[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-327383-105":59,"doc-detail-327383-en":134},{"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":127,"head_meta":129,"extra_data":131,"updated_unix":133},105,"en","sab-lvlm-significance-aware-binarization-for-large-vision-language-models-abstract","SAB-LVLM - Significance-Aware Binarization for Large Vision-Language Models - Abstract","","Large Vision-Language Models (LVLMs) deliver strong multimodal understanding, yet their huge parameter scale and cross-modal computation create major memory and latency costs that hinder deployment on resource-limited devices. Binarization reduces storage and computation, but prior methods ignore that weight importance varies across layers and modalities, causing task-irrelevant parameters to remain while modality-critical ones are not properly optimized. SAB-LVLM builds Hessian matrices for textual and visual inputs, derives a spatial significance map, integrates modality-guided significance across modalities, and uses it to reweight the binarization objective with alternating significance-weighted updates. Experiments show clear gains over existing binary PTQ methods under an about 1-bit compression constraint.",{"@graph":69,"@context":126},[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/sab-lvlm-significance-aware-binarization-for-large-vision-language-models-abstract/327383/",{"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/sab-lvlm-significance-aware-binarization-for-large-vision-language-models-abstract/327383.png","ImageObject",300,407,{"name":92,"@type":93},"Aurora","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-31","2026-09-21",true,{"@type":102,"interactionType":103,"userInteractionCount":44},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118,122],{"name":109,"@type":110,"acceptedAnswer":111},"Why is binarization difficult for large vision-language models (LVLMs)?","Question",{"text":112,"@type":113},"LVLMs combine visual encoders with large language backbones, and parameters across different layers and modalities contribute unevenly to downstream tasks. Many PTQ binarization methods minimize global quantization error while ignoring this heterogeneous importance.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"How does SAB-LVLM estimate which weights are significant?",{"text":117,"@type":113},"SAB-LVLM constructs Hessian matrices for textual and visual inputs, then uses a spatial significance map to separate full-precision weights activated by a single modality versus those activated across modalities.",{"name":119,"@type":110,"acceptedAnswer":120},"How is the significance map used during binarization training?",{"text":121,"@type":113},"The significance-aware binarization map is incorporated into the binarization objective as an error reweighting term, and binarization fitting is performed via an alternating significance-weighted update scheme.",{"name":123,"@type":110,"acceptedAnswer":124},"What do experiments indicate about SAB-LVLM compared with existing binary PTQ methods?",{"text":125,"@type":113},"Extensive experiments show SAB-LVLM outperforms existing binary PTQ methods under an approximately 1-bit compression constraint.","https://schema.org",{"og:url":83,"og:type":128,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":130,"canonical":83},"index,follow",{"doc_id":132,"site_id":62},327383,1790831006,{"code":4,"msg":5,"data":135},{"doc_id":132,"user_id":136,"nickname":92,"user_avatar":137,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":138,"file_id":139,"file_url":140,"file_type":141,"file_size":142,"view_count":44,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":52,"language":143,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":144,"faqs":145,"seo_title":146,"seo_description":67,"update_tm":147,"read_time":148,"preview_page_count":4},4810365810221,"https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d","SAB-LVLM: Significance-Aware Binarization for Large Vision-Language Models  \narXiv :2607 .0 1876v 1 [ cs .CV] 2 Jul 2026  \nQi Lyu 1,2,3†, Jiahua Dong4†, Baichen Liu 1,2 , Xudong Wang 1,2,3 , Mingfei Han4 , Yulun Zhang5 , Fahad Shahbaz Khan4 , Salman Khan4 , Lianqing Liu 1,2 , and Zhi Han 1,2  \n1 State Key Laboratory of Robotics and Intelligent Systems  \n2 Shenyang Institute of Automation, Chinese Academy of Sciences  \n3 University of Chinese Academy of Sciences  \n4 Mohamed bin Zayed University of Artificial Intelligence  \n5 Shanghai Jiao Tong University  \n~~ ~~ ✦ ~~ ~~  \nAbstract—Large Vision-Language Models (LVLMs) have achieved remarkable progress in multimodal understanding, yet their enormous parameter scale and cross-modal computation incur substantial memory and latency overhead, severely