[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-155907-105":59,"doc-detail-155907-en":131},{"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":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","probing-gender-bias-in-multilingual-llms-a-case-study-of-stereotypes-in-persian","Probing Gender Bias in Multilingual LLMs - A Case Study of Stereotypes in Persian","","Multilingual Large Language Models (LLMs) are widely deployed, increasing the need to detect and mitigate gender bias that can cause representational harm. This paper addresses a research gap for low-resource languages by presenting a template-based probing methodology validated with real-world data. It introduces the Domain Specific Gender Skew Index (DS GSI) to quantify deviations from gender parity and evaluates four prominent models across four semantic domains in Persian, showing stronger and more rigid stereotypes than in English, especially in sports.",{"@graph":69,"@context":123},[70,84,106],{"@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/probing-gender-bias-in-multilingual-llms-a-case-study-of-stereotypes-in-persian/155907/",{"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/probing-gender-bias-in-multilingual-llms-a-case-study-of-stereotypes-in-persian/155907.png","ImageObject",300,407,{"name":92,"@type":93},"Violet","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-03","2026-08-28",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",15,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What gap does the paper address regarding gender bias research in LLMs?","Question",{"text":113,"@type":114},"Prior work focuses mainly on high-resource languages; low-resource languages remain underexplored. The paper targets Persian, a low-resource language, to evaluate gender bias systematically.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"How does the proposed framework measure gender stereotypes in Persian LLM outputs?",{"text":118,"@type":114},"It uses a template-based probing methodology and introduces the Domain Specific Gender Skew Index (DS GSI) to quantify deviations from gender parity across semantic domains.",{"name":120,"@type":111,"acceptedAnswer":121},"Which models and semantic domains are evaluated, and what do the results show?",{"text":122,"@type":114},"The study evaluates GPT 4o mini, DeepSeek R1, Gemini 2.0 Flash, and Qwen QwQ 32B across academic disciplines, professions, colors, and sports. All models show gender stereotypes in Persian, with larger disparities than in English, and sports display the most rigid biases.","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},155907,1787941017,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":44,"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":130,"read_time":144,"preview_page_count":4},4398048950312,"https://ap-avatar.wpscdn.com/avatar/400002538284de19e3c?_k=1778320343897328908","Probing Gender Bias in Multilingual LLMs: A Case Study of Stereotypes in  \nPersian  \nGhazal Kalhor1 Behnam Bahrak2  \n1 School of Electrical and Computer Engineering,  \nCollege of Engineering, University of Tehran, Tehran, Iran,  \n2 Tehran Institute for Advanced Studies, Tehran, Iran,  \nCorrespondence: [kalhor.ghazal@ut.ac.ir](kalhor.ghazal@ut.ac.ir), [b.bahrak@teias.institute](b.bahrak@teias.institute)  \nAbstract  \nMultilingual Large Language Models (LLMs) are increasingly used worldwide, making it essential to ensure they are free from gender bias to prevent representational harm. While prior studies have examined such biases in high resource languages, low resource languages remain understudied. In this paper, we propose a template based probing methodology, validated against real world data, to uncover gender stereotypes in LLMs. As part of this framework, we introduce the Domain Specific Gender Skew Index (DS GSI), a metric that quantifies deviations from gender parity. We evaluate four prominent models, GPT 4o mini, DeepSeek R1, Gemini 2.0 Flash, and Qwen QwQ 32B, across four semantic domains, focusing on Persian, a low resource language with distinct linguistic features. Our results show that all models exhibit gender stereotypes, with greater disparities in Persian than in English across all domains. Among these, sports reflect the most rigid gender biases. This study underscores the need for inclusive NLP practicesand provides a framework for assessing bias in other low resource languages.  \n1 Introduction  \nLarge Language Models (LLMs) have seen rapid adoption across languages and domains, from everyday use to complex industrial tasks. Ensuring these technologies are fair and unbiased is essential. Gender bias, in particular, can lead to harmful stereotypes and representational harm (Kotek et al., 2023) . Despite advancements in multilingual LLMs, most research focuses on high resource languages, especially English, leaving low resource languages underexplored (Ranjan et al., 2024) .  \nPersian is a low resource language in the multilingual LLM landscape, largely due to the scarcity of structured, diverse training corpora. Most available data come from unstructured sources like social media, and open source resources and NLP  \ntools for Persian are limited. Despite these challenges, Persian offers a unique case for studying gender bias, given linguistic features such as the absence of gendered pronouns, which may affect how bias appears. However, there are currently no standardized benchmarks or tools for evaluating gender bias in LLMs for Persian.  \nTo address this gap, we propose a novel template based probing method to uncover implicit gender biases in multilingual LLMs applied to Persian. Our approach targets four semantic domains, academic disciplines, professions, colors, and sports, chosen to span a spectrum from professional identity to cultural concepts, where stereotypes are well documented in the sociological literature (Archer and Freedman, 1989; Matheus and Quinn, 2017; Cunningham and Macrae, 2011; Liu et al., 2023) . We evaluate four prominent, publicly accessible multilingual LLMs, GPT 4o mini, DeepSeek R1, Gemini 2.0 Flash, and Qwen QwQ 32B, developed by different organizations and representing a diverse range of architectures and training data (OpenAI, 2024; DeepSeek AI et al., 2025; DeepMind, 2025; Team, 2025) . All four models are capable of handling Persian, making them suitable for our evaluation.1  \nThis study investigates the following research questions: RQ1: To what extent do prominent multilingual LLMs exhibit gender bias when prompted in Persian across various semantic domains? RQ2: Are gender biases in LLMs more pronounced or expressed differently in Persian (a low resource language) compared to a high resource language like English?  \nOur results show that LLMs reflect strong gender stereotypes across all four domains in Persian. Generated names for academic fields and professions display c","cbCaisx7vi9hmoQU","https://ap.wps.com/l/cbCaisx7vi9hmoQU","pdf",394201,"English","# Abstract\n# Introduction\n# Related Work","[{\"question\":\"What gap does the paper address regarding gender bias research in LLMs?\",\"answer\":\"Prior work focuses mainly on high-resource languages; low-resource languages remain underexplored. The paper targets Persian, a low-resource language, to evaluate gender bias systematically.\"},{\"question\":\"How does the proposed framework measure gender stereotypes in Persian LLM outputs?\",\"answer\":\"It uses a template-based probing methodology and introduces the Domain Specific Gender Skew Index (DS GSI) to quantify deviations from gender parity across semantic domains.\"},{\"question\":\"Which models and semantic domains are evaluated, and what do the results show?\",\"answer\":\"The study evaluates GPT 4o mini, DeepSeek R1, Gemini 2.0 Flash, and Qwen QwQ 32B across academic disciplines, professions, colors, and sports. All models show gender stereotypes in Persian, with larger disparities than in English, and sports display the most rigid biases.\"}]","Probing Gender Bias in Multilingual LLMs - A Case Study of Stereotypes in Persian | PDF",23]