Hugging Face Daily Papers · · 3 min read

DataClaw0: Agentic Tailoring Multimodal Data from Raw Streams

Mirrored from Hugging Face Daily Papers for archival readability. Support the source by reading on the original site.

Make any data you want</p>\n","updatedAt":"2026-06-23T05:51:32.352Z","author":{"_id":"66d7e9519b7da501cd2ceeaa","avatarUrl":"/avatars/f23c207ef9264851a4cdefb000b22fc4.svg","fullname":"Cong","name":"Coneonewan","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8247194290161133},"editors":["Coneonewan"],"editorAvatarUrls":["/avatars/f23c207ef9264851a4cdefb000b22fc4.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.21337","authors":[{"_id":"6a3a16b3fdcd3514343bb6a3","user":{"_id":"66d7e9519b7da501cd2ceeaa","avatarUrl":"/avatars/f23c207ef9264851a4cdefb000b22fc4.svg","isPro":false,"fullname":"Cong","user":"Coneonewan","type":"user","name":"Coneonewan"},"name":"Cong Wan","status":"claimed_verified","statusLastChangedAt":"2026-06-23T13:56:05.018Z","hidden":false},{"_id":"6a3a16b3fdcd3514343bb6a4","name":"Zeyu Guo","hidden":false},{"_id":"6a3a16b3fdcd3514343bb6a5","name":"Zijian Cai","hidden":false},{"_id":"6a3a16b3fdcd3514343bb6a6","name":"Jiangyang Li","hidden":false},{"_id":"6a3a16b3fdcd3514343bb6a7","name":"SongLin Dong","hidden":false},{"_id":"6a3a16b3fdcd3514343bb6a8","name":"Lin Peng","hidden":false},{"_id":"6a3a16b3fdcd3514343bb6a9","user":{"_id":"668df98de9e585e8718f767f","avatarUrl":"/avatars/2be52f4ae88a0991c8ae584f8e870734.svg","isPro":false,"fullname":"Xiangyang Luo","user":"XiangyangLuo02","type":"user","name":"XiangyangLuo02"},"name":"Xiangyang Luo","status":"claimed_verified","statusLastChangedAt":"2026-06-23T13:56:01.079Z","hidden":false},{"_id":"6a3a16b3fdcd3514343bb6aa","name":"Zhiheng Ma","hidden":false},{"_id":"6a3a16b3fdcd3514343bb6ab","name":"Yihong Gong","hidden":false}],"publishedAt":"2026-06-19T00:00:00.000Z","submittedOnDailyAt":"2026-06-23T00:00:00.000Z","title":"DataClaw0: Agentic Tailoring Multimodal Data from Raw Streams","submittedOnDailyBy":{"_id":"66d7e9519b7da501cd2ceeaa","avatarUrl":"/avatars/f23c207ef9264851a4cdefb000b22fc4.svg","isPro":false,"fullname":"Cong","user":"Coneonewan","type":"user","name":"Coneonewan"},"summary":"Massive unstructured multimodal streams suffer from high \"data entropy,\" impeding both efficient human knowledge acquisition and high-quality AI post-training. Existing passive annotation paradigms, heavily reliant on heuristic rules or general VLMs, are costly, monotonous, and fail to unlock the deep procedural logic embedded in raw data. We elevate data processing to a learnable capability, proposing a paradigm shift towards Agentic Data Tailoring, which actively refining and structuring data to align with diverse user and downstream intents. To overcome the data scarcity bottleneck in training such high-order capabilities, we design a two-stage pipeline grounding generative semantic synthesis in deterministic Factual Anchors, yielding a large-scale dataset spanning five core physical and digital domains. Building upon this, DataClaw_0-9B model synergizes Supervised Fine-Tuning (SFT) with Group Relative Policy Optimization (GRPO), achieving robust alignment with complex refinement and tailoring intents. To systematically quantify this capability, we construct DataClaw_0-val, the first benchmark dedicated to data refinement. Crucially, we adopt downstream post-training as the ultimate validation touchstone. Evaluations on video generation, real-world VQA, and GUI navigation confirm that DataClaw_0 delivers high-information-density tailored data, facilitating efficient model adaptation to new tasks under limited training data regimes. Project page: https://czjdsg.github.io/MakeAnyData","upvotes":65,"discussionId":"6a3a16b3fdcd3514343bb6ac","projectPage":"https://czjdsg.github.io/MakeAnyData/#cases","githubRepo":"https://github.com/vancyland/DataClaw0","githubRepoAddedBy":"user","ai_summary":"Agentic Data Tailoring paradigm uses learnable data processing to structure high-entropy multimodal streams, with DataClaw_0-9B model achieving robust alignment through SFT and GRPO on a novel benchmark.","ai_keywords":["Agentic Data Tailoring","generative semantic synthesis","deterministic Factual Anchors","Supervised Fine-Tuning","Group Relative Policy Optimization","data refinement","multimodal streams","data entropy","downstream post-training"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":24},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6a146ff99c5e89e04b70e4b9","avatarUrl":"/avatars/3a434e78443a2005dd4c7101b8e159cd.svg","isPro":false,"fullname":"Zheng Yifan","user":"zhengyifan20","type":"user"},{"_id":"6442c7d97feb866811b2935c","avatarUrl":"/avatars/16c18db3b33785203c3023307fcc547e.svg","isPro":false,"fullname":"Zijian Cai","user":"yzxjb","type":"user"},{"_id":"668df98de9e585e8718f767f","avatarUrl":"/avatars/2be52f4ae88a0991c8ae584f8e870734.svg","isPro":false,"fullname":"Xiangyang Luo","user":"XiangyangLuo02","type":"user"},{"_id":"6a15cb24785864c80e393b2e","avatarUrl":"/avatars/3f9f78771acca608e769a7653c4584d5.svg","isPro":false,"fullname":"山下愛","user":"heanderson","type":"user"},{"_id":"6a15e8bdf28e1a7f4f2b4061","avatarUrl":"/avatars/06e1fb99c1f7421938c558569f613962.svg","isPro":false,"fullname":"罗子轩","user":"harperclark","type":"user"},{"_id":"6a1591afe68cf1a41e7c0bce","avatarUrl":"/avatars/5f2930a52af392950df761b06c4eade9.svg","isPro":false,"fullname":"優奈 阿部","user":"asher-garciahb","type":"user"},{"_id":"6a1470b7b28ec6a2ad92193c","avatarUrl":"/avatars/6243b0a5d5d3c8bec740d07bdbb12950.svg","isPro":false,"fullname":"Grayson Scott","user":"grascott99","type":"user"},{"_id":"6a14801df432ac02b881e091","avatarUrl":"/avatars/0cf3308977f6066b02e083598cb4726f.svg","isPro":false,"fullname":"Julian Perez","user":"julianpere32","type":"user"},{"_id":"6a147f9828c29f18c7dca547","avatarUrl":"/avatars/a0b42a9bda610a7fbc951085332c6dcb.svg","isPro":false,"fullname":"木村結月","user":"ameliarobinson","type":"user"},{"_id":"6a15c469d57ab19bdd02eb7d","avatarUrl":"/avatars/c576b96ba3190641d74ac85e938509f1.svg","isPro":false,"fullname":"佐藤颯太","user":"mateom8","type":"user"},{"_id":"6a15ed18c2fc8c23990d384e","avatarUrl":"/avatars/dd363f08ed49f8b0d4c7699391263b10.svg","isPro":false,"fullname":"加藤 大輝","user":"wyatth34","type":"user"},{"_id":"6a147d3193b9759a7f062980","avatarUrl":"/avatars/167f360f649939b767745a7b30e2ac7a.svg","isPro":false,"fullname":"Aurora Hill","user":"aurorahi8","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":3,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.21337.md","query":{}}">
Papers
arxiv:2606.21337

