[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-139169-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-139169-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","hyperphantasia-a-benchmark-for-evaluating-the-mental-visualization-capabilities-of-multimodal-llms","Hyperphantasia - A Benchmark for Evaluating the Mental Visualization Capabilities of Multimodal LLMs","","Mental visualization—the ability to construct and manipulate internal visual representations—is fundamental to human reasoning, prediction, and abstraction. While multimodal large language models (MLLMs) advance quickly, existing benchmarks largely test passive visual perception rather than active, internally generated visual simulation for problem solving. Hyperphantasia is introduced as a synthetic benchmark with four procedurally generated puzzles at three difficulty levels to enable controlled evaluation, revealing a substantial performance gap between humans and MLLMs and highlighting reinforcement learning as a possible path to improved visual simulation.",{"@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/hyperphantasia-a-benchmark-for-evaluating-the-mental-visualization-capabilities-of-multimodal-llms/139169/",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/hyperphantasia-a-benchmark-for-evaluating-the-mental-visualization-capabilities-of-multimodal-llms/139169.png","ImageObject",300,407,{"name":42,"@type":43},"Emma Mercer","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-29","2026-08-23",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",14,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What problem does Hyperphantasia address in multimodal LLM evaluation?","Question",{"text":63,"@type":64},"It targets the gap between current benchmarks that mainly assess passive visual perception and the need to evaluate active mental visualization, i.e., the internal construction and manipulation of visual patterns for reasoning.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"How is the Hyperphantasia benchmark structured?",{"text":68,"@type":64},"It consists of four distinct procedurally generated puzzle types, each provided at three difficulty levels to support controlled analysis across increasing complexity.",{"name":70,"@type":61,"acceptedAnswer":71},"What do the benchmark results suggest about current MLLMs?",{"text":72,"@type":64},"They indicate a substantial gap between human performance and MLLMs, with models showing only partial competence at recognizing visual patterns while robust mental visualization remains an open challenge.","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},139169,1787505540,{"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},962084925502,"https://ap-avatar.wpscdn.com/davatar_6f874abed73319feea01a86fa6f0fab8","Hyperphantasia: A Benchmark for Evaluating the Mental Visualization Capabilities of Multimodal LLMs  \nMohammad Shahab Sepehri∗ Berk Tinaz† Zalan Fabian‡ Mahdi Soltanolkotabi§  \nDepartment of Electrical and Computer Engineering  \nUniversity of Southern California, Los Angeles, CA, USA  \n∗ [sepehri@usc.edu](sepehri@usc.edu) †[tinaz@usc.edu](tinaz@usc.edu)[ ](tinaz@usc.edu)‡[fabian.zalan@gmail.com](fabian.zalan@gmail.com) § [soltanol@usc.edu](soltanol@usc.edu)  \nAbstract  \nMental visualization, the ability to construct and manipulate visual representations internally, is a core component of human cognition and plays a vital role in tasks involving reasoning, prediction, and abstraction. Despite the rapid progress of Multimodal Large Language Models (MLLMs), current benchmarks primarily assess passive visual perception, offering limited insight into the more active capability of internally constructing visual patterns to support problem solving. Yet mental visualization is a critical cognitive skill in humans, supporting abilities such as spatial navigation, predicting physical trajectories, and solving complex visual problems through imaginative simulation. To bridge this gap, we introduce Hyperphantasia, a synthetic benchmark designed to evaluate the mental visualization abilities of MLLMs through four carefully constructed puzzles. Each puzzle is procedurally generated and presented at three difficulty levels, enabling controlled analysis of model performance across increasing complexity. Our comprehensive evaluation of state-of-the-art models reveals a substantial gap between the performance of humans and MLLMs. Additionally, we explore the potential of reinforcement learning to improve visual simulation capabilities. Our findings suggest that while some models exhibit partial competence in recognizing visual patterns, robust mental visualization remains an open challenge for current MLLMs. Our dataset is publicly available at Huggingface 1 , and the evaluation code can be found at GitHub2.  \n1 Introduction  \nThe human capacity for mental visualization – the ability to internally simulate scenes, structures, and dynamics – is central to perception and reasoning. Decades of work in cognitive science have demonstrated that people can mentally rotate objects in three-dimensional space [25], infer the future trajectory of moving bodies from a static snapshot [4], and fill in occluded or missing information in a scene based on prior experience [3] . These abilities reflect an underlying mental model of the physical and spatial world, enabling prediction in situations where direct perceptual input is limited or ambiguous.  \nIn recent years, large language models (LLMs) have demonstrated strong performance on a wide range of linguistic and reasoning tasks, mostly driven by scale and training on massive text corpora. Building on this success, vision-language models (VLMs), which fuse textual and visual representations, have extended these capabilities to multimodal domains. VLMs have achieved impressive results on  \n1[https://huggingface.co/datasets/shahab7899/Hyperphantasia](https://huggingface.co/datasets/shahab7899/Hyperphantasia)  \n2[https://github.com/AIF4S/Hyperphantasia](https://github.com/AIF4S/Hyperphantasia)  \n39th Conference on Neural Information Processing Systems (NeurIPS 2025) Track on Datasets and Benchmarks.  \nInterpolation  \nSeven Segments  \nConnect 3 to 4,   What number do you see?  \nConnect the Dots  \nWhat do you see in this image?  \nExtrapolation  \nTriangle Trajectory  \nWhich point is a node of the triangle?  \nBall Trajectory  \nWhich one is the impact point?  \nPuzzle (image)  \nVisual Thought  \nFigure 1: Overview of puzzles in Hyperphantasia  \nbenchmarks that assess a variety of skills, including visual problem-solving [33], visual Math reasoning [19], and visual question answering (VQA) [9] . These benchmarks have become de facto standards for evaluating VLM performance. However, the majority of these tasks involve","cbCairKZ2SymR4xs","https://ap.wps.com/l/cbCairKZ2SymR4xs","pdf",2143899,28,"English","# Abstract\n# Introduction\n# Background\n# Hyperphantasia Puzzles\n## Interpolation\n## Extrapolation","[{\"question\":\"What problem does Hyperphantasia address in multimodal LLM evaluation?\",\"answer\":\"It targets the gap between current benchmarks that mainly assess passive visual perception and the need to evaluate active mental visualization, i.e., the internal construction and manipulation of visual patterns for reasoning.\"},{\"question\":\"How is the Hyperphantasia benchmark structured?\",\"answer\":\"It consists of four distinct procedurally generated puzzle types, each provided at three difficulty levels to support controlled analysis across increasing complexity.\"},{\"question\":\"What do the benchmark results suggest about current MLLMs?\",\"answer\":\"They indicate a substantial gap between human performance and MLLMs, with models showing only partial competence at recognizing visual patterns while robust mental visualization remains an open challenge.\"}]","Hyperphantasia - A Benchmark for Evaluating the Mental Visualization Capabilities of Multimodal LLMs | PDF",71]