limiting real-world deployment on resource-constrained devices. Binarization offers an attractive solution by drastically reducing storage and computational costs. However, existing binarization methods neglect the varying importance of weights across different layers and modalities. This causes parameters irrelevant to downstream tasks to be unnecessarily retained, whereas modalitycritical weights may not be adequately optimized, resulting in significant performance degradation. To address these challenges, we develop a novel Significance-Aware Binarization for Large Vision-Language Models (SAB-LVLM) . Specifically, after constructing Hessian matrices for textual and visual inputs, we propose a spatial significance map to distinguish full-precision weights activated under a single modality from those activated across modalities. We then devise a modalityguided integration strategy to obtain the significance-aware binarization map, which measures weight significance across layers and modalities. Subsequently, this binarization map is incorporated into the binarization objective as an error reweighting term, and binarization fitting is performed through an alternating significance-weighted update scheme. Extensive experiments illustrate the superiority of our SAB-LVLM over existing binary PTQ methods under an approximately 1-bit compression constraint. Our code is accessible at [https://github.com/LyuQi127/SAB](https://github.com/LyuQi127/SAB)  LVLM.  \nIndex Terms—Binarization, post-training quantization, large visionlanguage models.  \n1 INTRODUCTION  \nIn recent years, large language models (LLMs) built upon the transformer architecture have achieved state-of-the-art results on a broad spectrum of natural language processing tasks [1] . Their strong empirical performance primarily stems from their enormous parameter counts, which frequently reach tens of billions. For instance, the open pretrained transformer (OPT) family [2] provides variants scaling up to 66 billion parameters, while the LLaMA series [3] includes even larger versions such as LLaMA3- 70B. Despite their effectiveness, these models impose an immense memory footprint and computational burden. For example, running a 70B model in full precision requires  \n†Qi Lyu and Jiahua Dong contributed equally to this work.  \nover 130 GB of memory, which poses serious challenges for deployment on resource-constrained mobile devices and restricts the widespread adoption of LLMs.  \nTo address these limitations, numerous compression techniques [4]–[6] have been developed for LLMs, such as weight quantization [7], [8], network pruning [9], [10], knowledge distillation [11], [12], and low-rank factorization [13], [14] . Among them, binarization is particularly notable since it reduces each parameter to approximately one bit, achieving an extremely high level of storage efficiency [15] . The post-training quantization (PTQ) framework enables the conversion of full-precision models into low-bit counterparts without incurring the substantial cost of retraining. As a PTQ approach, binarization enables approximately 1-bit weight quantization. Recent approaches [16]–[18] within this framework, inc","cbCait3pIkK3lHGF","https://ap.wps.com/l/cbCait3pIkK3lHGF","pdf",8022196,"English","# Introduction\n## Background and motivation\n## Challenges in extending PTQ to LVLMs\n## Proposed approach (SAB-LVLM)","[{\"question\":\"Why is binarization difficult for large vision-language models (LVLMs)?\",\"answer\":\"LVLMs combine visual encoders with large language backbones, and parameters across different layers and modalities contribute unevenly to downstream tasks. Many PTQ binarization methods minimize global quantization error while ignoring this heterogeneous importance.\"},{\"question\":\"How does SAB-LVLM estimate which weights are significant?\",\"answer\":\"SAB-LVLM constructs Hessian matrices for textual and visual inputs, then uses a spatial significance map to separate full-precision weights activated by a single modality versus those activated across modalities.\"},{\"question\":\"How is the significance map used during binarization training?\",\"answer\":\"The significance-aware binarization map is incorporated into the binarization objective as an error reweighting term, and binarization fitting is performed via an alternating significance-weighted update scheme.\"},{\"question\":\"What do experiments indicate about SAB-LVLM compared with existing binary PTQ methods?\",\"answer\":\"Extensive experiments show SAB-LVLM outperforms existing binary PTQ methods under an approximately 1-bit compression constraint.\"}]","SAB-LVLM - Significance-Aware Binarization for Large Vision-Language Models - Abstract | PDF",1789977107,25]