DataClaw0: Agentic Tailoring Multimodal Data from Raw Streams

Published on Jun 19
· Submitted by
Cong
on Jun 23
#3 Paper of the day
Authors:
,
,
,
,
,
,

Abstract

Agentic Data Tailoring paradigm uses learnable data processing to structure high-entropy multimodal streams, with DataClaw_0-9B model achieving robust alignment through SFT and GRPO on a novel benchmark.

Massive unstructured multimodal streams suffer from high "data entropy," impeding both efficient human knowledge acquisition and high-quality AI post-training. Existing passive annotation paradigms, heavily reliant on heuristic rules or general VLMs, are costly, monotonous, and fail to unlock the deep procedural logic embedded in raw data. We elevate data processing to a learnable capability, proposing a paradigm shift towards Agentic Data Tailoring, which actively refining and structuring data to align with diverse user and downstream intents. To overcome the data scarcity bottleneck in training such high-order capabilities, we design a two-stage pipeline grounding generative semantic synthesis in deterministic Factual Anchors, yielding a large-scale dataset spanning five core physical and digital domains. Building upon this, DataClaw_0-9B model synergizes Supervised Fine-Tuning (SFT) with Group Relative Policy Optimization (GRPO), achieving robust alignment with complex refinement and tailoring intents. To systematically quantify this capability, we construct DataClaw_0-val, the first benchmark dedicated to data refinement. Crucially, we adopt downstream post-training as the ultimate validation touchstone. Evaluations on video generation, real-world VQA, and GUI navigation confirm that DataClaw_0 delivers high-information-density tailored data, facilitating efficient model adaptation to new tasks under limited training data regimes. Project page: https://czjdsg.github.io/MakeAnyData

Community

Paper author Paper submitter about 19 hours ago

Make any data you want

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images

· Sign up or log in to comment

Get this paper in your agent:

hf papers read 2606.21337
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2606.21337 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/2606.21337 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/2606.21337 in a Space README.md to link it from this page.

Collections including this paper 1

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.

More from Hugging Face Daily Papers