[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-83471-en":3,"doc-seo-83471-105":30,"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":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},83471,1099513958762,"Logic","https://ap-avatar.wpscdn.com/avatar/1000023916a998db790?x-image-process=image/resize,m_fixed,w_180,h_180&k=1784791008015729253",8,"Research & Report","Wake up for Touch Mask-isolated Tactile Alignment Learning in MLLMs","Touch provides the grounding needed to perceive intrinsic material properties such as friction and compliance that vision alone often cannot resolve. Prior work on adding tactile sense to multimodal LLMs faces a zero-sum constraint in compact models: limited parameters force a choice between new sensory capability and maintaining vision-language reasoning. Splash introduces mask-isolated tactile alignment for MLLMs by partitioning parameters into dormant and critical subspaces, freezing critical weights and updating dormant ones to prevent catastrophic forgetting and enable non-destructive modality expansion.","arXiv :2607 .00302v 1 [ cs .CV] 1 Jul 2026  \nWake up for Touch! Mask-isolated Tactile Alignment Learning in MLLMs  \nYoonhyung Park∗ , Minji Kim∗ , Sungwon Moon, and Jiyoung Lee† Division of Artificial Intelligence & Software, Ewha Womans University {pyoon0820, [xxinzzi03}@ewhain.net](xxinzzi03}@ewhain.net) , {sungwon268,[lee.jiyoung}@ewha.ac.kr](lee.jiyoung}@ewha.ac.kr)  \n[http://mmai.ewha.ac.kr/splash/](http://mmai.ewha.ac.kr/splash/)  \nAbstract. Touch supplies the physical grounding needed to perceive intrinsic material properties, such as friction and compliance, that vision alone often cannot resolve. Recent efforts for equipping multimodal LLMs with this tactile sense, however, expose a zero-sum trade-off: the limited parameter budget of compact models forces a choice between acquiring the new sensory modality and preserving the established visionlanguage reasoning. We present Splash, a mask-isolated tactile alignment learning framework for MLLMs. Splash quantifies the significance of each pretrained parameter, and partitions the parameter space into a dormant and critical subspace. While the frozen critical subspace acts as a stable anchor to safeguard general visual knowledge, Splash updates the isolated dormant subspace to internalize tactile alignment towards LLMs. This selective, non-destructive expansion effectively prevents catastrophic forgetting and ensures non-destructive modality expansion. Extensive experiments show that Splash effectively achieves tactile reasoning without additional inference overhead in the LLM part, demonstrating state-of-the-art performance on visuo-tactile benchmarks, including SSVTP, TVL, and TacQuad, while preserving its original generalpurpose capabilities.  \nKeywords: Visuo-Tactile-Language Learning · Catastrophic Forgetting  \n· Multimodal Large Language Models  \n1 Introduction  \nHumans interact with the physical world not merely by looking at it, but by touching or pressing it to judge whether a surface is slippery, compliant, or rigid. This innate ability to perceive intrinsic material properties such as friction and compliance is essential feedback for precise physical interaction. For embodied robots to achieve similar dexterity, perceiving fine-grained contact dynamics is crucial for manipulating objects. However, vision-based systems often struggle to infer such properties, which are frequently occluded or visually indistinguishable during active manipulation [4, 50] .  \n*  \n†  \nThese authors contributed equally to this work. Corresponding author.  \n2 Y. Park and M. Kim et al.  \nFig. 1: Example of the catastrophic forgetting problem in tactile alignment for MLLMs (e.g ., TVL [21] w/ Qwen2.5-VL-3B) . A small amount of tactile training set often forgets the visual sense in the base MLLM. More failure cases in Appendix.  \nTo close this gap, recent approaches [9, 16, 30, 47] have explored learning cross-modal associations between visual appearance and tactile feedback. These approaches typically learn a shared visuo-tactile-language (VTL) representation space. While effective for discriminative tasks such as retrieval and classification, they lack the capacity for the higher-level semantic reasoning and instruction following that complex embodied decision-making demands. Moreover, their learned representations generalize poorly to novel objects with visually ambiguous textures. These observations indicate the need for models capable of reasoning over multimodal sensory inputs.  \nMultimodal large language models (MLLMs) [2, 8, 34] are a natural fit. By drawing on the world knowledge encoded in their LLM backbones, MLLMs support open-vocabulary reasoning over sensory inputs, allowing robots not only to perceive but also to reason about the physical consequences of their actions. Most existing MLLMs, however, rely on computationally intensive backbones (e.g ., LLaMA-7B [43]), which limit their practicality for resource-constrained edge robots [45] . This motivates compact MLLMs capabl","cbCaii7cmNFHZuss","https://ap.wps.com/l/cbCaii7cmNFHZuss","pdf",4184739,3,1,27,"English","en",105,"# Introduction\n## Background and Motivation\n## Challenge: Catastrophic Forgetting in Compact MLLMs\n## Proposed Approach: Splash\n## Experimental Validation","[{\"question\":\"Why is tactile perception important for multimodal LLM-based physical interaction?\",\"answer\":\"Tactile feedback reveals intrinsic material properties like friction and compliance that vision often cannot determine, enabling more precise physical manipulation and reasoning about action outcomes.\"},{\"question\":\"What core problem does Splash address when aligning tactile signals to compact MLLMs?\",\"answer\":\"Aligning tactile signals in small MLLMs can distort pretrained visual features, causing catastrophic forgetting and degrading vision-language reasoning, creating a trade-off between sensory expansion and preserved reasoning.\"},{\"question\":\"How does Splash mitigate catastrophic forgetting during tactile alignment?\",\"answer\":\"Splash estimates parameter importance and partitions parameters into dormant and critical subspaces, freezing the critical (stable anchor) parameters while updating only the isolated dormant subspace to incorporate tactile alignment without destructive changes.\"}]",1784188199,68,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"wake-up-for-touch-mask-isolated-tactile-alignment-learning-in-mllms","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":20},"https://docshare.wps.com/document/research-report/",{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/wake-up-for-touch-mask-isolated-tactile-alignment-learning-in-mllms/83471/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-26","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why is tactile perception important for multimodal LLM-based physical interaction?","Question",{"text":75,"@type":76},"Tactile feedback reveals intrinsic material properties like friction and compliance that vision often cannot determine, enabling more precise physical manipulation and reasoning about action outcomes.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What core problem does Splash address when aligning tactile signals to compact MLLMs?",{"text":80,"@type":76},"Aligning tactile signals in small MLLMs can distort pretrained visual features, causing catastrophic forgetting and degrading vision-language reasoning, creating a trade-off between sensory expansion and preserved reasoning.",{"name":82,"@type":73,"acceptedAnswer":83},"How does Splash mitigate catastrophic forgetting during tactile alignment?",{"text":84,"@type":76},"Splash estimates parameter importance and partitions parameters into dormant and critical subspaces, freezing the critical (stable anchor) parameters while updating only the isolated dormant subspace to incorporate tactile alignment without destructive changes.","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":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]