Kairos Technical Report</p>\n","updatedAt":"2026-06-18T02:49:24.721Z","author":{"_id":"65d60765d484956b5ac7d688","avatarUrl":"/avatars/2b6b2be00b852097e9b2e11076e8c4ba.svg","fullname":"Shan You","name":"shanyou92","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.4585890471935272},"editors":["shanyou92"],"editorAvatarUrls":["/avatars/2b6b2be00b852097e9b2e11076e8c4ba.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.16533","authors":[{"_id":"6a3359b659127a45e2c1c5c2","name":"Kairos Team","hidden":false},{"_id":"6a3359b659127a45e2c1c5c3","name":"Fei Wang","hidden":false},{"_id":"6a3359b659127a45e2c1c5c4","name":"Shan You","hidden":false},{"_id":"6a3359b659127a45e2c1c5c5","name":"Qiming Zhang","hidden":false},{"_id":"6a3359b659127a45e2c1c5c6","name":"Tao Huang","hidden":false},{"_id":"6a3359b659127a45e2c1c5c7","name":"Zuoyi Fu","hidden":false},{"_id":"6a3359b659127a45e2c1c5c8","name":"Zhisheng Zheng","hidden":false},{"_id":"6a3359b659127a45e2c1c5c9","name":"Yunlong Xi","hidden":false},{"_id":"6a3359b659127a45e2c1c5ca","name":"Feng Lv","hidden":false},{"_id":"6a3359b659127a45e2c1c5cb","name":"Xiaoming Wu","hidden":false},{"_id":"6a3359b659127a45e2c1c5cc","name":"Zeyu Liu","hidden":false},{"_id":"6a3359b659127a45e2c1c5cd","name":"Cong Wan","hidden":false},{"_id":"6a3359b659127a45e2c1c5ce","name":"Pu Li","hidden":false},{"_id":"6a3359b659127a45e2c1c5cf","name":"Ruiqing Yang","hidden":false},{"_id":"6a3359b659127a45e2c1c5d0","name":"Xiaoou Li","hidden":false},{"_id":"6a3359b659127a45e2c1c5d1","name":"Wei Wang","hidden":false},{"_id":"6a3359b659127a45e2c1c5d2","name":"Kangkang Zhu","hidden":false},{"_id":"6a3359b659127a45e2c1c5d3","name":"Yuwei Zhang","hidden":false},{"_id":"6a3359b659127a45e2c1c5d4","name":"Shi Fu","hidden":false},{"_id":"6a3359b659127a45e2c1c5d5","name":"Zheng Zhang","hidden":false},{"_id":"6a3359b659127a45e2c1c5d6","name":"Xiaoning Wu","hidden":false},{"_id":"6a3359b659127a45e2c1c5d7","name":"Xuzeng Fan","hidden":false},{"_id":"6a3359b659127a45e2c1c5d8","name":"Dacheng Tao","hidden":false},{"_id":"6a3359b659127a45e2c1c5d9","name":"Xiaogang Wang","hidden":false}],"publishedAt":"2026-06-16T00:00:00.000Z","submittedOnDailyAt":"2026-06-18T00:00:00.000Z","title":"Kairos: A Native World Model Stack for Physical AI","submittedOnDailyBy":{"_id":"65d60765d484956b5ac7d688","avatarUrl":"/avatars/2b6b2be00b852097e9b2e11076e8c4ba.svg","isPro":false,"fullname":"Shan You","user":"shanyou92","type":"user","name":"shanyou92"},"summary":"World models are transitioning from passive visual generators to foundational, operational infrastructure for Physical AI: they must natively acquire world knowledge from heterogeneous experience, maintain persistent states over long horizons, and execute efficiently within real deployment constraints. We introduce Kairos, a native world model stack designed around these requirements. (1) Kairos learns the world by pioneering a Native Pre-training Paradigm governed by a Cross-Embodiment Data Curriculum, which organizes open-world videos, human behavioral data, and robot interactions into a progressive developmental pathway. (2) Kairos maintains the world by unified world understanding, generation, and prediction within a Native Unified Architecture equipped with Hybrid Linear Temporal Attention, where sliding-window attention captures local dynamics, dilated sliding windows capture mid-range dependencies, and gated linear attention maintains persistent global memory. We establish formal theoretical bounds demonstrating that this temporal factorization strictly limits error accumulation, mathematically guaranteeing state propagation across extended horizons. (3) Kairos runs the world by incorporating a Deployment-Aware System Co-Design to support low-latency rollout generation on server and