[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-289683-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-289683-en":131},{"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","were-different-were-the-same-creative-homogeneity-across-llms","We’re Different, We’re the Same - Creative Homogeneity Across LLMs","","Large language models are widely used as creative partners, yet prior studies show that LLM-assisted creation can produce a narrower range of outputs. The research question addressed is whether this homogeneity arises from using a specific LLM type or from LLMs in general. The work elicits creative responses from humans and a broad set of LLMs using standardized creativity tests and compares population-level diversity. Results show LLM responses are far more similar across models than humans are to each other, implying model-agnostic convergence.",{"@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/document/","Document",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/document/research-report/","Research & Report",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/were-different-were-the-same-creative-homogeneity-across-llms/289683/",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/were-different-were-the-same-creative-homogeneity-across-llms/289683.png","ImageObject",300,407,{"name":42,"@type":43},"Đào","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-18","2026-09-17",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",22,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What problem does the paper investigate about LLM creativity?","Question",{"text":63,"@type":64},"The paper investigates whether LLMs used as creative partners produce narrowed, homogeneous outputs, and whether that effect depends on using a particular LLM or occurs across LLMs generally.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"How do the authors test their question?",{"text":68,"@type":64},"They solicit creative responses from humans and many LLMs using standardized creativity tests and measure population-level variability of responses.",{"name":70,"@type":61,"acceptedAnswer":71},"What is the main finding about response diversity across models?",{"text":72,"@type":64},"LLM responses are much more similar to other LLM responses than human responses are to each other, even after controlling for response structure and other variables.","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},289683,1789636312,{"code":4,"msg":82,"data":83},"success",[84,88,92,96,101,106,111,115,120,123,127],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":85,"show_sort_weight":86,"slug":87},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":89,"show_sort_weight":90,"slug":91},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":93,"show_sort_weight":94,"slug":95},"Exam",70,"exam",{"id":97,"doc_module":4,"doc_module_name":25,"category_name":98,"show_sort_weight":99,"slug":100},5,"Comic",60,"comic",{"id":102,"doc_module":4,"doc_module_name":25,"category_name":103,"show_sort_weight":104,"slug":105},6,"Technology",50,"technology",{"id":107,"doc_module":4,"doc_module_name":25,"category_name":108,"show_sort_weight":109,"slug":110},7,"Healthcare",40,"healthcare",{"id":112,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":113,"slug":114},8,30,"research-report",{"id":116,"doc_module":4,"doc_module_name":25,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":25,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":25,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":25,"category_name":129,"show_sort_weight":97,"slug":130},19,"General","general",{"code":4,"msg":82,"data":132},{"doc_id":79,"user_id":133,"nickname":42,"user_avatar":134,"doc_module":4,"category_id":112,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":140,"language":141,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":12,"update_tm":80,"read_time":145,"preview_page_count":4},1374402968488,"https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0","arXiv :2501 . 19361v1 [ cs .CY] 31 Jan 2025  \nWe’re Different, We’re the Same: Creative Homogeneity Across LLMs  \nEMILY WENGER∗ , Duke University  \nYOED KENETT, Technion-Israel Institute of Technology  \nNumerous powerful large language models (LLMs) are now available for use as writing support tools, idea generators, and beyond. Although these LLMs are marketed as helpful creative assistants, several works have shown that using an LLM as a creative partner results in a narrower set of creative outputs. However, these studies only consider the effects of interacting with a single LLM, begging the question of whether such narrowed creativity stems from using a particular LLM—which arguably has a limited range of outputs—or from using LLMs in general as creative assistants. To study this question, we elicit creative responses from humans anda broad set ofLLMs using standardized creativity tests and compare the population-level diversity of responses. We find that LLM responses are much more similar to other LLM responses than human responses are to each other, even after controlling for response structure and other key variables. This finding of significant homogeneity in creative outputs across the LLMs we evaluate adds anew dimension to the ongoing conversation about creativity and LLMs. If today’s LLMs behave similarly, using them as a creative partners—regardless of the model used—may drive all users towards a limited set of “creative” outputs.  \n1 INTRODUCTION  \nLarge language models (LLMs) have moved out of research labs and into our everyday lives. Given their advanced abilities to generate text and respond to prompts, LLMs are often marketed as creativity support tools that allow users to write drafts, edit documents, and generate novel ideas with ease [2, 4, 20, 21] . Consumers have responded eagerly to these suggestions. According to a 2024 survey by Adobe, over half of Americans have used generative AI tools like LLMsas creative partners for brainstorming, drafting written content, creating images, or writing code. An overwhelming majority of LLM users surveyed believe these models will help them be more creative [39] .  \nWhile appealing, outsourcing our creative thinking to LLMs could have unintended consequences and demands further scrutiny. For example, recent work has unearthed complications around the use of LLMs as creativity support tools. Researchers found that LLM-aided creative outputs look individually creative but are often quite similar to other LLM-aided outputs. Such “homogeneity” in LLM-aided creative outputs has been observed in a variety of settings, from creative writing to online survey responses to research idea generation and beyond [7, 16, 37, 43, 53] .  \nWhile concerning, these works typically only look at a single LLM and it’s effect on downstream creative content. In a prototypical example, Doshi and Hauser [16] compared the individual and collective creativity of two groups of writers—humans alone and humans aided by ChatGPT—and found that stories produced by the ChatGPT-aided group were more homogeneous. Related work from Moon, Green, and Kushlev [37] compared college essays written by humans and GPT models and found that LLM-authored essays contributed fewer new ideas and were more homogeneous than human-authored essays. However, such work begs the question: does the observed homogeneity occur because only a single type of LLM (GPT variants) is studied? It could be reasonably argued that a single LLM must have a limited range of outputs, causing the homogeneity. Perhaps if writers all used different LLMs, creativity would be restored.  \nRecent work studying feature space alignment in LLMs suggests otherwise. There is a long line of work measuring feature space similarity in machine learning models, since this is believed to indicate overall model similarity [8, 31, 33, 34, 46] . Some initial work has applied these techniques to large-scale LLMs and found evidence of “feature universality”in","cbCaieJX1MLHRl9R","https://ap.wps.com/l/cbCaieJX1MLHRl9R","pdf",1351561,18,"English","# Introduction\n## Background: LLMs as creativity support tools\n## Prior evidence of LLM output homogeneity\n## Open question: single-LLM effects vs LLMs in general\n## Hypothesis from feature-space alignment\n## Consequences of cross-LLM homogeneity\n## Contributions and testing methodology","[{\"question\":\"What problem does the paper investigate about LLM creativity?\",\"answer\":\"The paper investigates whether LLMs used as creative partners produce narrowed, homogeneous outputs, and whether that effect depends on using a particular LLM or occurs across LLMs generally.\"},{\"question\":\"How do the authors test their question?\",\"answer\":\"They solicit creative responses from humans and many LLMs using standardized creativity tests and measure population-level variability of responses.\"},{\"question\":\"What is the main finding about response diversity across models?\",\"answer\":\"LLM responses are much more similar to other LLM responses than human responses are to each other, even after controlling for response structure and other variables.\"}]","We’re Different, We’re the Same - Creative Homogeneity Across LLMs | PDF",45]