Check out our demo at <a href=\"https://vaynexie.github.io/insight-doc-demo-display/demo_display.html\" rel=\"nofollow\">https://vaynexie.github.io/insight-doc-demo-display/demo_display.html</a> (🎬 replay) and <a href=\"https://huggingface.co/spaces/leoyu112211/insight-doc-online-demo\">https://huggingface.co/spaces/leoyu112211/insight-doc-online-demo</a> (🚀 live) !</p>\n","updatedAt":"2026-08-12T09:40:04.046Z","author":{"_id":"65d5b967eeb590ea7435ad07","avatarUrl":"/avatars/ca0a5e123d5da5aca97cfd8a2d07e60e.svg","fullname":"Kaican Li","name":"m-Just","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":3,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.6782253384590149},"editors":["m-Just"],"editorAvatarUrls":["/avatars/ca0a5e123d5da5aca97cfd8a2d07e60e.svg"],"reactions":[{"reaction":"🔥","users":["m-Just"],"count":1}],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.10628","authors":[{"_id":"6a7c2f891653ef87c6af1df6","user":{"_id":"65d5b967eeb590ea7435ad07","avatarUrl":"/avatars/ca0a5e123d5da5aca97cfd8a2d07e60e.svg","isPro":true,"fullname":"Kaican Li","user":"m-Just","type":"user","name":"m-Just"},"name":"Kaican Li","status":"claimed_verified","statusLastChangedAt":"2026-08-12T16:45:05.170Z","hidden":false},{"_id":"6a7c2f891653ef87c6af1df7","name":"Weiyan Xie","hidden":false},{"_id":"6a7c2f891653ef87c6af1df8","name":"Lewei Yao","hidden":false},{"_id":"6a7c2f891653ef87c6af1df9","name":"Jiannan Wu","hidden":false},{"_id":"6a7c2f891653ef87c6af1dfa","name":"Lanqing Hong","hidden":false},{"_id":"6a7c2f891653ef87c6af1dfb","name":"Yongxiang Huang","hidden":false},{"_id":"6a7c2f891653ef87c6af1dfc","name":"Nevin L. Zhang","hidden":false}],"publishedAt":"2026-08-11T00:00:00.000Z","submittedOnDailyAt":"2026-08-12T00:00:00.000Z","title":"InSight-doc: Agentic Visual Perception for Long-Document Understanding","submittedOnDailyBy":{"_id":"65d5b967eeb590ea7435ad07","avatarUrl":"/avatars/ca0a5e123d5da5aca97cfd8a2d07e60e.svg","isPro":true,"fullname":"Kaican Li","user":"m-Just","type":"user","name":"m-Just"},"summary":"Long-document understanding often requires reasoning over many visually rich pages, making inference costly and prone to context rot. In this work, we propose InSight-doc, an agentic visual perception framework that treats visual resolution as an adaptive reasoning-time resource. InSight-doc starts from low resolution and selectively zooms into high-resolution regions for finer evidence, without relying on any external retriever. To train such an agent, we construct an active-perception corpus of 17.9K high-quality SFT examples with region-level zoom-in trajectories, accompanied by 19.2K hard RL examples. Through SFT+RL, InSight-doc-8B improves the baseline by 4.3--16.4 accuracy points over document VQA benchmarks. On long documents, it reduces hallucination by more than 40% and inference latency by 41%--68% while maintaining an accuracy lead. Our code, datasets, and model are released at https://github.com/m-Just/InSight-doc .","upvotes":5,"discussionId":"6a7c2f8a1653ef87c6af1dfd","projectPage":"https://github.com/m-Just/InSight-doc","githubRepo":"https://github.com/m-Just/InSight-doc","githubRepoAddedBy":"user","ai_summary":"InSight-doc adaptively allocates visual resolution during reasoning to improve long-document understanding while reducing latency and hallucinations.","ai_keywords":["agentic visual perception","adaptive reasoning-time resource","region-level zoom-in","active-perception corpus","SFT","RL","document VQA","hallucination","inference latency"],"ai_summary_model":"thinkingmachines/Inkling-Small","githubStars":4},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"65d5b967eeb590ea7435ad07","avatarUrl":"/avatars/ca0a5e123d5da5aca97cfd8a2d07e60e.svg","isPro":true,"fullname":"Kaican Li","user":"m-Just","type":"user"},{"_id":"69fd5c5a154c7dfe3c9c799f","avatarUrl":"/avatars/95f7f9c1312d7789c1dcfdecebce2dce.svg","isPro":false,"fullname":"Jiannan Wu","user":"wjn922-01","type":"user"},{"_id":"670382b1bf8ff45a4954939d","avatarUrl":"/avatars/f86118b2dcf9f4a9520cef9de25aceae.svg","isPro":true,"fullname":"leoyu","user":"leoyu112211","type":"user"},{"_id":"6918d79225d42ee7e6741161","avatarUrl":"/avatars/9958442e0988ae40da059e4fd90f4d6c.svg","isPro":false,"fullname":"InSight-o3","user":"InSight-o3","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,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.10628.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?