[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-356114-105":59,"doc-detail-356114-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","a-benchmark-study-of-vision-and-pathology-foundation-models-for-computational-pathology","A benchmark study of vision and pathology foundation models for computational pathology","","Benchmarking foundation models for computational pathology evaluates how well vision and vision-language AI systems generalize across datasets, tissue types, and clinical tasks. The study benchmarks 32 AI foundation models spanning general vision (VM), general vision-language (VLM), pathology-specific vision (Path-VM), and pathology-specific vision-language (Path-VLM) approaches. Using slide- and patch-level tasks from TCGA, CPTAC, external benchmarks, and out-of-domain datasets, Path-VMs rank among strongest on TCGA, while cross-dataset rankings shift modestly, with top-model gaps small and task dependent. Model scale and pretraining size are not reliable predictors, and late decision ensembling improves aggregate performance across datasets and tissues.",{"@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/a-benchmark-study-of-vision-and-pathology-foundation-models-for-computational-pathology/356114/",{"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/a-benchmark-study-of-vision-and-pathology-foundation-models-for-computational-pathology/356114.png","ImageObject",300,407,{"name":92,"@type":93},"dhado","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-02","2026-09-23",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",14,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What is the main goal of the benchmark study?","Question",{"text":113,"@type":114},"To assess how comparative performance and generalizability of AI foundation models behave in computational pathology across diverse datasets, tissues, and clinical tasks.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"How are the 32 models categorized and evaluated?",{"text":118,"@type":114},"Models are grouped into general vision (VM), general vision-language (VLM), pathology-specific vision (Path-VM), and pathology-specific vision-language (Path-VLM) and evaluated on slide- and patch-level tasks.",{"name":120,"@type":111,"acceptedAnswer":121},"Which model category performs best on TCGA tasks and how does this change across datasets?",{"text":122,"@type":114},"Path-VMs consistently rank among the strongest performers on TCGA. Across CPTAC and out-of-domain datasets, rankings show nuanced generalization behavior with modest but consistent shifts across datasets and task categories.","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},356114,1790247307,{"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":145,"read_time":146,"preview_page_count":81,"preview_urls":147},5909893218464,"https://ap-avatar.wpscdn.com/davatar_085a072bc5b1113ac321206ff7593b45","Article [https://doi.org/10.1038/s41467-026-76004-6](https://doi.org/10.1038/s41467-026-76004-6)  \nA benchmark study of vision and pathology foundation models for computational pathology  \n\n| Received: 9 September 2025\u003Cbr>Accepted: 8 July 2026\u003Cbr>| Rohan Bareja1, Francisco Carrillo-Perez1,5, Yuanning Zheng 1, Marija Pizurica1, Tarak Nath Nandi2, Lu Tian3, Jeanne Shen 4, Ravi Madduri 2 &\u003Cbr>Olivier Gevaert 1,3  |  |\n| --- | --- | --- |\n| | To advance precision medicine in pathology, artiﬁcial intelligence (AI)-driven foundation models must generalize across diverse datasets, tissues, and clinical tasks. However, their comparative performance and generalizability in computational pathology remain incompletely characterized. Here, we benchmark 32 AI foundation models across four categories, including general vision models (VM), general vision-language models (VLM), pathology-speciﬁc vision models (Path-VM), and pathology-speciﬁc vision-language models (Path-VLM), using slide-and patch-level tasks from The Cancer Genome Atlas (TCGA), Clinical Proteomic Tumor Analysis Consortium (CPTAC), external benchmarking datasets, and out-of-domain datasets. Across TCGA tasks, PathVMs consistently rank among the strongest performers. Evaluation across CPTAC and out-of-domain datasets reveals more nuanced generalization behavior, with model rankings showing modest but consistent shifts across datasets and task categories. Pairwise statistical comparisons indicate that differences among top-performing models are often