Domain Arithmetic: One-Shot VLA Adaptation under Environmental Shifts
Mirrored from Hugging Face Daily Papers for archival readability. Support the source by reading on the original site.
Accepted at ECCV 2026.<br>Domain Arithmetic (DART) adapts multi-task VLAs to environmental shifts (e.g., camera-pose changes, embodiment changes) using a single demo of a single task through subspace-aligned weight arithmetic.</p>\n","updatedAt":"2026-07-02T02:37:44.342Z","author":{"_id":"654e14776337c78f3c72e680","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/654e14776337c78f3c72e680/As2L8hOQjUzgjFaMxDcKo.jpeg","fullname":"Taewook Kang","name":"twkang43","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":1,"identifiedLanguage":{"language":"en","probability":0.8323928713798523},"editors":["twkang43"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/654e14776337c78f3c72e680/As2L8hOQjUzgjFaMxDcKo.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2607.00666","authors":[{"_id":"6a45cccc4f1dd35e48fb8eed","name":"Taewook Kang","hidden":false},{"_id":"6a45cccc4f1dd35e48fb8eee","name":"Taeheon Kim","hidden":false},{"_id":"6a45cccc4f1dd35e48fb8eef","name":"Donghyun Shin","hidden":false},{"_id":"6a45cccc4f1dd35e48fb8ef0","name":"Jonghyun Choi","hidden":false}],"publishedAt":"2026-07-01T00:00:00.000Z","submittedOnDailyAt":"2026-07-02T00:00:00.000Z","title":"Domain Arithmetic: One-Shot VLA Adaptation under Environmental Shifts","submittedOnDailyBy":{"_id":"654e14776337c78f3c72e680","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/654e14776337c78f3c72e680/As2L8hOQjUzgjFaMxDcKo.jpeg","isPro":false,"fullname":"Taewook Kang","user":"twkang43","type":"user","name":"twkang43"},"summary":"Vision-Language-Action (VLA) models often fail to perform the same learned tasks under environmental shifts, such as changes in camera pose and shifts to a different but similar robot (e.g., from Panda to UR5e). Adapting these models to the shifted environment (i.e., target domain) often requires training on multiple demonstrations for each task, which are costly to collect. To reduce the burden of data curation and training, we propose an analogy-based method that adapts VLA models under environmental shifts through weight vector arithmetic with domain-specific information addition, named Domain ARiThmetic (DART). Unlike prior approaches, DART requires collecting only a single demonstration, enabling efficient adaptation. To accurately isolate domain-specific information for addition, DART performs subspace alignment between singular components in weight vectors to filter out noisy components. In both simulated and real-world experiments, DART outperforms existing VLA adaptation methods in one-shot scenarios across diverse visual and embodiment shifts. Code is available at https://github.com/snumprlab/dart.","upvotes":16,"discussionId":"6a45cccc4f1dd35e48fb8ef1","projectPage":"https://twkang43.github.io/projects/dart/","githubRepo":"https://github.com/snumprlab/dart","githubRepoAddedBy":"user","ai_summary":"Vision-Language-Action models can be efficiently adapted to new environments using a single demonstration through weight vector arithmetic that isolates domain-specific information via subspace alignment.","ai_keywords":["Vision-Language-Action models","domain-specific information","weight vector arithmetic","subspace alignment","one-shot adaptation","environmental shifts","embodiment shifts"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":10,"organization":{"_id":"66d54dc8033492801db2bf5a","name":"SeoulNatlUniv","fullname":"Seoul National University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/659ccc9d18897eb6594e897f/_-0BM-1UyM-d-lRiahFnf.