🤗 Data: <a href=\"https://huggingface.co/datasets/Zhengbo-Zhang/Visual-Seeker-train-data\">https://huggingface.co/datasets/Zhengbo-Zhang/Visual-Seeker-train-data</a><br>💻 Code: <a href=\"https://github.com/ZhengboZhang/Visual-Seeker\" rel=\"nofollow\">https://github.com/ZhengboZhang/Visual-Seeker</a><br>📄Paper: <a href=\"https://arxiv.org/abs/2606.15231\" rel=\"nofollow\">https://arxiv.org/abs/2606.15231</a></p>\n","updatedAt":"2026-06-17T11:55:42.439Z","author":{"_id":"671906d1827bea3146f0b457","avatarUrl":"/avatars/4cf0b766143768e58f3cc41673f77cb1.svg","fullname":"Bo","name":"Zhengbo-Zhang","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.5004693865776062},"editors":["Zhengbo-Zhang"],"editorAvatarUrls":["/avatars/4cf0b766143768e58f3cc41673f77cb1.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.15231","authors":[{"_id":"6a328a3859127a45e2c1c379","name":"Zhengbo Zhang","hidden":false},{"_id":"6a328a3859127a45e2c1c37a","name":"Changtao Miao","hidden":false},{"_id":"6a328a3859127a45e2c1c37b","name":"Jinbo Su","hidden":false},{"_id":"6a328a3859127a45e2c1c37c","name":"Zhaowen Zhou","hidden":false},{"_id":"6a328a3859127a45e2c1c37d","name":"Chunxia Zhang","hidden":false},{"_id":"6a328a3859127a45e2c1c37e","name":"Xukai Wang","hidden":false},{"_id":"6a328a3859127a45e2c1c37f","name":"Ruiqi Liu","hidden":false},{"_id":"6a328a3859127a45e2c1c380","name":"Kaiyuan Zheng","hidden":false},{"_id":"6a328a3859127a45e2c1c381","name":"Jiansheng Cai","hidden":false},{"_id":"6a328a3859127a45e2c1c382","name":"Bo Zhang","hidden":false},{"_id":"6a328a3859127a45e2c1c383","name":"Zhe Li","hidden":false},{"_id":"6a328a3859127a45e2c1c384","name":"Shiming Xiang","hidden":false},{"_id":"6a328a3859127a45e2c1c385","name":"Ying Yan","hidden":false}],"publishedAt":"2026-06-13T00:00:00.000Z","submittedOnDailyAt":"2026-06-17T00:00:00.000Z","title":"Visual-Seeker: Towards Visual-Native Multimodal Agentic Search via Active Visual Reasoning","submittedOnDailyBy":{"_id":"671906d1827bea3146f0b457","avatarUrl":"/avatars/4cf0b766143768e58f3cc41673f77cb1.svg","isPro":false,"fullname":"Bo","user":"Zhengbo-Zhang","type":"user","name":"Zhengbo-Zhang"},"summary":"Multimodal large language models (MLLMs) have demonstrated impressive capabilities in many visual tasks, but they often struggle with factual grounding when confronted with complex, open-world scenarios. While recent multimodal deep search agents attempt to address this issue by utilizing external tools, the visual-native search paradigm remains underexplored. Existing methods primarily rely on simple images with explicit semantics and text-only evidence trajectories, limiting the agent's ability to perform multi-hop, cross-modal reasoning and search. To address these limitations, we propose Visual-Seeker, a visual-native multimodal deep search agent via active visual reasoning. Rather than treating vision as a static input, our agent actively attends to fine-grained visual details, dynamically harvests visual evidence throughout the search process. To unlock its visual-native potential, we design an active visual reasoning data pipeline and synthesize 5K high-quality multimodal trajectories for model training. Extensive experiments demonstrate the state-of-the-art performance across five challenging multimodal search benchmarks, even surpassing several proprietary models, validating robust visual-native reasoning and search in real-world web environments. The code and data can be accessed at: https://github.com/ZhengboZhang/Visual-Seeker.","upvotes":3,"discussionId":"6a328a3959127a45e2c1c386","ai_summary":"Visual-Seeker enables visual-native multimodal deep search through active visual reasoning, outperforming proprietary models on real-world web environments.","ai_keywords":["multimodal large language models","visual-native search","deep search agents","active visual reasoning","multimodal trajectories","visual evidence harvesting"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct"},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"671906d1827bea3146f0b457","avatarUrl":"/avatars/4cf0b766143768e58f3cc41673f77cb1.svg","isPro":false,"fullname":"Bo","user":"Zhengbo-Zhang","type":"user"},{"_id":"6a32a2e74db83dccae1476b3","avatarUrl":"/avatars/94090490a59743e8163bcaa71a60214f.svg","isPro":false,"fullname":"panfeng yao","user":"ypf1101","type":"user"},{"_id":"6270324ebecab9e2dcf245de","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6270324ebecab9e2dcf245de/cMbtWSasyNlYc9hvsEEzt.jpeg","isPro":false,"fullname":"Kye Gomez","user":"kye","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.15231.md","query":{}}">
Visual-Seeker: Towards Visual-Native Multimodal Agentic Search via Active Visual Reasoning
Published on Jun 13
· Submitted by Bo on Jun 17 Authors: ,
,
,
,
,
,
,
,
,
,
,
,
Abstract
Visual-Seeker enables visual-native multimodal deep search through active visual reasoning, outperforming proprietary models on real-world web environments.
Multimodal large language models (MLLMs) have demonstrated impressive capabilities in many visual tasks, but they often struggle with factual grounding when confronted with complex, open-world scenarios. While recent multimodal deep search agents attempt to address this issue by utilizing external tools, the visual-native search paradigm remains underexplored. Existing methods primarily rely on simple images with explicit semantics and text-only evidence trajectories, limiting the agent's ability to perform multi-hop, cross-modal reasoning and search. To address these limitations, we propose Visual-Seeker, a visual-native multimodal deep search agent via active visual reasoning. Rather than treating vision as a static input, our agent actively attends to fine-grained visual details, dynamically harvests visual evidence throughout the search process. To unlock its visual-native potential, we design an active visual reasoning data pipeline and synthesize 5K high-quality multimodal trajectories for model training. Extensive experiments demonstrate the state-of-the-art performance across five challenging multimodal search benchmarks, even surpassing several proprietary models, validating robust visual-native reasoning and search in real-world web environments. The code and data can be accessed at: https://github.com/ZhengboZhang/Visual-Seeker.
Community
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/2606.15231 in a model README.md to link it from this page.
Cite arxiv.org/abs/2606.15231 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2606.15231 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.