[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-267685-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-267685-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","evaluating-multimodal-generative-ai-with-korean-educational-standards","Evaluating Multimodal Generative AI with Korean Educational Standards","","This paper introduces KoNET, a Korean National Educational Test benchmark for evaluating Multimodal Generative AI systems using four Korean examinations spanning elementary, middle, high, and college-level assessments. KoNET supports rigorous cross-level analysis by measuring question difficulty, subject diversity, and human error rates where available. The study evaluates a spectrum of open-source, open-access, and closed-API models to compare performance under increasingly advanced curricula, and enables insights into model behavior for less-explored languages. Code and dataset builder are released.",{"@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/evaluating-multimodal-generative-ai-with-korean-educational-standards/267685/",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/evaluating-multimodal-generative-ai-with-korean-educational-standards/267685.png","ImageObject",300,407,{"name":42,"@type":43},"Kyle","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-11-11","2026-09-14",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",12,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What is KoNET and how does it evaluate multimodal generative AI?","Question",{"text":63,"@type":64},"KoNET is a benchmark built from four Korean national educational tests. It evaluates multimodal generative AI by examining question difficulty, subject diversity, and (for some items) human error rates.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"Which four Korean exams make up KoNET?",{"text":68,"@type":64},"KoNET includes KoEGED (Korean Elementary General Educational Development Test), KoMGED, KoHGED, and KoCSAT (College Scholastic Ability Test).",{"name":70,"@type":61,"acceptedAnswer":71},"What kinds of models are evaluated in the paper?",{"text":72,"@type":64},"The paper evaluates open-source, open-access, and closed-API multimodal generative AI models using KoNET’s multiple exam levels and analytical views.","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},267685,1789424046,{"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},3985741905716,"https://ap-avatar.wpscdn.com/davatar_994ba38a5ba835b3df7d355c54d3ed8d","Evaluating Multimodal Generative AI with Korean Educational Standards  \nSanghee Park˚  \nNAVER Cloud AI [parksangheeeee@gmail.com](parksangheeeee@gmail.com)  \nGeewook Kim˚:  \nNAVER Cloud AI KAIST AI  \n[gwkim.rsrch@gmail.com](gwkim.rsrch@gmail.com)  \nAbstract  \nThis paper presents the Korean National Educational Test Benchmark (KoNET), a new benchmark designed to evaluate Multimodal Generative AI Systems using Korean national educational tests. KoNET comprises four exams: the Korean Elementary General Educational Development Test (KoEGED), Middle (KoMGED), High (KoHGED), and College Scholastic Ability Test (KoCSAT) . These exams are renowned for their rigorous standards and diverse questions, facilitating a comprehensive analysis of AI performance across different educational levels. By focusing on Korean, KoNET provides insights into model performance in lessexplored languages. We assess a range of models—open-source, open-access, and closed APIs—by examining difficulties, subject diversity, and human error rates. The code and dataset builder will be made fully open-sourced at [https://github.com/naver-ai/KoNET](https://github.com/naver-ai/KoNET).  \n1 Introduction  \nThe advancement of Large Language Models (LLMs) has spurred the integration of sophisticated generative AI systems into various applications (OpenAI, 2023) . Recent developments combining LLMs with computer vision have resulted in powerful Multimodal LLMs (MLLMs) (Liu et al., 2023, 2024b ; Laurençon et al., 2024b,a) . However, questions remain about the true intelligence of these systems, especially their ability to generalize across novel tasks similar to human cognition.  \nCurrent benchmarks predominantly focus on English, overlooking the linguistic diversity worldwide and offering limited insights into low-resource languages like Korean. Moreover, many benchmarks do not compare AI performance to that of  \n˚ Sanghee Park and Geewook Kim contributed equally to this work and share first authorship.  \n: Corresponding author.  \n(A) Examples of QA types in KoNET  \n(B) Average accuracy(%) of Top 30 models by subset of KoNET  \nFigure 1: Examples and Performance Overview of KoNET. (a) Illustration of mathematics problem examples, highlighting the increased complexity and difficulty as the educational level progresses. (b) Demonstration of how the accuracy of contemporary AI models decreases with more advanced curricula. A detailed analysis is provided in Section 4.  \nhumans, making it difficult to precisely measure AI proficiency. Some benchmarks are also less connected to real-world application scenarios, hindering the applicability of MLLMs.  \nTo address these challenges, we introduce KoNET, a benchmark dataset leveraging four key Korean educational tests (refer to Figure 1) . Each  \n671  \nProceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies  \n(Volume 2: Short Papers), pages 671–688  \nApril 29-May 4, 2025 ©2025 Association for Computational Linguistics  \n\n| Statistic | KoEGED | KoMGED | KoHGED | KoCSAT |\n| --- | --- | --- | --- | --- |\n| Images | 400 | 540 | 540 | 897 |\n| Questions | 400 | 540 | 540 | 897 |\n| ˚ K-QA | 62 (15.5%) | 65 (12.0%) | 62 (11.5%) | 57 (6.4%) |\n| : TC-QA | 123 (30.8%) | 249 (46.1%) | 284 (52.6%) | 388 (43.3%) |\n| ; MC-QA | 215 (53.8%) | 226 (41.9%) | 194 (35.9%) | 452 (50.3%) |\n| Subjects | 10 | 11 | 11 | 41 |\n| Choices | 4 (100.0%) | 4 (100.0%) | 4 (100.0%) | 5 (98.8%) |\n| Avg word | 29.9 | 42.7 | 48.0 | 113.0 |\n| Max word | 106 | 362 | 410 | 786 |\n| Avg Char | 113.0 | 167.2 | 193.6 | 475.9 |\n| Max Char | 417 | 1,408 | 1,678 | 3,300 |\n| \\#choice | 4 | 4 | 4 | 5 |\n\nTable 1: Key statistics of the KoNET benchmark. ˚ KQA: Knowledge QA, :TC-QA: Text Comprehension QA, and ;MC-QA: Mutimodal Comprehension QA.  \n\n| Bench | Lang. | \\#Q | \\#I | \\#choice | ˚ D | : H |\n| --- | --- | --- | --- | --- | --- | --- |\n| AI2D | En | 3,088 | 3,088 | “ 4 (100 .0%) | ✗ |","cbCaieHzPZM69Mhy","https://ap.wps.com/l/cbCaieHzPZM69Mhy","pdf",1853190,18,"English","# 1 Introduction\n# 2 Related Work","[{\"question\":\"What is KoNET and how does it evaluate multimodal generative AI?\",\"answer\":\"KoNET is a benchmark built from four Korean national educational tests. It evaluates multimodal generative AI by examining question difficulty, subject diversity, and (for some items) human error rates.\"},{\"question\":\"Which four Korean exams make up KoNET?\",\"answer\":\"KoNET includes KoEGED (Korean Elementary General Educational Development Test), KoMGED, KoHGED, and KoCSAT (College Scholastic Ability Test).\"},{\"question\":\"What kinds of models are evaluated in the paper?\",\"answer\":\"The paper evaluates open-source, open-access, and closed-API multimodal generative AI models using KoNET’s multiple exam levels and analytical views.\"}]","Evaluating Multimodal Generative AI with Korean Educational Standards | PDF",45]