[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84615-en":3,"doc-seo-84615-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},84615,1649267921044,"Ava Thompson","https://us-avatar.wpscdn.com/avatar/1800007509477c92dfb?_k=1782875107921204101",8,"Research & Report","Emergence of Preferential Attachment and Glass Ceiling Effects in Autonomous Networks of LLMs","Investigates structural disparities in communication networks formed by autonomous large language model (LLM) agents. Each agent selects collaborators based on a prompted LLM type defined by its base model, size, and system prompt. When agents choose partners, the resulting directed weighted graph exhibits preferential attachment, where already-prominent types gain more connections. In some settings, weaker agents can become centrally influential, interpreted as a type-dependent glass-ceiling effect. A mean-field model with contraction arguments proves convergence of type centrality to unique stable equilibria and, with a cross-attention-inspired utility, yields predictive limiting structures.","arXiv :2607 .0 1 148v2 [ cs . SI] 6 Jul 2026  \nEmergence of Preferential Attachment and Glass-Ceiling Effects in Autonomous Networks of LLMs ∗  \nYiming Zhang [yz2926@cornell. edu](yz2926@cornell. edu)  \nDepartment of Electrical and Computer Engineering Cornell University  \nVikram Krishnamurthy [vk342@cornell. edu](vk342@cornell. edu)  \nDepartment of Electrical and Computer Engineering Cornell University  \nAbstract  \nWe investigate the emergence of structural disparities in networks comprising large language model (LLM) agents. Each LLM agent refers to a prompted LLM of a specified type determined by its base model, model size, and system prompt. When LLM agents autonomously choose collaborators, the resulting communication network exhibits preferential-attachment dynamics: agent types that are already prominent become increasingly likely to attract additional connections. In some cases, weaker LLM agents (agents with smaller base model or older version) can disproportionately occupy central and influential network positions relative to stronger LLM agents. We interpret this misalignment between task capability and network prominence as a type-dependent glass-ceiling effect (GCE) .  \nWe model the network of LLM agents as a time-evolving sequence of directed weighted graphs, where the vector-valued edge weights represent cumulative tokens exchanged, number of interaction rounds, and reasoning effort. Using a contraction mapping argument on the mean-field dynamics, we prove that the importance (centrality) of each agent type converges to a unique stable equilibrium. To anchor the model in LLM decision mechanisms, we introduce a cross-attention-inspired utility for collaborator selection. This utility specifies the local connection dynamics and, together with the mean-field model, yields a predictive characterization of the limiting network structure and its type-dependent centrality gaps.  \nTo validate the theory, we develop an experimental testbed with 100 LLM agents. Our experiments show that autonomous network formation can generate persistent centrality disparities, with their magnitude and direction depending on model family, model size, system-prompt design, and task context. They further show that the effect of preferential attachment depends on its alignment with model capability: reinforcing it improves collective performance when stronger agents become central, whereas weakening it improves performance when network dynamics instead favor weaker agents. All results are reproducible; the code and datasets are available in an anonymous GitHub repository.  \n1 Introduction  \nMulti-agent LLM networks, where each node is an LLM agent 1 are becoming increasingly important as large language models are deployed not only as isolated assistants, but also as interacting agents capable of collaboration, specialization, and collective problem solving. Such networks of LLM agents have been  \n∗ This work was supported by National Science Foundation grant CCF-2312198 and Army Research Office grant W911NF- 24-1-0083.  \n1 Throughout this paper, an LLM agent refers to a prompted LLM of a specified type. An agent’s type is determined by its base model (e.g., Gemini or GPT), model size (number of parameters), and system prompt defining its role (e.g., answer provider or answer checker) .  \nexplored in a range of settings, including complex problem solving (Qian et al., 2025), software development (Qian et al., 2024; Hong et al., 2024), automated debate and collective judgment (Li et al., 2024; Hu et al. , 2026), and large-scale social simulation (Piao et al., 2026; Guan et al., 2025) . As LLMs continue to improve through scaling and instruction tuning, they are increasingly studied not only as problem-solving tools, but also as agents that can exhibit communicative behaviors, preferences, biases, and social interactions (Park et al., 2023; Ashery et al., 2025; Madmoun & Lahlou, 2026) . Recent work examines multi-agent LLM systems from a socia","cbCaiirlnH7JsAD5","https://ap.wps.com/l/cbCaiirlnH7JsAD5","pdf",1905630,1,29,"English","en",105,"# Abstract\n# Introduction\n## Multi-agent LLM networks and motivation\n## Preferential attachment and two dominance modes","[{\"question\":\"What causes preferential-attachment dynamics in autonomous LLM agent networks?\",\"answer\":\"As agents autonomously choose collaborators, types that are already prominent become increasingly likely to attract additional connections, producing preferential-attachment dynamics.\"},{\"question\":\"What is the glass-ceiling effect (GCE) in this work?\",\"answer\":\"A type-dependent misalignment where weaker LLM agents can disproportionately occupy central and influential network positions relative to stronger agents, interpreted as a glass-ceiling effect.\"},{\"question\":\"How is the limiting network structure and type centrality characterized theoretically?\",\"answer\":\"The work models the evolving network as directed weighted graphs, uses mean-field dynamics with a contraction-mapping argument to prove convergence of type centrality to a unique stable equilibrium, and introduces a cross-attention-inspired utility to connect collaborator selection to the local connection dynamics.\"}]",1784197138,73,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"emergence-of-preferential-attachment-and-glass-ceiling-effects-in-autonomous-networks-of-llms","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/emergence-of-preferential-attachment-and-glass-ceiling-effects-in-autonomous-networks-of-llms/84615/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What causes preferential-attachment dynamics in autonomous LLM agent networks?","Question",{"text":75,"@type":76},"As agents autonomously choose collaborators, types that are already prominent become increasingly likely to attract additional connections, producing preferential-attachment dynamics.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is the glass-ceiling effect (GCE) in this work?",{"text":80,"@type":76},"A type-dependent misalignment where weaker LLM agents can disproportionately occupy central and influential network positions relative to stronger agents, interpreted as a glass-ceiling effect.",{"name":82,"@type":73,"acceptedAnswer":83},"How is the limiting network structure and type centrality characterized theoretically?",{"text":84,"@type":76},"The work models the evolving network as directed weighted graphs, uses mean-field dynamics with a contraction-mapping argument to prove convergence of type centrality to a unique stable equilibrium, and introduces a cross-attention-inspired utility to connect collaborator selection to the local connection dynamics.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]