[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-83679-en":3,"doc-seo-83679-105":29,"detail-sidebar-cat-0-en-105":91},{"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":11,"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},83679,4810365810221,"Aurora","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","Provable Pruning for Efficient 3D Gaussian Splatting via Coresets","3D Gaussian Splatting (3DGS) supports high-quality real-time novel-view synthesis, yet real scenes may contain millions of Gaussians, making compression critical for deployment on limited hardware. Existing pruning methods are often heuristic and lack multiplicative approximation guarantees, leading to expensive post-pruning finetuning. This work asks whether a 3DGS scene can be provably replaced by a much smaller weighted coreset while preserving the rendering objective. Results show no non-trivial multiplicative coreset in the unrestricted case, but resolution-dependent weighted coreset constructions are possible. Gaussian sampling uses sensitivity-based importance scores, and objective guarantees extend to rendering under stability assumptions.","arXiv :2607 .0272 1v 1 [ cs .CV] 2 Jul 2026  \nProvable Pruning for Efficient 3D Gaussian Splatting via  \nCoresets  \nWaseem Mousa 1 Alaa Maalouf1  \n1Department of Computer Science, University of Haifa  \nProject page and open-source code: [github.com/waseem-m/3dgs_provable_coresets](github.com/waseem-m/3dgs_provable_coresets)  \nAbstract  \n3D Gaussian Splatting (3DGS) enables high-quality real-time novel-view synthesis, but practical scenes often contain millions of Gaussians, making compression essential for deployment on limited hardware. Existing reduction methods are effective but mostly heuristic: they provide no multiplicative approximation guarantee for the rendered objective, and thus rely heavily on costly post-pruning finetuning to recover quality. We ask a basic question: can a 3DGS scene be provably replaced by a much smaller weighted subset (coreset) while preserving the objective of interest? We first show that, in the unrestricted setting, no non-trivial multiplicative 3DGS coreset exists. We then show that multiplicative guarantees are not impossible, but resolution-dependent. For a prescribed rendering resolution, such as representative views or grids of views/rays, we provide the first weighted coreset construction theorem for 3DGS. The construction samples Gaussians by sensitivity: provable importance scores measuring each Gaussian’s role in the full-scene objective. Finally, under explicit validity and log-transmittance stability assumptions, we turn this objective guarantee into a rendering guarantee. Empirically, our method is strongest where deployment needs it most: aggressive compression with no or minimal recovery compute. In prune-only and very short finetuning regimes, it achieves state-of-the-art performance, showing that principled importance estimation can be both theoretically meaningful and practically useful. Open-source code is available [at github.com/waseem-m/3dgs_provable_coresets](at github.com/waseem-m/3dgs_provable_coresets).  \nFigure 1: Our method prunes a pretrained 3D Gaussian scene while provably preserving the quality on a set of queries. In the regime where no finetuning is used, our approach preserves rendering quality better than competing pruning heuristics, especially at aggressive pruning ratios.  \n1 Introduction  \n3D Gaussian Splatting (3DGS) has become a central representation for real-time novel-view synthesis, combining explicit Gaussian scene models with visibility-aware differentiable splatting and strong fidelity– speed tradeoffs [1–8] . Yet, high-quality reconstructions often require large Gaussian sets, creating significant  \nmemory, storage, bandwidth, and runtime costs, especially on resource-limited hardware. This has motivated growing work on 3DGS pruning and global compaction [9–22] .  \nThe gap. While these methods demonstrate that strong compression is often possible, they leave two key bottlenecks. First, the compressed Gaussian set is often followed by a costly recovery stage, requiring substantial hardware resources and many fine-tuning iterations. Second, existing approaches are largely heuristic, providing no provable guarantees on approximation error or subset size.  \nOur approach: Coresets for 3DGS. To this end, we study 3DGS pruning through the lens of coresets: small weighted subsets that approximately preserve a target objective over a prescribed family of queries. This leads to our central question: Can a full 3DGS scene be provably replaced by a much smaller weighted subset of the gaussian while preserving the target rendering objective, and under what assumptions is such a guarantee possible? This question is especially important in low-compute settings. When only a prune-only evaluation or a very short recovery phase is feasible, the selected subset directly determines performance. A principled compression method should therefore identify which Gaussians are most important for preserving the rendered objective, while providing explicit control over t","cbCaiupEV773mIw4","https://ap.wps.com/l/cbCaiupEV773mIw4","pdf",7820614,1,39,"English","en",105,"# Abstract\n# Introduction\n## The gap\n## Our approach: Coresets for 3DGS\n## The challenge\n## A glimpse into our findings\n## Our Contribution","[{\"question\":\"Why is pruning important in 3D Gaussian Splatting deployments?\",\"answer\":\"High-quality 3DGS reconstructions often require very large sets of Gaussians, causing heavy memory, storage, bandwidth, and runtime costs. Pruning reduces the Gaussian set to meet limited hardware constraints.\"},{\"question\":\"Do existing 3DGS pruning methods provide provable approximation guarantees?\",\"answer\":\"Existing reduction methods are described as mostly heuristic. They typically do not provide multiplicative approximation guarantees for the rendered objective, which is why costly post-pruning finetuning is often needed to recover quality.\"},{\"question\":\"Under what conditions can a provable weighted coreset be constructed for 3DGS?\",\"answer\":\"No non-trivial multiplicative coreset exists in the unrestricted setting, especially when the query family can isolate individual Gaussians. A weighted coreset construction becomes possible under a finite set of queries induced by a fixed rendering resolution, together with assumptions such as stable transmittance/occlusion and other validity and log-transmittance stability conditions.\"}]",1784189689,98,{"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":86,"head_meta":88,"extra_data":90,"updated_unix":27},"provable-pruning-for-efficient-3d-gaussian-splatting-via-coresets","",{"@graph":35,"@context":85},[36,53,68],{"@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/provable-pruning-for-efficient-3d-gaussian-splatting-via-coresets/83679/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-26","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":11},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why is pruning important in 3D Gaussian Splatting deployments?","Question",{"text":75,"@type":76},"High-quality 3DGS reconstructions often require very large sets of Gaussians, causing heavy memory, storage, bandwidth, and runtime costs. Pruning reduces the Gaussian set to meet limited hardware constraints.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"Do existing 3DGS pruning methods provide provable approximation guarantees?",{"text":80,"@type":76},"Existing reduction methods are described as mostly heuristic. They typically do not provide multiplicative approximation guarantees for the rendered objective, which is why costly post-pruning finetuning is often needed to recover quality.",{"name":82,"@type":73,"acceptedAnswer":83},"Under what conditions can a provable weighted coreset be constructed for 3DGS?",{"text":84,"@type":76},"No non-trivial multiplicative coreset exists in the unrestricted setting, especially when the query family can isolate individual Gaussians. A weighted coreset construction becomes possible under a finite set of queries induced by a fixed rendering resolution, together with assumptions such as stable transmittance/occlusion and other validity and log-transmittance stability conditions.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]