[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-137934-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-137934-en":130},{"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","beyondmix-leveraging-structural-priors-and-long-range-dependencies-for-domain-invariant-lidar-segmentation-paper","BeyondMix - Leveraging Structural Priors and Long-Range Dependencies for Domain-Invariant LiDAR Segmentation - Paper","","Domain adaptation for LiDAR semantic segmentation remains challenging due to the complex structural properties of point cloud data. While mix-based paradigms show promise, they often fail to fully leverage rich structural priors in 3D LiDAR point clouds. BeyondMix introduces a framework that uses State Space Models (Mamba) to exploit permutation invariance, local consistency, and geometric consistency, while modeling long-range dependencies beyond the limited receptive fields of voxel-based methods. Extensive experiments show consistent SOTA gains on benchmarks and establish a new direction for unsupervised domain adaptation in 3D perception.",{"@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/beyondmix-leveraging-structural-priors-and-long-range-dependencies-for-domain-invariant-lidar-segmentation-paper/137934/",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/beyondmix-leveraging-structural-priors-and-long-range-dependencies-for-domain-invariant-lidar-segmentation-paper/137934.png","ImageObject",300,407,{"name":42,"@type":43},"นรินทร์","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-09-18","2026-08-23",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",6,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What problem does BeyondMix address in LiDAR semantic segmentation?","Question",{"text":63,"@type":64},"BeyondMix targets unsupervised domain adaptation, where models trained on a labeled source domain must generalize to an unlabeled target domain despite domain shifts in 3D point cloud structure.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"Which structural priors does BeyondMix leverage?",{"text":68,"@type":64},"BeyondMix identifies and exploits three structural priors: permutation invariance, local consistency, and geometric consistency, aiming to better preserve 3D structural information during adaptation.",{"name":70,"@type":61,"acceptedAnswer":71},"How does BeyondMix model long-range dependencies in point clouds?",{"text":72,"@type":64},"BeyondMix uses State Space Models (specifically Mamba) to capture dependencies that go beyond the limited receptive fields of conventional voxel-based approaches, improving domain-invariant representation learning.","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},137934,1787471077,{"code":4,"msg":82,"data":83},"success",[84,88,92,96,101,105,110,114,119,122,126],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":85,"show_sort_weight":86,"slug":87},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":89,"show_sort_weight":90,"slug":91},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":93,"show_sort_weight":94,"slug":95},"Exam",70,"exam",{"id":97,"doc_module":4,"doc_module_name":25,"category_name":98,"show_sort_weight":99,"slug":100},5,"Comic",60,"comic",{"id":55,"doc_module":4,"doc_module_name":25,"category_name":102,"show_sort_weight":103,"slug":104},"Technology",50,"technology",{"id":106,"doc_module":4,"doc_module_name":25,"category_name":107,"show_sort_weight":108,"slug":109},7,"Healthcare",40,"healthcare",{"id":111,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":112,"slug":113},8,30,"research-report",{"id":115,"doc_module":4,"doc_module_name":25,"category_name":116,"show_sort_weight":117,"slug":118},9,"Religion & Spirituality",20,"religion-spirituality",{"id":117,"doc_module":4,"doc_module_name":25,"category_name":120,"show_sort_weight":117,"slug":121},"World Cup","world-cup",{"id":123,"doc_module":4,"doc_module_name":25,"category_name":124,"show_sort_weight":123,"slug":125},10,"Lifestyle","lifestyle",{"id":127,"doc_module":4,"doc_module_name":25,"category_name":128,"show_sort_weight":97,"slug":129},19,"General","general",{"code":4,"msg":82,"data":131},{"doc_id":79,"user_id":132,"nickname":42,"user_avatar":133,"doc_module":4,"category_id":111,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":139,"language":140,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":12,"update_tm":80,"read_time":144},2336475104957,"https://ap-avatar.wpscdn.com/avatar/22000c4c6bd8a5076e1?x-image-process=image/resize,m_fixed,w_180,h_180&k=1787554080175789136","BeyondMix: Leveraging Structural Priors and Long-Range Dependencies for Domain-Invariant LiDAR Segmentation  \nYujia Chen 1,2 , Rui Sun3 , Wangkai Li 1,2 , Huayu Mai 1,2 , Si Chen 1,2 , Zhuoyuan Li 1,2 , Zhixin Cheng 1,2 , Tianzhu Zhang 1,2 ∗  \n1University of Science and Technology of China  \n2National Key Laboratoray of Deep Space Exploration, Deep Space Exploration Laboratory  \n3 Shenzhen International Graduate School, Tsinghua