[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-295543-105":59,"doc-detail-295543-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","passing-the-driving-knowledge-test-driveqa-open-access-iccv-paper","Passing the Driving Knowledge Test - DriveQA - Open-Access ICCV Paper","","The ICCV paper investigates whether current Large Language Models can pass a driving knowledge test that requires full understanding of traffic rules, signage, and right-of-way. It introduces DriveQA, an open-source text-and-vision benchmark covering regulations and traffic scenarios from U.S. driving manuals with synthetic visual variants. Experiments show strong performance on basic rules but weaknesses in numerical reasoning, sign variation, and complex right-of-way. Fine-tuning improves multiple categories, and pretraining boosts downstream driving QA on nuScenes and BDD.",{"@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/passing-the-driving-knowledge-test-driveqa-open-access-iccv-paper/295543/",{"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/passing-the-driving-knowledge-test-driveqa-open-access-iccv-paper/295543.png","ImageObject",300,407,{"name":92,"@type":93},"Đào","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-30","2026-09-18",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",17,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What question does the paper study about LLMs?","Question",{"text":113,"@type":114},"It asks whether a Large Language Model could pass a driving knowledge test today, beyond common spatial/visual QA benchmarks.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"What is DriveQA and what does it cover?",{"text":118,"@type":114},"DriveQA is an extensive open-source text-and-vision benchmark that exhaustively covers traffic regulations and scenarios, including text questions and visual sign/right-of-way tasks.",{"name":120,"@type":111,"acceptedAnswer":121},"How do models perform on DriveQA and what improvements are shown?",{"text":122,"@type":114},"The paper reports good performance on basic traffic rules but notable weaknesses in numerical reasoning and complex right-of-way, while fine-tuning improves accuracy across categories such as sign recognition and intersection decisions.","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},295543,1789720068,{"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":140,"language":141,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":67,"update_tm":130,"read_time":41,"preview_page_count":4},1374402968488,"https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0","This ICCV paper is the Open Access version, provided by the Computer Vision Foundation.  \nExcept for this watermark, it is identical to the accepted version; the final published version of the proceedings is available on IEEE Xplore.  \nPassing the Driving Knowledge Test  \nMaolin Wei 1 * Wanzhou Liu2 * Eshed Ohn-Bar1  \n1Boston University 2Washington University in St. Louis  \nAbstract  \nIf a Large Language Model (LLM) were to take a driving knowledge test today, would it pass? Beyond standard spatial and visual question-answering (QA) tasks on current autonomous driving benchmarks, driving knowledge tests require a complete understanding of all traffic rules, signage, and right-of-way principles. To pass this test, human drivers must discern various edge cases that rarely appear in real-world datasets. In this work, we present DriveQA, an extensive open-source text and vision-based benchmark that exhaustively covers traffic regulations and scenarios. Through our experiments using DriveQA, we show that (1) state-of-the-art LLMs and Multimodal LLMs (MLLMs) perform well on basic traffic rules but exhibit significant weaknesses in numerical reasoning and complex right-of-way scenarios, traffic sign variations, and spatial layouts, (2) fine-tuning on DriveQA improves accuracy across multiple categories, particularly in regulatory sign recognition and intersection decision-making,(3) controlled variations in DriveQA-V provide insights into model sensitivity to environmental factors such as lighting, perspective, distance, and weather conditions, and (4) pretraining on DriveQA enhances downstream driving task performance, leading to improved results on real-world datasets such as nuScenes and BDD, while also demonstrating that models can internalize text and synthetic traffic knowledge to generalize effectively across downstream QA tasks.  \n1. Introduction  \nSafe navigation in traffic requires not only recognizing and interpreting visual information but also reasoning over traffic rules and making decisions that align with regulations.  \nTo ensure drivers develop these critical skills, before receiving their license they must first pass a written knowledge test—a structured (multiple choice questions) assessment designed to evaluate precise understanding of traffic laws, right-of-way rules, and complex driving scenarios [32, 56] . Driving tests are not merely procedural; they assess a driver’s ability to apply reasoning across a wide range  \n*Equally contributed.  \nFigure 1 . Can LLMs Pass a Driving Knowledge Test? We introduce a comprehensive multimodal dataset to evaluate the traffic rule-following capabilities of MLLMs. While most question answering (QA) benchmarks in autonomous driving focus on spatial understanding and common planning tasks, our DriveQA dataset assesses broad driving knowledge. The challenging benchmark comprises text-based questions derived from various U.S. state driving manuals, as well as visual tasks for traffic sign recognition and right-of-way judgment. We evaluate both text-only and imagetext QA using synthetic images (varying perspectives, weather, time of day, and sign types) while showing transferability and generalization to downstream real-world driving tasks. Ths figure shows one example for a right-of-way question, a category where models frequently struggle. Incorrect responses are highlighted in red and correct answers in green.  \nof traffic conditions. While primarily textual, these tests may also include graphical illustrations to ground questions in real-world scenarios. Recent advances in Multimodal Large Language Models (MLLMs) [3, 26, 47, 75, 97] as general-purpose reasoning models provide an opportunity to explore a key question: how well do current vision-andlanguage models perform when faced with the same driving knowledge assessments? Even without targeted fine-tuning, MLLMs may inherit some traffic rule knowledge from their pretraining data (however, our findings indicate that both such","cbCaitywFnKeUuq8","https://ap.wps.com/l/cbCaitywFnKeUuq8","pdf",16659226,12,"English","# Abstract\n# 1. Introduction","[{\"question\":\"What question does the paper study about LLMs?\",\"answer\":\"It asks whether a Large Language Model could pass a driving knowledge test today, beyond common spatial/visual QA benchmarks.\"},{\"question\":\"What is DriveQA and what does it cover?\",\"answer\":\"DriveQA is an extensive open-source text-and-vision benchmark that exhaustively covers traffic regulations and scenarios, including text questions and visual sign/right-of-way tasks.\"},{\"question\":\"How do models perform on DriveQA and what improvements are shown?\",\"answer\":\"The paper reports good performance on basic traffic rules but notable weaknesses in numerical reasoning and complex right-of-way, while fine-tuning improves accuracy across categories such as sign recognition and intersection decisions.\"}]","Passing the Driving Knowledge Test - DriveQA - Open-Access ICCV Paper | PDF"]