[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82122-en":3,"doc-seo-82122-105":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":20,"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},82122,1099514067415,"Rowan","https://ap-avatar.wpscdn.com/avatar/100002539d78ffe74a7?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779092875211072502",8,"Research & Report","CLAP: Direct VLM-to-VLA Adaptation via Language-Action Grounding","Vision-language-action models (VLAs) transfer semantic abilities from pretrained VLMs, yet extensive robot-data post-training and architectural changes make it hard to attribute which parts of control come from the original VLM. Direct conversion with minimal structural modification enables clearer study of cross-scale capability transfer. The key challenge is output-distribution mismatch: predicting action tokens as bare numeric sequences departs from the VLM’s pretrained language distribution. CLAP (Causal Language-Action Prediction) prepends each numeric action sequence with a natural-language action description, causally conditioning action-token generation without altering the backbone. With single-epoch fine-tuning, 2B CLAP reaches 90.8% on LIBERO (+14.9 points over VLA-0) and improves robustness on LIBEROPRO under language, object, and spatial perturbations. Open-weight releases at 0.8B, 2B, and 4B from a single VLM lineage support controlled analysis.","CLAP: Direct VLM-to-VLA Adaptation via Language-Action Grounding  \nYuri Ishitoya† Ochanomizu University  \nJeremy Siburian† University of Tokyo  \nMasashi Hamaya  \nOMRON SINIC X Corp.  \narXiv :2607 .08974v 1 [ cs .RO] 9 Jul 2026  \nKuniaki Saito  \nOMRON SINIC X Corp.  \nCristian C. Beltran-Hernandez  \nOMRON SINIC X Corp.  \nMai Nishimura∗ OMRON SINIC X Corp.  \n{masashi.hamaya, kuniaki.saito, cristian.beltran, [mai.nishimura](mai.nishimura}@sinicx.com)[}](mai.nishimura}@sinicx.com)[@sinicx.com](mai.nishimura}@sinicx.com)  \nAbstract: Vision-language-action models (VLAs) inherit semantic capabilities from pretrained VLMs, yet large-scale post-training on robot data and architectural modifications can reshape the backbone so extensively that it becomes difficult to isolate what the VLM contributes to control. Directly converting pretrained VLMs into VLAs with minimal architectural change offers a more transparent path to understanding how VLM capabilities transfer across model scales. The core obstacle is output-distribution mismatch: predicting actions as bare numeric token sequences moves generation away from the VLM’s pretrained language distribution, degrading the capabilities we seek to preserve. To address this, we propose CLAP (Causal Language-Action Prediction), which prepends each numeric action sequence with a natural-language action description, causally conditioning precise action-token prediction on a language-action plan without modifying the backbone architecture. With single-epoch fine-tuning alone, 2B CLAP achieves 90.8% on LIBERO (+14.9 pt over VLA-0) and improves robustness on LIBEROPRO under language, object, and spatial perturbations. We release CLAP at 0.8B, 2B, and 4B as an open-weight, multi-scale compact VLA family from a single VLM lineage, enabling controlled analysis of VLM-to-VLA capability transfer.  \nKeywords: Vision-Language-Action Models, Robot Manipulation  \n1 Introduction  \nRapid progress in vision-language models (VLMs) continually expands the capabilities of robot perception and reasoning, yet transferring these capabilities to robot control remains disproportionately difficult. Vision-language-action models (VLAs), which sit at the intersection of visionlanguage modeling and robot learning, pursue this transfer by fine-tuning pretrained VLMs on robot demonstrations [1, 2, 3] . Recent systems have improved action prediction through discretized action tokens, learned action experts, or diffusion action heads [1, 4], but each addition moves the VLA further from its underlying VLM in both architecture and training distribution. This architectural and data complexity makes it difficult to isolate what the pretrained VLM actually contributes to control, limiting both systematic capability analysis and the transfer of advances from the broader VLM community. What would it take to make VLA research as accessible as VLM research?  \nA central source of complexity in VLM-to-VLA adaptation is an output-distribution mismatch: standard VLA fine-tuning trains a VLM pretrained to generate semantically structured language to instead emit bare action-token sequences such as 4 12 98 3 0 0. These tokens carry little of the  \n∗Equal contribution.  \n†Work done as an intern at OMRON SINIC X Corporation.  \nFigure 1: CLAP converts a pretrained VLM directly into a deployable VLA by prepending a naturallanguage action description to numeric action tokens, keeping every prediction step closer to the VLM pretraining distribution. CLAP requires no action expert or architectural change.  \nlinguistic or spatial structure seen during pretraining, so learning to act can degrade the representations that support semantic generalization. This unresolved mismatch is especially important for compact VLAs, which use lightweight VLM backbones for efficient deployment but may be more sensitive to distribution shift during fine-tuning [5, 6] . Recent architecture-free VLM-to-VLA work provides cleaner settings for studying this issue by avoi","cbCaideZPBRRm7BU","https://ap.wps.com/l/cbCaideZPBRRm7BU","pdf",2604344,1,18,"English","en",105,"# Introduction\n## Output-distribution mismatch in VLM-to-VLA adaptation\n## CLAP: Causal Language-Action Prediction","[{\"question\":\"What main problem does CLAP address in adapting VLMs to VLAs?\",\"answer\":\"CLAP targets output-distribution mismatch: standard VLA fine-tuning makes the model generate bare numeric action tokens instead of staying close to the VLM’s pretrained language distribution, which can degrade learned representations for generalization.\"},{\"question\":\"How does CLAP reformulate action prediction?\",\"answer\":\"CLAP generates a natural-language action description first, then causally conditions the subsequent numeric action-token prediction on that language-action plan in a single autoregressive sequence.\"},{\"question\":\"What results does CLAP achieve after single-epoch fine-tuning?\",\"answer\":\"With single-epoch fine-tuning, the 2B CLAP model achieves 90.8% success on LIBERO, outperforming VLA-0 by +14.9 percentage points, and improves robustness on LIBEROPRO under language, object, and spatial perturbations.\"}]",1784178327,45,{"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},"clap-direct-vlm-to-vla-adaptation-via-language-action-grounding","",{"@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/clap-direct-vlm-to-vla-adaptation-via-language-action-grounding/82122/",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","2026-07-17","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},"What main problem does CLAP address in adapting VLMs to VLAs?","Question",{"text":75,"@type":76},"CLAP targets output-distribution mismatch: standard VLA fine-tuning makes the model generate bare numeric action tokens instead of staying close to the VLM’s pretrained language distribution, which can degrade learned representations for generalization.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does CLAP reformulate action prediction?",{"text":80,"@type":76},"CLAP generates a natural-language action description first, then causally conditions the subsequent numeric action-token prediction on that language-action plan in a single autoregressive sequence.",{"name":82,"@type":73,"acceptedAnswer":83},"What results does CLAP achieve after single-epoch fine-tuning?",{"text":84,"@type":76},"With single-epoch fine-tuning, the 2B CLAP model achieves 90.8% success on LIBERO, outperforming VLA-0 by +14.9 percentage points, and improves robustness on LIBEROPRO under language, object, and spatial perturbations.","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"]