University  \n{yujia_chen, issunrui, lwklwk, mai556, sa23010094, [chengzhixin}@mail.ustc.edu.cn](chengzhixin}@mail.ustc.edu.cn), [tzzhang@ustc.edu.cn](tzzhang@ustc.edu.cn)  \nAbstract  \nDomain adaptation for LiDAR semantic segmentation remains challenging due to the complex structural properties of point cloud data. While mix-based paradigmshave shown promise, they often fail to fully leverage the rich structural priors inherent in 3D LiDAR point clouds. In this paper, we identify three critical yet underexploited structural priors: permutation invariance, local consistency, and geometric consistency. We introduce BeyondMix, a novel framework that harnesses the capabilities of State Space Models (specifically Mamba) to construct and exploit these structural priors while modeling long-range dependencies that transcend the limited receptive fields of conventional voxel-based approaches. By employing space-filling curves to impose sequential ordering on point cloud data and implementing strategic spatial partitioning schemes, BeyondMix effectively captures domain-invariant representations. Extensive experiments on challenging LiDAR semantic segmentation benchmarks demonstrate that our approach consistently outperforms existing state-of-the-art methods, establishing a new paradigm for unsupervised domain adaptation in 3D point cloud understanding.  \n1 Introduction  \nLiDAR sensors maintain operational integrity under adverse conditions [41, 64, 6, 20] where camerabased perception fails. Semantic segmentation enables critical scene understanding for autonomous navigation safety [46, 39, 7] . Despite significant advancements through deep learning methodologies [16, 33, 79, 62, 1], LiDAR segmentation requires extensive annotated datasets—a substantial challenge given the prohibitive resource requirements for manually labeling point clouds comprising about 105 points per scan. While synthetic data provides readily available annotations, it introduces domain shift, violating the i.i.d. assumption between training and deployment distributions in statistical learning theory [58], consequently degrading model performance. Unsupervised domain adaptation (UDA) techniques have been extensively studied to address this issue by transferring knowledge from a labeled source domain to an unlabeled target domain, and improve the model’s performance on the target dataset without requiring additional annotations. The unstructured nature of LiDAR point clouds coupled with challenges in designing effective alignment methodologies renders UDA for LiDAR segmentation particularly difficult.  \n∗ Corresponding author  \n39th Conference on Neural Information Processing Systems (NeurIPS 2025) .  \nmIoU  \n(a)  \nCategory Semantic Integrity Z axis  \nRegion 1  \nRegion 2  \nSpatial Region Integrity  \nLocal 1  \nLocal 2  \nLocal Consistency  \n(b)  \n| Properties\\Prior | Category Semantic Integrity | Azimuthal Semantic Consistency | Spatial Region Integrity | Permutation Invariance | Local Consistency | Geometric Consistency |\n| --- | --- | --- | --- | --- | --- | --- |\n| Feasibility of Mix | 􀂗 | 􀂗 | 􀂗 | - | - | - |\n| Effective Receptive\u003Cbr>Field | 􀁵 | 􀁵 | 􀁵 | - | - | - |\n| Feasibility of Mix | 􀂗 | 􀂗 | 􀂗 | 􀁵 | 􀁵 | 􀁵 |\n| Effective Receptive\u003Cbr>Field | 􀂗 | 􀂗 | 􀂗 | 􀂗 | 􀂗 | 􀂗 |\n\nPrevious  \n􀀯  \nOurs  \n☺  \n(c)  \nFigure 1: (a) Performance comparison across diverse mixing methodologies. (b) Schematic illustration of distinct structural priors. (c) Comparative analysis of properties between our proposed approach and existing methods.  \nIn previous work, the ","cbCaibzC9PdgqPi8","https://ap.wps.com/l/cbCaibzC9PdgqPi8","pdf",2676444,27,"English","# Introduction\n## Domain adaptation for LiDAR semantic segmentation\n## Mix-based paradigms and their limitations\n## Structural priors and the proposed BeyondMix idea\n## Permutation invariance prior\n## Local consistency prior\n## Geometric consistency prior","[{\"question\":\"What problem does BeyondMix address in LiDAR semantic segmentation?\",\"answer\":\"BeyondMix targets unsupervised domain adaptation, where models trained on a labeled source domain must generalize to an unlabeled target domain despite domain shifts in 3D point cloud structure.\"},{\"question\":\"Which structural priors does BeyondMix leverage?\",\"answer\":\"BeyondMix identifies and exploits three structural priors: permutation invariance, local consistency, and geometric consistency, aiming to better preserve 3D structural information during adaptation.\"},{\"question\":\"How does BeyondMix model long-range dependencies in point clouds?\",\"answer\":\"BeyondMix uses State Space Models (specifically Mamba) to capture dependencies that go beyond the limited receptive fields of conventional voxel-based approaches, improving domain-invariant representation learning.\"}]","BeyondMix - Leveraging Structural Priors and Long-Range Dependencies for Domain-Invariant LiDAR Segmentation - Paper | PDF",68]