consumer-grade hardware for real-world observation-action-feedback loops. Experiments on embodied world-model, long-horizon, and action-policy benchmarks show that Kairos achieves top level performance while offering a strong efficiency-capability trade-off. Together, these results position Kairos as a cohesive operational foundation for future self-evolving physical intelligence.","upvotes":19,"discussionId":"6a3359b659127a45e2c1c5da","githubRepo":"https://github.com/kairos-agi/kairos-sensenova","githubRepoAddedBy":"user","ai_summary":"Kairos is a native world model framework that learns from diverse experiences, maintains persistent states through hybrid temporal attention, and supports efficient deployment for physical AI applications.","ai_keywords":["world models","native pre-training paradigm","cross-embodiment data curriculum","native unified architecture","hybrid linear temporal attention","sliding-window attention","dilated sliding windows","gated linear attention","temporal factorization","error accumulation","deployment-aware system co-design","embodied world-model","long-horizon benchmarks","action-policy benchmarks"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":712},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"64d1cfaf44d373d70673b627","avatarUrl":"/avatars/151b9f83ca91498ca8e201ecf0289c77.svg","isPro":false,"fullname":"Charlie","user":"Dyland2","type":"user"},{"_id":"691f05ac8c7f4781804ba19a","avatarUrl":"/avatars/33e8c87596c779a4d5b4a22fd3b3e24b.svg","isPro":false,"fullname":"Delancy","user":"Delancy1","type":"user"},{"_id":"65d60765d484956b5ac7d688","avatarUrl":"/avatars/2b6b2be00b852097e9b2e11076e8c4ba.svg","isPro":false,"fullname":"Shan You","user":"shanyou92","type":"user"},{"_id":"67ea50a4ccb7286fd33fc6f0","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/gxvQ2AiAJXMKYX5CKC8gE.png","isPro":false,"fullname":"Zeyu Liu","user":"Zane-Liu","type":"user"},{"_id":"653f979daf499ca76195eabb","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/653f979daf499ca76195eabb/EI4F4X5-C0OBhhx90KRed.jpeg","isPro":false,"fullname":"finn zheng","user":"superzzs","type":"user"},{"_id":"6a335f4c4b5c1c0ebeec94fc","avatarUrl":"/avatars/2bd336422a9aa03052604720da9a093e.svg","isPro":false,"fullname":"Chen","user":"LiangRobot","type":"user"},{"_id":"668f7e38709bfdf42c4d09a3","avatarUrl":"/avatars/cef08784f71559d5bb805693ff556a08.svg","isPro":false,"fullname":"Tao Huang","user":"Hunto98","type":"user"},{"_id":"69eb5cff11b847ca607c41ae","avatarUrl":"/avatars/fdd148677daf967b297dc5296e0e3569.svg","isPro":false,"fullname":"wu xiaoming","user":"draven-alg","type":"user"},{"_id":"685904363c490e43cdcf0373","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/lq4HHkv2699mdmkIUd3_e.png","isPro":false,"fullname":"YU","user":"liyudut123","type":"user"},{"_id":"61dd2b7389dddd97daead12f","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/61dd2b7389dddd97daead12f/xUL7YK3Mtvzz4rEhorFKF.jpeg","isPro":false,"fullname":"Xiao-Ming Wu","user":"DravenALG","type":"user"},{"_id":"692fec4a01fb75217ab086b1","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/692fec4a01fb75217ab086b1/yMRzYceN_5eUiH4nhYXSa.png","isPro":false,"fullname":"kairos-agi","user":"kairos-agi","type":"user"},{"_id":"65260280b36077356b44bc99","avatarUrl":"/avatars/bd50b39cd131a278e20a011a71f0eec9.svg","isPro":false,"fullname":"w","user":"huggwan","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":3,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.16533.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2606.16533 in a model README.md to link it from this page.

Cite arxiv.org/abs/2606.16533 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2606.16533 in a Space README.md to link it from this page.

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