[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-203750-105":59,"doc-detail-203750-en":130},{"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":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","resi-a-comprehensive-benchmark-for-representational-similarity-measures","RESI - A COMPREHENSIVE BENCHMARK FOR REPRESENTATIONAL SIMILARITY MEASURES","","Representational similarity measurement is a core challenge in machine learning, because it offers a way to compare how neural architectures encode information. This paper introduces the RESI benchmark, the first comprehensive evaluation framework grounded in explicit similarity assumptions. RESI includes six tests, 24 representational similarity measures, 14 neural network architectures, and seven datasets spanning graph, language, and vision. The full benchmark is publicly available to support reproducible research and extensions.",{"@graph":69,"@context":122},[70,84,105],{"@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/resi-a-comprehensive-benchmark-for-representational-similarity-measures/203750/",{"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/resi-a-comprehensive-benchmark-for-representational-similarity-measures/203750.png","ImageObject",300,407,{"name":92,"@type":93},"eBook King","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-08","2026-09-04",true,{"@type":102,"interactionType":103,"userInteractionCount":39},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"What does the RESI benchmark evaluate?","Question",{"text":112,"@type":113},"RESI evaluates the quality of representational similarity measures by testing how well they quantify similarity between neural representations under well-defined ground-truth assumptions.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"What components are included in the RESI benchmark?",{"text":117,"@type":113},"RESI consists of six carefully designed tests, 24 similarity measures, 14 neural network architectures, and seven datasets covering graph, language, and vision domains.",{"name":119,"@type":110,"acceptedAnswer":120},"How does RESI ensure similarity ground truth in practice?",{"text":121,"@type":113},"RESI constructs tests by designing model sets where similarity between representations can be grounded in practice, then ranks measures by how well their similarity estimates match the ground truth.","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},203750,1788563198,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":39,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":36,"language":139,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":140,"faqs":141,"seo_title":142,"seo_description":67,"update_tm":129,"read_time":143},962088006270,"https://ap-avatar.wpscdn.com/davatar_085a072bc5b1113ac321206ff7593b45","RESI: A COMPREHENSIVE BENCHMARK FOR REPRESENTATIONAL SIMILARITY MEASURES  \nMax Klabunde 1 ∗ Tassilo Wald2 ,3 ,4∗ Tobias Schumacher5 ,6∗†  \nKlaus Maier-Hein2 ,3 ,4 ,7 ,8 Markus Strohmaier5 ,9 , 10 Florian Lemmerich 1  \n1University of Passau 2Medical Image Computing, German Cancer Research Center (DKFZ)  \n3Helmholtz Imaging, DKFZ 4University of Heidelberg 5University of Mannheim  \n6RWTH Aachen University 7Heidelberg University Hospital  \n8National Center for Tumor Diseases (NCT) Heidelberg  \n9 GESIS-Leibniz Institute for the Social Sciences 10 Complexity Science Hub  \nABSTRACT  \nMeasuring the similarity of different representations of neural architectures is a fundamental task and an open research challenge for the machine learning community. This paper presents the first comprehensive benchmark for evaluating representational similarity measures based on well-defined groundings of similarity. The representational similarity (ReSi) benchmark consists of (i) six carefully designed tests for similarity measures,(ii) 24 similarity measures,(iii) 14 neural network architectures, and (iv) seven datasets, spanning the graph, language, and vision domains. The benchmark opens up several important avenues of research on representational similarity that enable novel explorations and applications of neural architectures. We demonstrate the utility of the ReSi benchmark by conducting experiments on various neural network architectures, real-world datasets, and similarity measures. All components of the benchmark are publicly available 1 and thereby facilitate systematic reproduction and production of research results.  \nThe benchmark is extensible; future research can build on it and expand on it. We believe that the ReSi benchmark can serve as a sound platform catalyzing future research that aims to systematically evaluate existing and explore novel ways of comparing representations of neural architectures.  \n1 INTRODUCTION  \nRepresentations are fundamental concepts of deep learning that have garnered significant interest due to their ability to shed light on the opaque inner workings of neural networks. Studying and analyzing them has enabled insight into numerous problems, for example understanding learning dynamics (Morcos et al., 2018; Mehrer et al., 2018), catastrophic forgetting (Ramasesh et al., 2021), and language changes over time (Hamilton et al., 2016a) . Such analyses commonly involve measuring similarity of representations, which resulted in a plethora of similarity measures proposed in the literature (Klabunde et al., 2023; Sucholutsky et al., 2023) . However, these similarity measures have often been proposed in an ad hoc manner, without a comprehensive comparison to existing similarity measures. Moreover, they have often been proposed in conjunction with new quality criteria that were deemed desirable, with previously defined quality criteria being ignored. So far, only the few most popular measures have been compared (Ding et al., 2021; Hayne et al., 2024) or analyzed in more detail (Dujmovi et al., 2023; Cui et al., 2022; Davari et al., 2022) .  \nIn this work, we present the first comprehensive benchmark for representational similarity measures. It comprises six tests that postulate different ground truth assumptions about the similarities between representations that measures could capture. We implemented these tests across several architecturesand datasets in the graph, language, and vision domains. The ReSi benchmark enables tests for 24 similarity measures that have been proposed in the literature, and we illustrate how the results can  \n∗Equal contribution. Author order among the co-first authors may be adjusted for individual use.†[Corresponding author. tobias.schumacher@uni-mannheim.de](Corresponding author. tobias.schumacher@uni-mannheim.de)[ ](Corresponding author. tobias.schumacher@uni-mannheim.de)1 [https://github.com/mklabunde/resi](https://github.com/mklabunde/resi)  \n\n| ReSi Benchmark Test |  |  |  |  |  |  ","cbCaifO4zc3mwYMg","https://ap.wps.com/l/cbCaifO4zc3mwYMg","pdf",5758233,"English","# Abstract\n# Introduction\n# Grounding Representational Similarity\n## Representational Similarity","[{\"question\":\"What does the RESI benchmark evaluate?\",\"answer\":\"RESI evaluates the quality of representational similarity measures by testing how well they quantify similarity between neural representations under well-defined ground-truth assumptions.\"},{\"question\":\"What components are included in the RESI benchmark?\",\"answer\":\"RESI consists of six carefully designed tests, 24 similarity measures, 14 neural network architectures, and seven datasets covering graph, language, and vision domains.\"},{\"question\":\"How does RESI ensure similarity ground truth in practice?\",\"answer\":\"RESI constructs tests by designing model sets where similarity between representations can be grounded in practice, then ranks measures by how well their similarity estimates match the ground truth.\"}]","RESI - A COMPREHENSIVE BENCHMARK FOR REPRESENTATIONAL SIMILARITY MEASURES | PDF",101]