[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-258333-105":3,"detail-sidebar-cat-1-en-105":81,"doc-detail-258333-en":127},{"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","distilling-large-language-models-using-skill-occupation-graph-context-resume-job-description-benchmark-rjdb","Distilling Large Language Models using Skill-Occupation Graph Context - Resume-Job Description Benchmark (RJDB)","","Numerous HR applications rely on resumes and job descriptions, but broad adoption of NLP with large language models remains limited by the lack of comprehensive public benchmarks for varied HR tasks and the absence of smaller models with competitive capability. This paper introduces the Resume-Job Description Benchmark (RJDB) and distills domain-specific knowledge from a large language model using a curated skill-occupation graph as contextual guidance. RJDB supports matching, skill/experience extraction, and resume editing.",{"@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/template/","Template",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/template/paper-templates/","Paper Templates",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/template/distilling-large-language-models-using-skill-occupation-graph-context-resume-job-description-benchmark-rjdb/258333/",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/distilling-large-language-models-using-skill-occupation-graph-context-resume-job-description-benchmark-rjdb/258333.png","ImageObject",442,249,{"name":42,"@type":43},"Grenda","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-14","2026-09-13",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",6,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What problem does RJDB address for HR-related NLP tasks?","Question",{"text":63,"@type":64},"RJDB targets the shortage of comprehensive benchmarks for many resume and job description tasks and the lack of smaller, competitive models trained from strong teachers like GPT-4.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"How is the skill-occupation graph used in generating RJDB data?",{"text":68,"@type":64},"The approach curates a bipartite skill-occupation graph and samples subgraphs to provide skills and occupations as context that guides GPT-4 during generation of resume-job description triples.",{"name":70,"@type":61,"acceptedAnswer":71},"What tasks can models be trained and evaluated on using RJDB?",{"text":72,"@type":64},"RJDB supports multiple HR tasks, including job-candidate matching, extracting skills and experiences from resumes, and editing resumes.","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},258333,1790341285,{"code":4,"msg":82,"data":83},"success",[84,89,94,99,104,109,114,119,123],{"id":85,"doc_module":22,"doc_module_name":25,"category_name":86,"show_sort_weight":87,"slug":88},11,"Presentations",90,"presentations",{"id":90,"doc_module":22,"doc_module_name":25,"category_name":91,"show_sort_weight":92,"slug":93},12,"Resumes",80,"resumes",{"id":95,"doc_module":22,"doc_module_name":25,"category_name":96,"show_sort_weight":97,"slug":98},14,"Invoices",70,"invoices",{"id":100,"doc_module":22,"doc_module_name":25,"category_name":101,"show_sort_weight":102,"slug":103},15,"Posters",60,"posters",{"id":105,"doc_module":22,"doc_module_name":25,"category_name":106,"show_sort_weight":107,"slug":108},16,"Social Media",50,"social-media",{"id":110,"doc_module":22,"doc_module_name":25,"category_name":111,"show_sort_weight":112,"slug":113},17,"Forms",40,"forms",{"id":115,"doc_module":22,"doc_module_name":25,"category_name":116,"show_sort_weight":117,"slug":118},18,"Letters",30,"letters",{"id":120,"doc_module":22,"doc_module_name":25,"category_name":29,"show_sort_weight":121,"slug":122},21,5,"papers-templates",{"id":124,"doc_module":22,"doc_module_name":25,"category_name":125,"show_sort_weight":4,"slug":126},158,"General","general-158",{"code":4,"msg":82,"data":128},{"doc_id":79,"user_id":129,"nickname":42,"user_avatar":130,"doc_module":22,"category_id":120,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":131,"file_id":132,"file_url":133,"file_type":134,"file_size":135,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":95,"language":136,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":137,"faqs":138,"seo_title":139,"seo_description":12,"update_tm":140,"read_time":121},7971474921005,"https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0","Distilling Large Language Models using Skill-Occupation Graph Context  \nfor HR-Related Tasks  \nPouya Pezeshkpour 1 Hayate Iso 1 Thom Lake2  \nNikita Bhutani 1 Estevam Hruschka 1  \n1Megagon Labs 2Indeed  \n{pouya,hayate,nikita,[estevam}@megagon.ai](estevam}@megagon.ai) [tlake@indeed.com](tlake@indeed.com)  \narXiv :2311 .06383v1 [ cs .CL] 10 Nov 2023  \nAbstract  \nNumerous HR applications are centered around resumes and job descriptions. While they can benefit from advancements in NLP, particularly large language models, their real-world adoption faces challenges due to absence of comprehensive benchmarks for various HR tasks, and lack of smaller models with competitive capabilities. In this paper, we aim to bridge this gap by introducing the Resume-Job Description Benchmark (RJDB). We meticulously craft this benchmark to cater to a wide array of HR tasks, including matching and explaining resumes to job descriptions, extracting skills and experiences from resumes, and editing resumes. To create this benchmark, we propose to distill domain-specific knowledge from a large language model (LLM) . We rely on a curated skill-occupation graph to ensure diversity and provide context for LLMs generation. Our benchmark includes over 50 thousand triples of job descriptions, matched resumes and unmatched resumes. Using RJDB, we train multiple smaller student models. Our experiments reveal that the student models achieve near/better performance than the teacher model (GPT-4), affirming the effectiveness of the benchmark.  \nAdditionally, we explore the utility of RJDB on out-of-distribution data for skill extraction and resume-job description matching, in zero-shot and weak supervision manner. We release our datasets and code 1 to foster further research and industry applications.  \n1 Introduction  \nIn organizational recruitment, resumes and job descriptions play a pivotal role, facilitating identification of potential candidates and informing hiring decisions (Zimmermann et al., 2016 ; Guo et al., 2021 ; Ali et al., 2022) . Natural Language Processing (NLP) algorithms play a crucial role in enhancing this process, unraveling valuable information  \n1 [https://github.com/megagonlabs/rjdb](https://github.com/megagonlabs/rjdb)  \nFigure 1: Creating a benchmark for HR-related tasks via large language model distillation. We start by meticulously sampling subgraphs from our curated skilloccupation graph. Then, we utilize the skills and occupations contained within these subgraphs as contextual guidance to steer GPT-4 in the generation of resumesand job descriptions encompassing a variety of tasks.  \nand insights embedded within resumes and job descriptions. This include information such as job titles, skill sets, work history, and educational background. Leveraging such information can facilitate a broad variety of HR tasks such as aligning candidates with job openings, streamlining the resume screening process, and aiding in salary negotiation. Deploying NLP models for such use cases requires a diverse and representative dataset covering a wide range of resumes and job descriptions. However, there is a noticeable absence of publicly available large-scale benchmarks tailored for this purpose. Additionally, many companies heavily rely on proprietary, in-house data to train NLP models. However, such datasets often carry inherent biases and noise, and only provide necessary annotation for limited downstream tasks.  \nIn recent years, several attempts have been made to create/augment datasets by distilling knowledge from existing large language models (LLMs) (Kim et al., 2022 ; Gu et al., 2023 ; Li et al., 2023 ; Gu  \net al., 2023) . The objective is to extract the knowledge stored in a large teacher model such as GPT-4 (OpenAI, 2023), into a benchmark for a specific task. Subsequently, this benchmark becomes a valuable resource for training smaller student models, allowing them to emulate the performance of the teacher model in the targeted task. These st","cbCaivRSPqfiCljq","https://ap.wps.com/l/cbCaivRSPqfiCljq","pdf",352044,"English","# Abstract\n# Introduction\n## Knowledge distillation for HR benchmarks\n## Skill-occupation graph as context","[{\"question\":\"What problem does RJDB address for HR-related NLP tasks?\",\"answer\":\"RJDB targets the shortage of comprehensive benchmarks for many resume and job description tasks and the lack of smaller, competitive models trained from strong teachers like GPT-4.\"},{\"question\":\"How is the skill-occupation graph used in generating RJDB data?\",\"answer\":\"The approach curates a bipartite skill-occupation graph and samples subgraphs to provide skills and occupations as context that guides GPT-4 during generation of resume-job description triples.\"},{\"question\":\"What tasks can models be trained and evaluated on using RJDB?\",\"answer\":\"RJDB supports multiple HR tasks, including job-candidate matching, extracting skills and experiences from resumes, and editing resumes.\"}]","Distilling Large Language Models using Skill-Occupation Graph Context - Resume-Job Description Benchmark (RJDB) | PDF",1789321297]