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"654e14776337c78f3c72e680","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/654e14776337c78f3c72e680/As2L8hOQjUzgjFaMxDcKo.jpeg","isPro":false,"fullname":"Taewook Kang","user":"twkang43","type":"user"},{"_id":"680dc2ea9f64391e4188d974","avatarUrl":"/avatars/3febf1501623d75ab01dddbd998224a1.svg","isPro":false,"fullname":"Jae Seok Heo","user":"J13947","type":"user"},{"_id":"6757a721f498e0436cc77055","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/GBxQKQ7E7FmfrIS6VQhJJ.png","isPro":false,"fullname":"Taeheon Kim","user":"thkim0305-SNU","type":"user"},{"_id":"663dd021e2910f13f96fb63c","avatarUrl":"/avatars/d057bcef274391691c83f20390c9dc5b.svg","isPro":false,"fullname":"Jimin Ryu","user":"nick11967","type":"user"},{"_id":"6690bf4a93ab304d32ab7464","avatarUrl":"/avatars/25ce5a6d3daafabfb0d9a6a17665cd98.svg","isPro":false,"fullname":"Reokyoung Kim","user":"furud","type":"user"},{"_id":"6753226bde07a89d0ceb9162","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/QwijcIGUT6FITZQ_-5Pk2.png","isPro":false,"fullname":"youhan lee","user":"youhan-lee","type":"user"},{"_id":"675288db99b478caa10a95e7","avatarUrl":"/avatars/e3c63a61b324a21b1853e1def2915910.svg","isPro":false,"fullname":"Hyeonbeom Choi","user":"violet-blue","type":"user"},{"_id":"6683fc5344a65be1aab25dc0","avatarUrl":"/avatars/e13cde3f87b59e418838d702807df3b5.svg","isPro":false,"fullname":"hjkim","user":"hojie11","type":"user"},{"_id":"668defa6cbb3c5e08b51401e","avatarUrl":"/avatars/47a35273d5477c48f0b68c3927a2738b.svg","isPro":false,"fullname":"Kim san","user":"mounKim","type":"user"},{"_id":"65b744d605c25412bb7f08b8","avatarUrl":"/avatars/5a1c6632bd4320b882b25ff8aa6f2fd8.svg","isPro":false,"fullname":"Seungyeon Jwa","user":"syhuggingface","type":"user"},{"_id":"6a2da6c8ca070ee12c6e396c","avatarUrl":"/avatars/0355287dcabaa67dbc7f0b10b87451f9.svg","isPro":false,"fullname":"Joe Mama","user":"JoeMama123123123","type":"user"},{"_id":"631c386bc73939ffc0716a37","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1662793811119-noauth.jpeg","isPro":false,"fullname":"SeongWan Kim","user":"idgmatrix","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"66d54dc8033492801db2bf5a","name":"SeoulNatlUniv","fullname":"Seoul National University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/659ccc9d18897eb6594e897f/_-0BM-1UyM-d-lRiahFnf.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2607/2607.00666.md","query":{}}">
Papers
arxiv:2607.00666
Domain Arithmetic: One-Shot VLA Adaptation under Environmental Shifts
Authors:
,
,
,
Abstract
Vision-Language-Action models can be efficiently adapted to new environments using a single demonstration through weight vector arithmetic that isolates domain-specific information via subspace alignment.
Generated by Qwen/Qwen2.5-Coder-32B-Instruct
Vision-Language-Action (VLA) models often fail to perform the same learned tasks under environmental shifts, such as changes in camera pose and shifts to a different but similar robot (e.g., from Panda to UR5e). Adapting these models to the shifted environment (i.e., target domain) often requires training on multiple demonstrations for each task, which are costly to collect. To reduce the burden of data curation and training, we propose an analogy-based method that adapts VLA models under environmental shifts through weight vector arithmetic with domain-specific information addition, named Domain ARiThmetic (DART). Unlike prior approaches, DART requires collecting only a single demonstration, enabling efficient adaptation. To accurately isolate domain-specific information for addition, DART performs subspace alignment between singular components in weight vectors to filter out noisy components. In both simulated and real-world experiments, DART outperforms existing VLA adaptation methods in one-shot scenarios across diverse visual and embodiment shifts. Code is available at https://github.com/snumprlab/dart.
Community
Accepted at ECCV 2026.
Domain Arithmetic (DART) adapts multi-task VLAs to environmental shifts (e.g., camera-pose changes, embodiment changes) using a single demo of a single task through subspace-aligned weight arithmetic.
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Get this paper in your agent:
hf papers read 2607.00666 curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 0
No model linking this paper
Cite arxiv.org/abs/2607.00666 in a model README.md to link it from this page.
Datasets citing this paper 0
No dataset linking this paper
Cite arxiv.org/abs/2607.00666 in a dataset README.md to link it from this page.
Spaces citing this paper 0
No Space linking this paper
Cite arxiv.org/abs/2607.00666 in a Space README.md to link it from this page.
Collections including this paper 0
No Collection including this paper
Add this paper to a collection to link it from this page.
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.