This paper presents JD Oxygen AIIC, an industrial-scale LLM/VLM-centric platform for item knowledge production serving tens of billions of SKUs on JD.com. The system achieves 94.2% precision and 82.8% recall in knowledge production, processes hundreds of millions of item updates daily, and has delivered measurable business gains across search, recommendation, and operations.</p>\n","updatedAt":"2026-07-09T00:47:55.281Z","author":{"_id":"6644461a294dd5ab53e9cdc5","avatarUrl":"/avatars/fcac49f1aefc51b09d996a29e4101628.svg","fullname":"Yongyi Liao","name":"Allen03","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.839877188205719},"editors":["Allen03"],"editorAvatarUrls":["/avatars/fcac49f1aefc51b09d996a29e4101628.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.28070","authors":[{"_id":"6a4b1743d66e8ea154bc8180","name":"Oxygen AIIC","hidden":false},{"_id":"6a4b1743d66e8ea154bc8181","name":"Chan Long","hidden":false},{"_id":"6a4b1743d66e8ea154bc8182","name":"Chao Liu","hidden":false},{"_id":"6a4b1743d66e8ea154bc8183","name":"Chaofan Chen","hidden":false},{"_id":"6a4b1743d66e8ea154bc8184","name":"Chaohui Dong","hidden":false},{"_id":"6a4b1743d66e8ea154bc8185","name":"Chunyuan Guo","hidden":false},{"_id":"6a4b1743d66e8ea154bc8186","name":"Danping Liu","hidden":false},{"_id":"6a4b1743d66e8ea154bc8187","name":"Debin Liu","hidden":false},{"_id":"6a4b1743d66e8ea154bc8188","name":"Deping Xiang","hidden":false},{"_id":"6a4b1743d66e8ea154bc8189","name":"Fulai Xu","hidden":false},{"_id":"6a4b1743d66e8ea154bc818a","name":"Guangyue Liu","hidden":false},{"_id":"6a4b1743d66e8ea154bc818b","name":"Hao Li","hidden":false},{"_id":"6a4b1743d66e8ea154bc818c","name":"Huichun Hu","hidden":false},{"_id":"6a4b1743d66e8ea154bc818d","name":"Jian Yang","hidden":false},{"_id":"6a4b1743d66e8ea154bc818e","name":"Jianan Wang","hidden":false},{"_id":"6a4b1743d66e8ea154bc818f","name":"Jianbo Zhao","hidden":false},{"_id":"6a4b1743d66e8ea154bc8190","name":"Jiaoyang Li","hidden":false},{"_id":"6a4b1743d66e8ea154bc8191","name":"Jiaxing Wang","hidden":false},{"_id":"6a4b1743d66e8ea154bc8192","name":"Jinglong Li","hidden":false},{"_id":"6a4b1743d66e8ea154bc8193","name":"Jinjin Guo","hidden":false},{"_id":"6a4b1743d66e8ea154bc8194","name":"Jun Fang","hidden":false},{"_id":"6a4b1743d66e8ea154bc8195","name":"Jun Liu","hidden":false},{"_id":"6a4b1743d66e8ea154bc8196","name":"Kai Zhou","hidden":false},{"_id":"6a4b1743d66e8ea154bc8197","name":"Li Wang","hidden":false},{"_id":"6a4b1743d66e8ea154bc8198","name":"Lili Gao","hidden":false},{"_id":"6a4b1743d66e8ea154bc8199","name":"Liying Chen","hidden":false},{"_id":"6a4b1743d66e8ea154bc819a","name":"Luning Yang","hidden":false},{"_id":"6a4b1743d66e8ea154bc819b","name":"Mengdi Zhou","hidden":false},{"_id":"6a4b1743d66e8ea154bc819c","name":"Pengzhang Liu","hidden":false},{"_id":"6a4b1743d66e8ea154bc819d","name":"Qi Lv","hidden":false},{"_id":"6a4b1743d66e8ea154bc819e","name":"Qianyun Wang","hidden":false},{"_id":"6a4b1743d66e8ea154bc819f","name":"Qixia Jiang","hidden":false},{"_id":"6a4b1743d66e8ea154bc81a0","name":"Ruyue Li","hidden":false},{"_id":"6a4b1743d66e8ea154bc81a1","name":"Shimu Liang","hidden":false},{"_id":"6a4b1743d66e8ea154bc81a2","name":"Shuxing Wang","hidden":false},{"_id":"6a4b1743d66e8ea154bc81a3","name":"Sijie Zhang","hidden":false},{"_id":"6a4b1743d66e8ea154bc81a4","name":"Siqi Li","hidden":false},{"_id":"6a4b1743d66e8ea154bc81a5","name":"Tianhao Gao","hidden":false},{"_id":"6a4b1743d66e8ea154bc81a6","name":"Wang Ke","hidden":false},{"_id":"6a4b1743d66e8ea154bc81a7","name":"Weihu Huang","hidden":false},{"_id":"6a4b1743d66e8ea154bc81a8","name":"Wencan Lai","hidden":false},{"_id":"6a4b1743d66e8ea154bc81a9","name":"Wenjie Zhang","hidden":false},{"_id":"6a4b1743d66e8ea154bc81aa","name":"Xiaohui Zhang","hidden":false},{"_id":"6a4b1743d66e8ea154bc81ab","name":"Xiaojing Dong","hidden":false},{"_id":"6a4b1743d66e8ea154bc81ac","name":"Ya Liu","hidden":false},{"_id":"6a4b1743d66e8ea154bc81ad","name":"Yifeng Zhang","hidden":false},{"_id":"6a4b1743d66e8ea154bc81ae","name":"Yixiang Wang","hidden":false},{"_id":"6a4b1743d66e8ea154bc81af","name":"Yongtai Zhang","hidden":false},{"_id":"6a4b1743d66e8ea154bc81b0","user":{"_id":"6644461a294dd5ab53e9cdc5","avatarUrl":"/avatars/fcac49f1aefc51b09d996a29e4101628.svg","isPro":false,"fullname":"Yongyi