[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-327239-105":59,"doc-detail-327239-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","explainable-novel-category-discovery-in-semantic-concept-space","Explainable Novel Category Discovery in Semantic Concept Space","","Novel category discovery identifies unseen classes from unlabeled data by transferring knowledge from labeled categories, but many approaches operate in opaque latent feature spaces, making discovered group semantics hard to justify. xNCD is an explainable framework that performs representation learning and pseudo-label assignment directly in a structured semantic concept space. It learns label-free concept representations by aligning visual features with vision-language similarity priors from pretrained multimodal models and uses a unified self-labeling objective over concept-space logits. Theoretical analysis shows a semantic bottleneck restricts the hypothesis class and biases partitions toward interpretable concept coordinates. Experiments on CIFAR-10, CIFAR-100, and CUB-200 show strong performance with human-readable explanations, including 92.63% on CIFAR-10, 76.45% on CIFAR-100, and intrinsic cluster-and-instance interpretability.",{"@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/explainable-novel-category-discovery-in-semantic-concept-space/327239/",{"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/explainable-novel-category-discovery-in-semantic-concept-space/327239.png","ImageObject",300,407,{"name":92,"@type":93},"Evangeline","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-24","2026-09-21",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",13,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What problem does this document address in novel category discovery?","Question",{"text":113,"@type":114},"It addresses how to identify previously unseen semantic categories from unlabeled data while transferring supervision from known categories.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"How does xNCD make discovered categories explainable?",{"text":118,"@type":114},"xNCD conducts representation learning and discovery directly in a structured semantic concept space, producing stable, human-readable concept signatures and instance-level concept evidence.",{"name":120,"@type":111,"acceptedAnswer":121},"What evidence does the document provide for xNCD’s effectiveness?",{"text":122,"@type":114},"Experiments on CIFAR-10, CIFAR-100, and CUB-200 show strong discovery performance and intrinsic explanations, including 92.63% accuracy on CIFAR-10 and improvement on CIFAR-100 from 73.2% to 76.45%.","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},327239,1790293796,{"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":26,"preview_page_count":4},13056703019662,"https://ap-avatar.wpscdn.com/avatar/be000253a8e92610077?_k=1778726343310543188","arXiv :2607 .04548v 1 [ cs .CV] 5 Jul 2026  \nExplainable Novel Category Discovery in Semantic  \nConcept Space  \nIfrat Ikhtear Uddin 1 , Yang Zhou2 , KC Santosh 1 , Longwei Wang 1†,  \n[ifratikhtear.uddin@coyotes.usd.edu](ifratikhtear.uddin@coyotes.usd.edu) , [longwei.wang@usd.edu](longwei.wang@usd.edu) , [yangzhou@auburn.edu](yangzhou@auburn.edu) , [kc.santosh@usd.edu](kc.santosh@usd.edu)  \n1Department of Computer Science, University of South Dakota, USA  \n2Department of Computer Science and Software Engineering, Auburn University, USA  \nAbstract  \nNovel category discovery aims to identify unseen classes from unlabeled data by transferring knowledge from labeled categories, but most existing methods perform discovery in opaque latent feature spaces. As a result, they may separate novel categories accurately while providing little insight into what semantic evidence defines each discovered group. We propose xNCD, an explainable novel category discovery framework that performs both representation-based discovery and pseudo-label assignment directly in a structured semantic concept space. Instead of clustering arbitrary deep features, xNCD learns a label-free concept representation by aligning visual features with vision-language similarity priors from pretrained multimodal models, and then applies a unified labeled-and-unlabeled self-labeling objective over concept-space logits. This design makes each discovered category explainable by construction through stable concept signatures and instance-level concept evidence. Theoretically, we show that routing discovery through a semantic concept bottleneck induces a strict restriction of the feature-space hypothesis class, excluding a large family of unconstrained decision rules and biasing induced partitions toward semantically interpretable concept coordinates. Experiments on CIFAR-10, CIFAR-100, and CUB-200 demonstrate that xNCD preserves strong discovery performance while providing intrinsic explanations. Under task-agnostic evaluation, xNCD achieves 92.63% overall accuracy on CIFAR-10, close to UNO’s 93.4%, and improves CIFAR-100 overall accuracy from 73.2% to 76.45%, while being the only compared method that provides human-readable cluster-and instance-level explanations.  \n1 Introduction  \nNovel Category Discovery addresses the problem of identifying previously unseen semantic categories from unlabeled data while leveraging supervision from a set of known categories [1, 2] . This setting naturally arises in realistic deployments, where labeled data are limited to a subset of categories and new categories emerge over time. Recent advances in representation learning and clustering [3– 18, 13, 19–25] have led to strong empirical performance on NCD benchmarks [26–28], demonstrating that high-quality embeddings can effectively separate known and novel categories.  \nDespite these successes, existing NCD methods largely operate in latent feature spaces whose dimensions lack explicit semantic meaning [29] . As a result, while novel categories may be correctly separated, the reason why a group of samples forms a distinct category remains opaque. This  \nlimitation makes it difficult to validate discovered categories, diagnose failure modes, or integrate † Corresponding Authors.  \n40th Conference on Neural Information Processing Systems (NeurIPS 2026) .  \nhuman knowledge into the discovery process. Such opacity is particularly problematic in scientific, safety-critical, and human-in-the-loop applications [30–32], where interpretability and accountability are essential.  \nIn this work, we argue that semantic interpretability should be an intrinsic property of the novel category discovery process, rather than a post hoc analysis applied after clustering. To this end, we propose an interpretable framework for NCD that performs representation learning and category discovery directly in a structured semantic concept space [33, 34] . Instead of relying on abstract embeddings, the framework learn","cbCairmE4QrOTH5j","https://ap.wps.com/l/cbCairmE4QrOTH5j","pdf",5445838,24,"English","# Abstract\n# 1 Introduction\n# 2 Related Works","[{\"question\":\"What problem does this document address in novel category discovery?\",\"answer\":\"It addresses how to identify previously unseen semantic categories from unlabeled data while transferring supervision from known categories.\"},{\"question\":\"How does xNCD make discovered categories explainable?\",\"answer\":\"xNCD conducts representation learning and discovery directly in a structured semantic concept space, producing stable, human-readable concept signatures and instance-level concept evidence.\"},{\"question\":\"What evidence does the document provide for xNCD’s effectiveness?\",\"answer\":\"Experiments on CIFAR-10, CIFAR-100, and CUB-200 show strong discovery performance and intrinsic explanations, including 92.63% accuracy on CIFAR-10 and improvement on CIFAR-100 from 73.2% to 76.45%.\"}]","Explainable Novel Category Discovery in Semantic Concept Space | PDF",1789976784]