DO-ALL is a plug-and-play module for Continual Test-Time Adaptation (CTTA) that revisits source information through Dataset Distillation (DD).</p>\n<p>Before deployment, DO-ALL distills the source dataset into a compact set of synthetic anchors. During adaptation, each target sample is matched to its nearest anchor, which guides the update via anchor replay, feature alignment, and harm-adaptive blending that rewinds unstable parameter groups toward the source model. DO-ALL plugs into any base CTTA method without changing its objective.</p>\n","updatedAt":"2026-06-25T10:25:54.086Z","author":{"_id":"67f514dd328e736dd0bc10f6","avatarUrl":"/avatars/7ccc2754318ca2ff274d65af4f6f9c20.svg","fullname":"Hyun-Kurl Jang","name":"blue531","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8579593896865845},"editors":["blue531"],"editorAvatarUrls":["/avatars/7ccc2754318ca2ff274d65af4f6f9c20.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.20196","authors":[{"_id":"6a3b39f00a86ac3098d5d640","user":{"_id":"67f514dd328e736dd0bc10f6","avatarUrl":"/avatars/7ccc2754318ca2ff274d65af4f6f9c20.svg","isPro":false,"fullname":"Hyun-Kurl Jang","user":"blue531","type":"user","name":"blue531"},"name":"Hyun-Kurl Jang","status":"claimed_verified","statusLastChangedAt":"2026-06-25T09:31:02.291Z","hidden":false},{"_id":"6a3b39f00a86ac3098d5d641","name":"Jihun Kim","hidden":false},{"_id":"6a3b39f00a86ac3098d5d642","name":"Hyeokjun Kweon","hidden":false},{"_id":"6a3b39f00a86ac3098d5d643","name":"Kuk-Jin Yoon","hidden":false}],"publishedAt":"2026-06-18T00:00:00.000Z","submittedOnDailyAt":"2026-06-25T00:00:00.000Z","title":"Distill Once, Adapt Life-Long: Exploring Dataset Distillation for Continual Test-Time Adaptation","submittedOnDailyBy":{"_id":"67f514dd328e736dd0bc10f6","avatarUrl":"/avatars/7ccc2754318ca2ff274d65af4f6f9c20.svg","isPro":false,"fullname":"Hyun-Kurl Jang","user":"blue531","type":"user","name":"blue531"},"summary":"Continual Test-Time Adaptation (CTTA) aims to maintain model performance under evolving target domains by adapting online without labeled data. However, practical deployments often cannot retain the source dataset due to privacy or licensing constraints, and purely source-free CTTA methods tend to become unstable under long-term distribution shift, suffering from compounding self-training errors and catastrophic forgetting. We introduce DO-ALL (Distill Once, Adapt Life-Long), a plug-and-play framework that revisits source information in a compact and privacy-conscious form via Dataset Distillation (DD). Before deployment, DO-ALL performs DD to produce a small set of synthetic distilled anchors that summarize the source distribution. During adaptation, each target sample is matched with its most semantically aligned anchor, which provides a stable reference for various CTTA via source replay, representation alignment, and manifold-smoothing regularization. DO-ALL can be seamlessly integrated into existing CTTA algorithms, consistently improving long-term robustness across CIFAR100-C, ImageNet-C, and the CCC benchmark. This demonstrates the potential of leveraging DD to enable stable and continuous adaptation without retaining raw source data. The code is available at https://github.com/blue-531/DOALL.","upvotes":1,"discussionId":"6a3b39f00a86ac3098d5d644","githubRepo":"https://github.com/blue-531/DOALL","githubRepoAddedBy":"user","ai_summary":"DO-ALL is a test-time adaptation framework that uses dataset distillation to create synthetic anchors for stable long-term model performance without retaining source data.","ai_keywords":["Continual Test-Time Adaptation","Dataset Distillation","source-free adaptation","catastrophic forgetting","semantic alignment","source replay","representation alignment","manifold-smoothing regularization","CIFAR100-C","ImageNet-C","CCC benchmark"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":2},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6a04484dfa6be2fefe2c0c6a","avatarUrl":"/avatars/3eb7713635c72623aa9963564e75be67.svg","isPro":false,"fullname":"Todd Harris","user":"ToddHarris","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.20196.md","query":{}}">
Distill Once, Adapt Life-Long: Exploring Dataset Distillation for Continual Test-Time Adaptation
Abstract
DO-ALL is a test-time adaptation framework that uses dataset distillation to create synthetic anchors for stable long-term model performance without retaining source data.
Continual Test-Time Adaptation (CTTA) aims to maintain model performance under evolving target domains by adapting online without labeled data. However, practical deployments often cannot retain the source dataset due to privacy or licensing constraints, and purely source-free CTTA methods tend to become unstable under long-term distribution shift, suffering from compounding self-training errors and catastrophic forgetting. We introduce DO-ALL (Distill Once, Adapt Life-Long), a plug-and-play framework that revisits source information in a compact and privacy-conscious form via Dataset Distillation (DD). Before deployment, DO-ALL performs DD to produce a small set of synthetic distilled anchors that summarize the source distribution. During adaptation, each target sample is matched with its most semantically aligned anchor, which provides a stable reference for various CTTA via source replay, representation alignment, and manifold-smoothing regularization. DO-ALL can be seamlessly integrated into existing CTTA algorithms, consistently improving long-term robustness across CIFAR100-C, ImageNet-C, and the CCC benchmark. This demonstrates the potential of leveraging DD to enable stable and continuous adaptation without retaining raw source data. The code is available at https://github.com/blue-531/DOALL.
Community
DO-ALL is a plug-and-play module for Continual Test-Time Adaptation (CTTA) that revisits source information through Dataset Distillation (DD).
Before deployment, DO-ALL distills the source dataset into a compact set of synthetic anchors. During adaptation, each target sample is matched to its nearest anchor, which guides the update via anchor replay, feature alignment, and harm-adaptive blending that rewinds unstable parameter groups toward the source model. DO-ALL plugs into any base CTTA method without changing its objective.
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2606.20196 in a model README.md to link it from this page.
Cite arxiv.org/abs/2606.20196 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2606.20196 in a Space README.md 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.