[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-327333-105":3,"doc-detail-327333-en":81,"detail-sidebar-cat-0-en-105":99},{"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","medstreambench-a-time-aware-benchmark-for-streaming-and-proactive-medical-video-understanding","MedStreamBench - A Time-Aware Benchmark for Streaming and Proactive Medical Video Understanding","","MedStreamBench is a time-aware temporal benchmark for medical video question answering that evaluates how vision-language models behave under strict evidence availability. The work contrasts conventional offline evaluation with streaming scenarios where a model must decide when evidence is sufficient to answer and when to abstain. It also introduces a proactive monitoring paradigm requiring timely decisions among no alert, uncertain, and alert. The benchmark spans 22 medical video and image-sequence datasets and provides thousands of question-answering records across retrospective, present, future, and proactive modes.",{"@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/medstreambench-a-time-aware-benchmark-for-streaming-and-proactive-medical-video-understanding/327333/",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/medstreambench-a-time-aware-benchmark-for-streaming-and-proactive-medical-video-understanding/327333.png","ImageObject",300,407,{"name":42,"@type":43},"Emma Wilson","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-31","2026-09-21",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",11,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What problem does MedStreamBench address in medical video understanding?","Question",{"text":63,"@type":64},"It focuses on time-aware evidence availability, where models must avoid using future information and should only answer or alert when sufficient visual evidence appears.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"How do the streaming and proactive paradigms differ?",{"text":68,"@type":64},"In streaming, the same question is asked over growing visual prefixes and the model must remain unanswerable until evidence is enough. In proactive monitoring, the model continuously watches the stream and chooses among no alert, uncertain, or alert.",{"name":70,"@type":61,"acceptedAnswer":71},"What temporal task modes does MedStreamBench cover?",{"text":72,"@type":64},"It includes retrospective questions about past evidence, present questions about the current visual state, future questions requiring deferred answerability or prediction, and proactive monitoring questions requiring timely alerting.","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},327333,1790551634,{"code":4,"msg":82,"data":83},"success",{"doc_id":79,"user_id":84,"nickname":42,"user_avatar":85,"doc_module":4,"category_id":86,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":87,"file_id":88,"file_url":89,"file_type":90,"file_size":91,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":92,"language":93,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":94,"faqs":95,"seo_title":96,"seo_description":12,"update_tm":97,"read_time":98,"preview_page_count":4},3848291630094,"https://eur-avatar.wpscdn.com/davatar_085a072bc5b1113ac321206ff7593b45",8,"MedStreamBench: A Time-Aware Benchmark for Streaming and Proactive Medical Video Understanding  \nthat assume full-video access, both single-turn and streaming  \ngeneral-purpose and medical vision-language grounded decision-making, with performance available at Venn2024/MedStreamBench.  \nMedical Video Benchmark, Streaming Video Understanding, Temporal Question Answering, Proactive Monitoring  \nArticle informations  \n©2026 Yuan and Shujian. License: CC-BY 4 .0  \nCorresponding author: [zuozhuliu@intl.zju.edu.cn](zuozhuliu@intl.zju.edu.cn)  \n1. Background  \nM edical video understanding is becoming increas  \ningly important for surgical intelligence, endo  \nscopic screening, procedural education, and realtime clinical decision support. Unlike static medical images,  \nadvanced the evaluation of multimodal large language models on video understanding, including long-video reasoning and online video analysis (Fu et al. , 2025 ; Wu et al. , 2024 ; Li et al. , 2025 ; Lin et al. , 2024) . However, most existing benchmarks assume either access to a complete video or a fixed clip, and primarily measure whether the final  \narXiv :2607 .01751v1  \nmedical videos encode evolving procedural context: instruments enter and leave the field, anatomical exposure changes over time, lesions may appear only transiently, and safety-relevant events often depend on the temporal order of visual evidence. Consequently, a clinically useful medical vision–language model should not only recognize what is visible, but also determine when sufficient evidence becomes available, when to abstain, and when to alert. This temporal discipline is essential in scenarios such as polyp screening, surgical workflow monitoring, view-quality assessment, and skill coaching, where premature conclusions can be misleading and delayed responses can reduce clinical utility.  \nRecent video-language benchmarks have substantially  \nanswer is semantically correct. Such offline evaluation is insufficient for medical streaming video analysis, where models must respect the information actually available ata specified time point. In parallel, medical and surgical datasets have enabled progress in phase recognition, instrument detection, action triplet recognition, polyp detection, capsule-endoscopy abnormality recognition, and surgical visual question answering (Stauder et al. , 2016 ; Nwoye et al. , 2023 ; Smedsrud et al. , 2021 ; Seenivasan et al. , 2022 ; He et al. , 2024 ; Jiang et al. , 2025a; Wang et al. , 2025 , 2026 ; Jiang et al. , 2025b) . Yet these resources typically evaluate recognition or question answering in an offline setting, and do not explicitly test whether a model can answer, defer, or alert at the correct moment under a strict evidence window .  \nTable 1: Comparison of MedStreamBench with representative surgical and medical video resources . We compare each resource by source breadth (\\#DS), evaluation protocol, and coverage of the four temporal task modes.  \n\n| Benchmark | \\#DS | Protocol |  |  | Temporal Tasks |  |  |  |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- |\n|  |  | Eval Unit | Prefix-Only | Streaming | Retrospective | Present | Future | Proactive |\n| \u003Cbr>Classical Surgical Datasets |  |  |  |  |  |  |  |  |\n| Cholec80 | 1 | frame / video | × | × | ✓ | ✓ | × | × |\n| M2CAI16-Workflow | 1 | frame / video | × | × | ✓ | ✓ | × | × |\n| JIGSAWS | 1 | gesture / video | × | × | ✓ | ✓ | × | × |\n| \u003Cbr>Recent Surgical / Medical Video Resources |  |  |  |  |  |  |  |  |\n| LLaVA-Surg | 1 | clip / video | × | × | ✓ | ✓ | × | × |\n| SurgPub-Video | 1 | clip / video | × | × | ✓ | ✓ | × | × |\n| SurgOnAir | 1 | streaming video | ✓ | ✓ | × | ✓ | × | × |\n| MedHorizon | 1 | long video | × | × | ✓ | ✓ | × | × |\n| \u003Cbr>MedStreamBench (Ours) | \u003Cbr>22 | \u003Cbr>long video / stream | \u003Cbr>✓ | \u003Cbr>✓ | \u003Cbr>✓ | \u003Cbr>✓ | \u003Cbr>✓ | \u003Cbr>✓ |\n\nWe introduce MedStreamBench, a diverse temporal benchmark for medical video question answering. To the best of our knowledge, MedStreamBench is the f","cbCaiqsPq7Es5Max","https://ap.wps.com/l/cbCaiqsPq7Es5Max","pdf",1211946,10,"English","# Background\n## Comparison with existing benchmarks\n## Proposed paradigms and evaluation goals\n## Benchmark composition and temporal taxonomy","[{\"question\":\"What problem does MedStreamBench address in medical video understanding?\",\"answer\":\"It focuses on time-aware evidence availability, where models must avoid using future information and should only answer or alert when sufficient visual evidence appears.\"},{\"question\":\"How do the streaming and proactive paradigms differ?\",\"answer\":\"In streaming, the same question is asked over growing visual prefixes and the model must remain unanswerable until evidence is enough. In proactive monitoring, the model continuously watches the stream and chooses among no alert, uncertain, or alert.\"},{\"question\":\"What temporal task modes does MedStreamBench cover?\",\"answer\":\"It includes retrospective questions about past evidence, present questions about the current visual state, future questions requiring deferred answerability or prediction, and proactive monitoring questions requiring timely alerting.\"}]","MedStreamBench - A Time-Aware Benchmark for Streaming and Proactive Medical Video Understanding | PDF",1789977000,25,{"code":4,"msg":82,"data":100},[101,105,109,113,118,123,128,131,136,139,142],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":102,"show_sort_weight":103,"slug":104},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":106,"show_sort_weight":107,"slug":108},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":110,"show_sort_weight":111,"slug":112},"Exam",70,"exam",{"id":114,"doc_module":4,"doc_module_name":25,"category_name":115,"show_sort_weight":116,"slug":117},5,"Comic",60,"comic",{"id":119,"doc_module":4,"doc_module_name":25,"category_name":120,"show_sort_weight":121,"slug":122},6,"Technology",50,"technology",{"id":124,"doc_module":4,"doc_module_name":25,"category_name":125,"show_sort_weight":126,"slug":127},7,"Healthcare",40,"healthcare",{"id":86,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":129,"slug":130},30,"research-report",{"id":132,"doc_module":4,"doc_module_name":25,"category_name":133,"show_sort_weight":134,"slug":135},9,"Religion & Spirituality",20,"religion-spirituality",{"id":134,"doc_module":4,"doc_module_name":25,"category_name":137,"show_sort_weight":134,"slug":138},"World Cup","world-cup",{"id":92,"doc_module":4,"doc_module_name":25,"category_name":140,"show_sort_weight":92,"slug":141},"Lifestyle","lifestyle",{"id":143,"doc_module":4,"doc_module_name":25,"category_name":144,"show_sort_weight":114,"slug":145},19,"General","general"]