Liao","user":"Allen03","type":"user","name":"Allen03"},"name":"Yongyi Liao","status":"admin_assigned","statusLastChangedAt":"2026-07-08T18:20:33.312Z","hidden":false},{"_id":"6a4b1743d66e8ea154bc81b1","name":"Zhaoru Chen","hidden":false},{"_id":"6a4b1743d66e8ea154bc81b2","name":"Zhen Chen","hidden":false},{"_id":"6a4b1743d66e8ea154bc81b3","name":"Zhiyong Ma","hidden":false},{"_id":"6a4b1743d66e8ea154bc81b4","name":"Zhiyuan Liu","hidden":false},{"_id":"6a4b1743d66e8ea154bc81b5","name":"Zhongwei Liu","hidden":false},{"_id":"6a4b1743d66e8ea154bc81b6","name":"Ziyan Xing","hidden":false}],"publishedAt":"2026-06-29T00:00:00.000Z","submittedOnDailyAt":"2026-07-08T00:00:00.000Z","title":"JD Oxygen AI Item Center (Oxygen AIIC) V1: An Industrial-Scale LLM/VLM-Centric Solution for Item Understanding, Management, and Applications","submittedOnDailyBy":{"_id":"6644461a294dd5ab53e9cdc5","avatarUrl":"/avatars/fcac49f1aefc51b09d996a29e4101628.svg","isPro":false,"fullname":"Yongyi Liao","user":"Allen03","type":"user","name":"Allen03"},"summary":"JD.com, one of the world's largest e-commerce platforms, serves over 700 million active users and millions of merchants, with a catalog of tens of billions of SKUs. At this scale, high-quality, structured item knowledge underpins a better consumer experience, lower management costs, and higher operational efficiency-yet producing and serving it poses three industrial-scale challenges: fast-emerging concepts, high-quality knowledge production for massive SKUs, and diverse downstream requirements. To address these challenges, we present the JD Oxygen AI Item Center (Oxygen AIIC), an industrial-scale platform built on LLMs/VLMs for item-knowledge production and service. Oxygen AIIC is built around four core pillars: (i) ontology engineering driven by efficient human-AI collaboration, which supports the dynamic evolution and agile expansion of an ontology with millions of entries; (ii) a \"Semantic Search then Discrimination\"(S2D) knowledge identification architecture that, combined with throughput improvement strategies, enables scalable, extensible, and high-throughput AI Item Library production for tens of billions of SKUs; (iii) self-evolving item-understanding LLMs/VLMs that improve in a stable and controllable manner, enabling knowledge production with 94.2% precision and 82.8% recall; and (iv) a unified item tunnel that serves as the data and service hub. Oxygen AIIC now covers tens of thousands of JD categories and processes hundreds of millions of item updates per day on Huawei Ascend NPUs. It has accumulated hundreds of billions of item-knowledge assets. Deployed across core business scenarios-including search, recommendation, operations, category planning-Oxygen AIIC has delivered measurable gains at scale. Search-traffic coverage reaches 80.4%, item-information quality issues drop by 37%, the automated fill rate of core attributes during item listing exceeds 80%.","upvotes":2,"discussionId":"6a4b1743d66e8ea154bc81b7","ai_summary":"A large-scale industrial platform leveraging LLMs and VLMs addresses key challenges in structured item knowledge production for e-commerce, achieving high precision and throughput across billions of products.","ai_keywords":["LLMs","VLMs","ontology engineering","Semantic Search then Discrimination","AI Item Library","self-evolving item-understanding","unified item tunnel","item-knowledge production"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct"},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6167d5df081d4f73c3ae4a37","avatarUrl":"/avatars/8aeb4761db3783611b0ffd0420cc8b8d.svg","isPro":false,"fullname":"FANGJUN","user":"fenshioneach","type":"user"},{"_id":"6644461a294dd5ab53e9cdc5","avatarUrl":"/avatars/fcac49f1aefc51b09d996a29e4101628.svg","isPro":false,"fullname":"Yongyi Liao","user":"Allen03","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.28070.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2606.28070 in a model README.md to link it from this page.

Cite arxiv.org/abs/2606.28070 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2606.28070 in a Space README.md to link it from this page.

Add this paper to a collection to link it from this page.