874149,"https://ap-avatar.wpscdn.com/avatar/3a001161165b4dc3d9e?x-image-process=image/resize,m_fixed,w_180,h_180&k=1790129441039064114","Tuning-Free Personalized Alignment via Trial-Error-Explain  \nIn-Context Learning  \nHyundong Cho1∗, Karishma Sharma2 , Nicolaas Jedema2 , Leonardo F. R. Ribeiro2 ,  \nAlessandro Moschitti2 , Ravi Krishnan2 , Jonathan May1  \n1University of Southern California, Information Sciences Institute 2Amazon AGI  \n[hd.justincho@gmail.com](hd.justincho@gmail.com)  \nAbstract  \nLanguage models are aligned to the collective voice of many, resulting in generic outputs that do not align with specific users’ styles. In this work, we present Trial-Error-Explain In-Context Learning (TICL), a tuning-free method that personalizes language models for text generation tasks with fewer than 10 examples per user. TICL iteratively expands an in-context learning prompt via a trial-errorexplain process, adding model-generated negative samples and explanations that provide finegrained guidance towards a specific user’s style. TICL achieves favorable win rates on pairwise comparisons with LLM-as-a-judge up to 91.5% against the previous state-of-the-art and outperforms competitive tuning-free baselines for personalized alignment tasks of writing emails, essays and news articles. Both lexical and qualitative analyses show that the negative samples and explanations enable language models to learn stylistic context more effectively and overcome the bias towards structural and formal phrases observed in their zero-shot outputs. By front-loading inference compute to create a user-specific in-context learning prompt that does not require extra generation steps at test time, TICL presents a novel yet simple approach for personalized alignment.  \n1 Introduction  \nLarge language models (LLMs) are biased towards generic outputs as they are trained to align to an aggregate preference to be generally useful (Padmakumar and He, 2024 ; Rafailov et al., 2024 ; Lee et al., 2023 ; Wang et al., 2023) . Yet end users often have specific needs that call for personalized text generation, such as writing an email (Kumar et al., 2024 ; Salemi et al., 2023) . Hence, adapting LLMs for personalized text generation has drawn a growing interest (Jang et al., 2023 ; Shaikh et al., 2024 ; Mysore et al., 2023 ; Li et al., 2024), but  \n∗ Work was done while HC was an intern at Amazon AGI.  \nFigure 1: Top: The author’s text contains rhetorical questions and more colloquial and informal phrases. Middle: Few-shot in-context learning is insufficient for adapting to an author’s style and model biases for formal phrases persist. Bottom: TICL helps overcome model biases and applies the author’s style more consistently. Model outputs are with gpt-4o-2024-0806  \nprior work on personalized text generation either rely on large amounts of personal data (Li et al., 2024, 2023 ; Mysore et al., 2023) or parameter updates that result in cumbersome per-user 1 model parameters (Shaikh et al., 2024 ; Liu et al., 2023) . These are critical limitations. Personalization is impractical when premised on users providing ample data because most are not willing to exert such effort (Tétard and Collan, 2009) or are concerned  \n1We use the term “user” and “author” interchangeably throughout this work. A user is anyone who provides personally written text that style can be learned from.  \nabout privacy (Plant et al., 2022) . In addition, finetuning may not even be a viable option. The best performing LLMs are most often only accessible via APIs or too large that fine-tuning them is prohibitive even if they are open-sourced (Touvron et al., 2023 ; Firsich and Rios, 2024 ; Anthropic, 2024) . Can we develop a method that overcomes these limitations such that it is effective with small amounts of data and requires no parameter updates?  \nTo address this problem, we propose TrialError-Explain In-Context Learning (TICL), a tuning-free adaptation of Trial-and-Error Finetuning (TEFT) (Shaikh et al., 2024 ; Gulcehre et al., 2023 ; Song et al., 2024) . The key idea of TEFT is to iteratively generate synthetic negative samples to cr","cbCaidM1Rm4atLzQ","https://ap.wps.com/l/cbCaidM1Rm4atLzQ","pdf",1319495,22,"English","# Abstract\n# 1 Introduction\n# 2 Methodology","[{\"question\":\"What problem does TICL address in language model outputs?\",\"answer\":\"TICL addresses the tendency of aligned language models to generate generic responses that do not match a specific user’s writing style.\"},{\"question\":\"How does Trial-Error-Explain In-Context Learning personalize a model without tuning?\",\"answer\":\"TICL expands an in-context learning prompt through trial-error, adding model-generated negative samples and explanations that highlight shortcomings relative to the user’s style.\"},{\"question\":\"What evidence shows TICL improves personalized writing tasks?\",\"answer\":\"Experiments report higher pairwise win rates up to 91.5% using LLM-as-a-judge comparisons, and ablations show explanations contribute the majority of the gains over a simple ICL baseline.\"}]","Tuning-Free Personalized Alignment via Trial-Error-Explain In-Context Learning - Paper | PDF",1791021020,55]