[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84181-en":3,"doc-seo-84181-105":29,"detail-sidebar-cat-0-en-105":90},{"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":4,"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},84181,1374391974468,"Eden","https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0",8,"Research & Report","K-Risk Knowledge-Augmented Dataset for High-Risk Driving Scenarios","Safe autonomous driving requires rapid reactions to common high-risk events and deeper reasoning for rare long-tail traffic safety scenarios. Such scenarios are under-represented in naturalistic driving data, while existing trajectory and language-augmented datasets rarely include high-risk event labels, semantic annotations, and verifiable safety signals. K-Risk combines structured trajectories with LLM-generated semantic annotations, integrating 20 driving datasets and curating 31,398 high-risk events, including 1,036 extreme near-collision cases, released as trajectory–metadata–language triplets with validated risk analyses and action recommendations.","arXiv :2607 .07 103v 1 [ cs .LG] 8 Jul 2026  \nA knowledge-augmented dataset of high-risk driving scenarios with LLM annotations for autonomous driving  \nHeye Huang 1 , Jingguang Li 1 , Zhiyuan Zhou2,* , Paul Liang3 , Mingyu Wu4 , Kitae Jang 1 , and Jianqiang Wang5,*  \n1 Cho Chun Shik Graduate School of Mobility, Korea Advanced Institute of Science and Technology, Daejeon 34051, South Korea 2 College of Computer Science and Artificial Intelligence, Fudan University, Shanghai 200433, China  \n3 Media Lab, Massachusetts Institute of Technology, Cambridge, MA 02139, USA  \n4 School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China  \n5 School of Vehicle and Mobility, Tsinghua University, Beijing 100084, China  \n* Correspondence ([zhouzhiyuan@pjlab.org.cn](zhouzhiyuan@pjlab.org.cn), [wjqlws@tsinghua.edu.cn](wjqlws@tsinghua.edu.cn)) .  \nABSTRACT  \nSafe autonomous driving requires both a rapid response to common high-risk events and deeper reasoning over the rare, extreme long-tail scenarios of traffic safety. These scenarios are severely under-represented in naturalistic driving data, and existing trajectory and language-augmented datasets seldom provide high-risk event labels, semantic annotations and verifiable safety signals. Here we present K-Risk, a knowledge-augmented dataset that combines structured driving trajectories with large language model (LLM)-generated semantic annotations for safety-critical driving scenarios. K-Risk integrates 20 human-driven and autonomous-vehicle trajectory datasets from Europe, China, and the United States, covering highways, urban freeways, intersections and roundabouts. Using a unified risk-centric extraction pipeline, K-Risk curates 31,398 highrisk events, together with a 1,036-event extreme subset of near-collision cases. Each event is released as a synchronized trajectory–metadata–language triplet containing structured scenario descriptions, abnormal-behavior notifications, and, fora representative subset, LLM-generated causal risk analyses and action recommendations validated through a closed-loop simulator with iterative reflection. By combining multi-dimensional risk annotations, interpretable language supervision, and verifiable decisions, K-Risk bridges structured traffic trajectories, semantic reasoning and verifiable decision supervision, providing a standardized foundation for developing and evaluating next-generation risk-aware autonomous driving agents.  \nBackground & Summary  \nAutonomous driving has progressed rapidly from partial to conditional automation, yet safety in rare and complex long-tail scenarios remains the central barrier to large-scale deployment and on-road testing 1. A safe driving agent must combine two capabilities that draw on different kinds of experience: a fast and reliable response to the common high-risk events encountered every day, and a slower, deeper form of reasoning over the rare and extreme situations that sit in the long tail of traffic safety. The second capability is harder to acquire, because the corresponding scenarios occur infrequently and are therefore severely under-represented in the naturalistic driving data used to train and evaluate driving models. Reliable behavior in these high-risk edge cases is decisive not only for technical performance but also for public trust and regulatory approval2. Agents based on large language models (LLMs) have recently emerged as a promising route to this second capability. By combining structured driving knowledge, learned experience and semantic understanding, they offer a interpretable reasoning process that interprets a scenario, reflects on prior cases and generalizes across situations. They can also be coupled with physics-based models to yield systems that are at once safer and more interpretable?, 4. Realizing this potential, however, depends on training and evaluation data that pair real high-risk trajectories with the semantic, interpretable and verifiable signals the","cbCaiiAaL4S1KoDR","https://ap.wps.com/l/cbCaiiAaL4S1KoDR","pdf",3467414,1,22,"English","en",105,"# Background & Summary\n## Motivation: Long-tail high-risk safety\n## Limits of existing trajectory datasets\n## Limits of current interaction and semantic datasets\n## K-Risk: knowledge-augmented risk annotations","[{\"question\":\"What problem does K-Risk address in autonomous driving safety data?\",\"answer\":\"K-Risk addresses the lack of high-risk event labels, semantic annotations, and verifiable safety signals for rare long-tail traffic scenarios, which are under-represented in naturalistic datasets.\"},{\"question\":\"How does K-Risk combine trajectories with LLM annotations?\",\"answer\":\"Each high-risk event is released as a synchronized trajectory–metadata–language triplet, including structured scenario descriptions, abnormal-behavior notifications, and LLM-generated causal risk analyses and action recommendations for a representative subset.\"},{\"question\":\"What coverage and event scale does K-Risk provide?\",\"answer\":\"K-Risk integrates 20 human-driven and autonomous-vehicle trajectory datasets across Europe, China, and the United States, curating 31,398 high-risk events plus an extreme subset of 1,036 near-collision cases.\"}]",1784193708,55,{"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":85,"head_meta":87,"extra_data":89,"updated_unix":27},"k-risk-knowledge-augmented-dataset-for-high-risk-driving-scenarios","",{"@graph":35,"@context":84},[36,53,67],{"@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/k-risk-knowledge-augmented-dataset-for-high-risk-driving-scenarios/84181/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What problem does K-Risk address in autonomous driving safety data?","Question",{"text":74,"@type":75},"K-Risk addresses the lack of high-risk event labels, semantic annotations, and verifiable safety signals for rare long-tail traffic scenarios, which are under-represented in naturalistic datasets.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How does K-Risk combine trajectories with LLM annotations?",{"text":79,"@type":75},"Each high-risk event is released as a synchronized trajectory–metadata–language triplet, including structured scenario descriptions, abnormal-behavior notifications, and LLM-generated causal risk analyses and action recommendations for a representative subset.",{"name":81,"@type":72,"acceptedAnswer":82},"What coverage and event scale does K-Risk provide?",{"text":83,"@type":75},"K-Risk integrates 20 human-driven and autonomous-vehicle trajectory datasets across Europe, China, and the United States, curating 31,398 high-risk events plus an extreme subset of 1,036 near-collision cases.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":105,"slug":137},19,"General","general"]