[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-1-en-105":3,"doc-seo-263040-105":53,"doc-detail-263040-en":126},{"code":4,"msg":5,"data":6},0,"success",[7,14,19,24,29,34,39,44,49],{"id":8,"doc_module":9,"doc_module_name":10,"category_name":11,"show_sort_weight":12,"slug":13},11,1,"Template","Presentations",90,"presentations",{"id":15,"doc_module":9,"doc_module_name":10,"category_name":16,"show_sort_weight":17,"slug":18},12,"Resumes",80,"resumes",{"id":20,"doc_module":9,"doc_module_name":10,"category_name":21,"show_sort_weight":22,"slug":23},14,"Invoices",70,"invoices",{"id":25,"doc_module":9,"doc_module_name":10,"category_name":26,"show_sort_weight":27,"slug":28},15,"Posters",60,"posters",{"id":30,"doc_module":9,"doc_module_name":10,"category_name":31,"show_sort_weight":32,"slug":33},16,"Social Media",50,"social-media",{"id":35,"doc_module":9,"doc_module_name":10,"category_name":36,"show_sort_weight":37,"slug":38},17,"Forms",40,"forms",{"id":40,"doc_module":9,"doc_module_name":10,"category_name":41,"show_sort_weight":42,"slug":43},18,"Letters",30,"letters",{"id":45,"doc_module":9,"doc_module_name":10,"category_name":46,"show_sort_weight":47,"slug":48},21,"Paper Templates",5,"papers-templates",{"id":50,"doc_module":9,"doc_module_name":10,"category_name":51,"show_sort_weight":4,"slug":52},158,"General","general-158",{"code":4,"msg":54,"data":55},"ok",{"site_id":56,"language":57,"slug":58,"title":59,"keywords":60,"description":61,"schema_data":62,"social_meta":119,"head_meta":121,"extra_data":123,"updated_unix":125},105,"en","human-and-llm-based-resume-matching-an-observational-study","Human and LLM-Based Resume Matching - An Observational Study","","Resume matching evaluates how well job applicants meet requirements using information contained in their resumes, and it increasingly relies on natural language processing to parse and score candidates efficiently. This observational study examines large language models for resume matching by comparing zero-shot GPT-4 ratings with human judgments on 736 real-world resumes across diverse fields. It also analyzes how prompt engineering, including chain-of-thought, affects GPT-4 scoring, and whether group differences related to race/ethnicity and gender differ between models and humans. Results show only minor correlation with human scores, improved rating quality via prompt engineering, and no larger demographic disparities for GPT-4 than for human raters.",{"@graph":63,"@context":118},[64,80,101],{"@type":65,"itemListElement":66},"BreadcrumbList",[67,71,74,77],{"item":68,"name":69,"@type":70,"position":9},"https://docshare.wps.com","Home","ListItem",{"item":72,"name":10,"@type":70,"position":73},"https://docshare.wps.com/template/",2,{"item":75,"name":16,"@type":70,"position":76},"https://docshare.wps.com/template/resumes/",3,{"item":78,"name":59,"@type":70,"position":79},"https://docshare.wps.com/template/human-and-llm-based-resume-matching-an-observational-study/263040/",4,{"url":78,"name":59,"@type":81,"image":82,"author":87,"headline":59,"publisher":90,"fileFormat":93,"inLanguage":57,"description":61,"dateModified":94,"datePublished":95,"encodingFormat":93,"isAccessibleForFree":96,"interactionStatistic":97},"DigitalDocument",{"url":83,"@type":84,"width":85,"height":86},"https://docshare.wps.com/thumbnails/human-and-llm-based-resume-matching-an-observational-study/263040.png","ImageObject",442,249,{"name":88,"@type":89},"Clementine","Person",{"url":68,"name":91,"@type":92},"DocShare","Organization","application/pdf","2026-09-20","2026-09-14",true,{"@type":98,"interactionType":99,"userInteractionCount":9},"InteractionCounter",{"@type":100},"ViewAction",{"@type":102,"mainEntity":103},"FAQPage",[104,110,114],{"name":105,"@type":106,"acceptedAnswer":107},"What is the main goal of the observational study on resume matching?","Question",{"text":108,"@type":109},"The study evaluates how zero-shot GPT-4 performs for resume matching compared with human ratings, using real-world resumes and job postings.","Answer",{"name":111,"@type":106,"acceptedAnswer":112},"How do GPT-4 ratings compare with human ratings overall?",{"text":113,"@type":109},"LLM scores show only minor correlation with humans, indicating they are not interchangeable.",{"name":115,"@type":106,"acceptedAnswer":116},"What impact does prompt engineering have on GPT-4 resume ratings?",{"text":117,"@type":109},"Prompt engineering techniques such as chain-of-thought improve the quality of GPT-4 ratings.","https://schema.org",{"og:url":78,"og:type":120,"og:title":59,"og:site_name":91,"og:description":61},"article",{"robots":122,"canonical":78},"index,follow",{"doc_id":124,"site_id":56},263040,1789366579,{"code":4,"msg":5,"data":127},{"doc_id":124,"user_id":128,"nickname":88,"user_avatar":129,"doc_module":9,"category_id":15,"category_name":16,"doc_title":59,"doc_description":61,"doc_content":130,"file_id":131,"file_url":132,"file_type":133,"file_size":134,"view_count":9,"is_deleted":4,"is_public":9,"is_downloadable":9,"audit_status":9,"page_count":30,"language":135,"language_code":57,"site_id":56,"html_lang":57,"table_of_contents":136,"faqs":137,"seo_title":138,"seo_description":61,"update_tm":125,"read_time":139},1374391974564,"https://ap-avatar.wpscdn.com/avatar/14000253aa45c000a9e?