Deep-research agents repeatedly reopen and prefill the same raw documents. Agentic Data Cracking lets agents speculate about useful structures and prefetch reusable views during reasoning. By materializing these structures we cut cost by 53% on FanOutQA and by 3× in a harder multi-hop study while revealing over an order of magnitude of remaining headroom.</p>\n","updatedAt":"2026-09-02T17:59:31.399Z","author":{"_id":"6787110f4d9e0e1baf11954b","avatarUrl":"/avatars/3ed4beebfa7b9378b18a227d9ad0119f.svg","fullname":"Milad Rezaei Hajidehi","name":"miladrzh","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9179239869117737},"editors":["miladrzh"],"editorAvatarUrls":["/avatars/3ed4beebfa7b9378b18a227d9ad0119f.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.31082","authors":[{"_id":"6a970a30fe3c2f89286c3706","user":{"_id":"6787110f4d9e0e1baf11954b","avatarUrl":"/avatars/3ed4beebfa7b9378b18a227d9ad0119f.svg","isPro":false,"fullname":"Milad Rezaei Hajidehi","user":"miladrzh","type":"user","name":"miladrzh"},"name":"Milad Rezaei Hajidehi","status":"claimed_verified","statusLastChangedAt":"2026-09-02T08:45:05.038Z","hidden":false},{"_id":"6a970a30fe3c2f89286c3707","name":"Qitong Wang","hidden":false},{"_id":"6a970a30fe3c2f89286c3708","name":"Stratos Idreos","hidden":false}],"publishedAt":"2026-08-31T00:00:00.000Z","submittedOnDailyAt":"2026-09-02T00:00:00.000Z","title":"Token-Efficient Data Reasoning Agents via Adaptive Structuring of Unstructured Data","submittedOnDailyBy":{"_id":"6787110f4d9e0e1baf11954b","avatarUrl":"/avatars/3ed4beebfa7b9378b18a227d9ad0119f.svg","isPro":false,"fullname":"Milad Rezaei Hajidehi","user":"miladrzh","type":"user","name":"miladrzh"},"summary":"Valuable data remains embedded in unstructured sources: web pages, reports, contracts, filings, earnings calls, and PDFs. The big bet in enterprise AI is deploying LLM agents that reason over this data to answer complex questions for every knowledge worker. Agents can do this today, but at prohibitive cost. Each question repeatedly opens large documents to recover scattered evidence, consuming up to a million tokens. However, if the data were already structured, the same question would reduce to a cheap database lookup. For example, on FanOutQA benchmark, reasoning over an ideal pre-structured store is 28X cheaper, and the gap grows to orders of magnitude as questions fan out over more documents. Yet structuring everything in advance is not viable: documents hold vastly more possible structure than any workload will use, and the useful structure and documents are unknown until queries arrive. We propose agentic data cracking, a method that structures unstructured data adaptively and speculatively as a byproduct of reasoning itself. Structuring is adaptive because observed queries decide when it happens and what matters, and speculative because it goes beyond the current question. Whenever the agent opens a document to answer, a cracking sub-agent forks from the already-loaded context at marginal cost and extracts grounded structure likely to serve related future queries. Over time, an increasing share of queries is fully covered by structured data and answered without opening a document, keeping agentic accuracy at close to RAG cost. On FanOutQA, extended with merely one related question per test question, cracking cuts cost by 53% while preserving accuracy. Agentic data cracking is a first step toward next-generation data infrastructure for agentic reasoning over unstructured data: a shared substrate beneath the model where knowledge that reasoning already paid to uncover accumulates.","upvotes":2,"discussionId":"6a970a30fe3c2f89286c3709","organization":{"_id":"63663038361a96184dbad334","name":"Harvard","fullname":"Harvard University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1667641389305-6366110f575c93cedafde54e.jpeg"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6a92ac7ef144f21950ef68ed","avatarUrl":"/avatars/f15849014b2f01f0d212d1d28bfe49c3.svg","isPro":false,"fullname":"Milad, Make It Work","user":"miladmakeitwork","type":"user"},{"_id":"63ac5701c21e60a3e9b58aa7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63ac5701c21e60a3e9b58aa7/g6EX7diOpuA94R2ab-rZC.png","isPro":true,"fullname":"Dipankar Sarkar","user":"dipankarsarkar","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"63663038361a96184dbad334","name":"Harvard","fullname":"Harvard University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1667641389305-6366110f575c93cedafde54e.jpeg"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.31082.md","query":{}}">
Token-Efficient Data Reasoning Agents via Adaptive Structuring of Unstructured Data
Abstract
Valuable data remains embedded in unstructured sources: web pages, reports, contracts, filings, earnings calls, and PDFs. The big bet in enterprise AI is deploying LLM agents that reason over this data to answer complex questions for every knowledge worker. Agents can do this today, but at prohibitive cost. Each question repeatedly opens large documents to recover scattered evidence, consuming up to a million tokens. However, if the data were already structured, the same question would reduce to a cheap database lookup. For example, on FanOutQA benchmark, reasoning over an ideal pre-structured store is 28X cheaper, and the gap grows to orders of magnitude as questions fan out over more documents. Yet structuring everything in advance is not viable: documents hold vastly more possible structure than any workload will use, and the useful structure and documents are unknown until queries arrive. We propose agentic data cracking, a method that structures unstructured data adaptively and speculatively as a byproduct of reasoning itself. Structuring is adaptive because observed queries decide when it happens and what matters, and speculative because it goes beyond the current question. Whenever the agent opens a document to answer, a cracking sub-agent forks from the already-loaded context at marginal cost and extracts grounded structure likely to serve related future queries. Over time, an increasing share of queries is fully covered by structured data and answered without opening a document, keeping agentic accuracy at close to RAG cost. On FanOutQA, extended with merely one related question per test question, cracking cuts cost by 53% while preserving accuracy. Agentic data cracking is a first step toward next-generation data infrastructure for agentic reasoning over unstructured data: a shared substrate beneath the model where knowledge that reasoning already paid to uncover accumulates.
Community
Deep-research agents repeatedly reopen and prefill the same raw documents. Agentic Data Cracking lets agents speculate about useful structures and prefetch reusable views during reasoning. By materializing these structures we cut cost by 53% on FanOutQA and by 3× in a harder multi-hop study while revealing over an order of magnitude of remaining headroom.
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/2608.31082 in a model README.md to link it from this page.
Cite arxiv.org/abs/2608.31082 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2608.31082 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.