small and task dependent. Path-VMs outperform Path-VLMs and remain competitive with VMs. Model size and pretraining dataset scale do not consistently predict downstream performance. Finally, late decision-level ensembling improves aggregate performance across external datasets and tissue types, highlighting complementary strengths across foundation models. PathBench: [https://](https://)[ ](https://)[pathbench.stanford.edu/](pathbench.stanford.edu/) |  |\n| Pathology plays a crucial role in cancer research, providing insights that drive the understanding, diagnosis, and treatment of cancer1. It involves thorough examination of tissue specimens for tumor diagnosis and identiﬁcation of important prognostic features such as tumor stage and grade, as well as biomarkers suggestive of treatment response2. Pathologists study digital pathology slides to analyze tumor histologic features, including architectural patterns and cellular |  | morphology, in order to detect and classify tumors3. Consequently, oncologic pathology plays an integral role in personalized medicine and treatment selection. Computational pathology (CPath) has enhanced the ﬁeld of pathology by integrating automated workﬂowsand processes and advanced computational techniques1. It has the potential to reduce the workload of pathologists considerably and can aid in speeding up the overall diagnostic process. With the advent of |\n\n1Division of Computational Medicine, Department of Medicine, Stanford University, Stanford, CA, USA. 2Data Science and Learning Division, Argonne National Laboratory, Lemont, IL, USA. 3Department of Biomedical Data Science, Stanford University, School of Medicine, Stanford, CA, USA. 4Department of Pathology, Stanford University, School of Medicine, Stanford, CA, USA. 5Present address: Bristol Myers Squibb, CITRE, Seville, Spain. [e-mail:](e-mail: ogevaert@stanford.edu)[ ogevaert@stanford.edu](e-mail: ogevaert@stanford.edu)  \nmachine learning and artiﬁcial intelligence, CPath has proven potential to facilitate detection, classiﬁcation, and predictive modeling of cancer4. Furthermore, CPath has proven effective in assisting pathologists with tumor grading, staging, and subtyping5, all of which support diagnostic decision-making. It has also been employed as a computational technique to study patterns, segment tumors6, and extract intricate features, thereby enhancing cancer detection7.  \nThe use of Artiﬁcial Intelligence (AI) has been at the forefron","cbCaijegSP5V56pU","https://ap.wps.com/l/cbCaijegSP5V56pU","pdf",4786335,23,"English","# Introduction\n## Motivation: Precision medicine and model generalization\n## Computational pathology workflow and AI foundations\n# Benchmark Design\n## Model categories and evaluated setups\n## Datasets and tasks\n# Results and Evaluation\n## Cross-dataset ranking behavior\n## Pairwise comparisons among top models\n## Effects of model size and pretraining scale\n## Late decision-level ensembling","[{\"question\":\"What is the main goal of the benchmark study?\",\"answer\":\"To assess how comparative performance and generalizability of AI foundation models behave in computational pathology across diverse datasets, tissues, and clinical tasks.\"},{\"question\":\"How are the 32 models categorized and evaluated?\",\"answer\":\"Models are grouped into general vision (VM), general vision-language (VLM), pathology-specific vision (Path-VM), and pathology-specific vision-language (Path-VLM) and evaluated on slide- and patch-level tasks.\"},{\"question\":\"Which model category performs best on TCGA tasks and how does this change across datasets?\",\"answer\":\"Path-VMs consistently rank among the strongest performers on TCGA. Across CPTAC and out-of-domain datasets, rankings show nuanced generalization behavior with modest but consistent shifts across datasets and task categories.\"}]","A benchmark study of vision and pathology foundation models for computational pathology | PDF",1790123482,58,[148,152,156],{"page_no":8,"url":149,"width":150,"height":151},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-08/450072d38e0c483d7ccd751084308763.webp",983,1306,{"page_no":14,"url":153,"width":154,"height":155},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-08/0c39c5a3d2a272d30e7141b084fa64c8.webp",982,1304,{"page_no":81,"url":157,"width":158,"height":159},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-08/adb1976c8a2c20fd455a9735690fe57f.webp",981,1303]