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779874745381141002","Human and LLM-Based Resume Matching: An Observational Study  \nSwanand Vaishampayan  \nVirginia Tech, USA [swanandsv@vt.edu](swanandsv@vt.edu)  \nHunter Leary  \nVirginia Tech, USA [hunterl22@vt.edu](hunterl22@vt.edu)  \nYoseph Berhanu Alebachew  \nVirginia Tech, USA  \n[yoseph@vt.edu](yoseph@vt.edu)  \nLouis Hickman  \nVirginia Tech, USA [louishickman@vt.edu](louishickman@vt.edu)  \nBrent Stevenor  \nNREMT, USA [brentstevenor@gmail.com](brentstevenor@gmail.com)  \nWeston Beck  \nDiscoveredATS, USA [weston@thehiretalent.com](weston@thehiretalent.com)  \nChris Brown  \nVirginia Tech, USA [dcbrown@vt.edu](dcbrown@vt.edu)  \nAbstract  \nResume matching assesses the extent to which candidates qualify for jobs based on the content of resumes. This process increasingly uses natural language processing (NLP) techniques to automate parsing and rating tasks—saving time and effort. Large language models (LLMs) are increasingly used for this purpose—thus, we explore their capabilities for resume matching inan observational study. We compare zero-shot GPT-4 and human ratings for 736 resumes submitted to job openings from diverse fields using real-world evaluation criteria. We also study the effects of prompt engineering techniques on GPT-4 ratings and compare differences in GPT- 4 and human ratings across racial and gender groups. Our results show: LLM scores correlate minorly with humans, suggesting they are not interchangeable; prompt engineering such as CoT improves the quality of LLM ratings;  \nand LLM scores do not show larger group differences (i.e., bias) than humans. Our findings provide implications for LLM-based resume rating to promote more fair NLP-based resume matching in a multicultural world.  \n1 Introduction  \nResume matching is the first step in multi-stage hiring pipelines, where recruiters rate resumes based on the extent to which applicants’ resume content matches job requirements to find ideal candidates (Li et al., 2020) . This rating involves assessing criteria such as work experience, skills, education, certifications and extracurricular activities (Tsai et al., 2011) . However, resume matching is challenging, involving rating hundreds of resumes per job opening (Torres, 2) . Thus, manual resume matching processes are laborious and time-consuming for recruiters. Modern hiring pipelines employ natural language processing (NLP)-powered support to automate resume matching, saving time and effort for recruiters in resume matching processes (Mujtaba and Mahapatra, 2019) .  \nRecent advancements of NLP-based Large Language Models (LLMs), such as OpenAI’s GPT-4, Meta’s Llama-3 and Google Gemini, have seen rapid adoption (Chen et al., 2024a) and revolutionized numerous domains such as software engineering (White et al., 2023), linguistics (Diandaru et al., 2024) and e-commerce (Roumeliotis et al., 2024) . Recent studies (Gan et al., 2024 ; Gaebler et al., 2024) evaluate LLMs for resume parsing and rating, reporting positive results. This motivates the use of LLMs for resume matching, with perceived benefits of easier, faster and efficient resume rating.  \nHowever, LLM evaluations can incorporate challenges, such as discrimination against candidates (Armstrong et al., 2024) . To safeguard against biased LLM-based resume ratings and maximizing perceived benefits, it is imperative to comprehend the differences between rationales used by humans and LLMs. Prior work fails to incorporate the intricacies of authentic resume matching processes in practice, relying on experimental approaches with tight control over factors that can influence LLM-generated scores (i.e., resumesand criteria) . It is important to use real-world resume rating constructs that contribute to hiring decisions (Stepanova et al., 2021 ; Tsai et al., 2011) .  \nAdditionally, to the best of our knowledge, no previous studies discern “how” resume ratings are computed by LLMs. We aim to fill this gap by exploring the capabilities of LLMs in resume matching. We compare ratings gene","cbCaisnfcKL13YJg","https://ap.wps.com/l/cbCaisnfcKL13YJg","pdf",572400,"English","# Abstract\n# 1 Introduction","[{\"question\":\"What is the main goal of the observational study on resume matching?\",\"answer\":\"The study evaluates how zero-shot GPT-4 performs for resume matching compared with human ratings, using real-world resumes and job postings.\"},{\"question\":\"How do GPT-4 ratings compare with human ratings overall?\",\"answer\":\"LLM scores show only minor correlation with humans, indicating they are not interchangeable.\"},{\"question\":\"What impact does prompt engineering have on GPT-4 resume ratings?\",\"answer\":\"Prompt engineering techniques such as chain-of-thought improve the quality of GPT-4 ratings.\"}]","Human and LLM-Based Resume Matching - An Observational Study | PDF",6]