<video src=\"https://cdn-uploads.huggingface.co/production/uploads/6340333a733f9eef46913dc8/D8P06xDmv5pYM1k364vA5.mp4\" controls=\"\" class=\"max-w-full!\"></video></p>","updatedAt":"2026-08-25T03:08:57.897Z","author":{"_id":"6340333a733f9eef46913dc8","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6340333a733f9eef46913dc8/BwhKgkqbuuxWsJv2s_JEH.png","fullname":"luozhou wang","name":"wileewang","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":27,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.6106469035148621},"editors":["wileewang"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/6340333a733f9eef46913dc8/BwhKgkqbuuxWsJv2s_JEH.png"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.23565","authors":[{"_id":"6a8d02b05add2537c32e96c3","name":"Zhifei Chen","hidden":false},{"_id":"6a8d02b05add2537c32e96c4","name":"Luozhou Wang","hidden":false},{"_id":"6a8d02b05add2537c32e96c5","name":"Guibao Shen","hidden":false},{"_id":"6a8d02b05add2537c32e96c6","name":"Dongyu Yan","hidden":false},{"_id":"6a8d02b05add2537c32e96c7","name":"Shuai Yang","hidden":false},{"_id":"6a8d02b05add2537c32e96c8","name":"Tianshuo Xu","hidden":false},{"_id":"6a8d02b05add2537c32e96c9","name":"Yihua Du","hidden":false},{"_id":"6a8d02b05add2537c32e96ca","name":"Wei Wang","hidden":false},{"_id":"6a8d02b05add2537c32e96cb","name":"Tianyi Gui","hidden":false},{"_id":"6a8d02b05add2537c32e96cc","name":"Lianghua Huang","hidden":false},{"_id":"6a8d02b05add2537c32e96cd","name":"Yingcong Chen","hidden":false}],"publishedAt":"2026-08-24T00:00:00.000Z","submittedOnDailyAt":"2026-08-25T00:00:00.000Z","title":"ReWorld: An Interactive World Model with Long-Horizon Memory","submittedOnDailyBy":{"_id":"6340333a733f9eef46913dc8","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6340333a733f9eef46913dc8/BwhKgkqbuuxWsJv2s_JEH.png","isPro":false,"fullname":"luozhou wang","user":"wileewang","type":"user","name":"wileewang"},"summary":"An interactive world model must follow the user's actions, remember the places it has shown, and stream in real time. The tension is structural: control wants a short horizon, memory wants an unbounded one. ReWorld separates the two during training and bounds them at inference. Mixed per-head attention windows confine most heads to the recent past while a small set of global heads attends over the entire history, and random head routing keeps either capability from binding to particular heads; random chunk dropping makes sparse histories in-distribution. At inference the whole past lives under a fixed budget: a bounded KV cache backed by a pose-indexed landmark bank, from which the model retrieves the landmarks nearest the current pose. A metric-scale-aligned data engine places eight sources -- Unreal-rendered fly-throughs, game roaming, and real-world footage -- on one physical action scale, so the same key press moves the camera the same distance in every source, and palindrome trajectories supply the revisit evidence that memory training needs. Distribution-matching distillation confined to a LoRA adapter then compresses sampling to four steps: one backbone serves both a high-fidelity multi-step mode and a real-time interactive one, streaming 704x1280 video across photorealistic, game-style, and stylized worlds. Under a three-axis protocol covering action following, long-horizon recall, and video quality, against six recent interactive world models it attains the best control fidelity (11.95^circ rotation error and the best camera-motion consistency) and the best generation quality; and on minute-long out-and-back rollouts (64\\,s, 384 latents), its fixed 12-chunk cache still regenerates the starting view -- at rollout lengths where a sliding window has long evicted the evidence and full-KV attention runs out of memory.","upvotes":11,"discussionId":"6a8d02b05add2537c32e96ce","projectPage":"https://zhifeichen097.github.io/ReWorld/","githubRepo":"https://github.com/zhifeichen097/ReWorld","githubRepoAddedBy":"user","ai_summary":"ReWorld separates short-horizon control and long-horizon memory during training, then bounds both at inference via mixed attention windows, a pose-indexed landmark bank, and distribution-matching LoRA distillation to enable real-time interactive world modeling with strong action fidelity and long-range recall.","ai_keywords":["mixed per-head attention windows","global heads","random head routing","random chunk dropping","pose-indexed landmark bank","KV cache","metric-scale-aligned data engine","palindrome trajectories","distribution-matching distillation","LoRA adapter"],"ai_summary_model":"thinkingmachines/Inkling-Small","githubStars":3,"organization":{"_id":"67d15cca6e2cf0e062dbfb54","name":"AlibabaTongyiLab","fullname":"TongyiLab","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/67d1502bfabfe9974d1f77bb/XdUSVf6HqBzE7zFBfSDQP.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6340333a733f9eef46913dc8","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6340333a733f9eef46913dc8/BwhKgkqbuuxWsJv2s_JEH.png","isPro":false,"fullname":"luozhou wang","user":"wileewang","type":"user"},{"_id":"634a91c53a0cd2d4985dc98a","avatarUrl":"/avatars/fe680e683761c347cfa4bc9273b67b0d.svg","isPro":false,"fullname":"ZhiFei Chen","user":"zhifeichen097","type":"user"},{"_id":"634ce90e741a5e37886a19e3","avatarUrl":"/avatars/0d1579039136b37db5b67282b0a34c33.svg","isPro":false,"fullname":"Syang","user":"Andyson","type":"user"},{"_id":"6039478ab3ecf716b1a5fd4d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg","isPro":true,"fullname":"taesiri","user":"taesiri","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"},{"_id":"65214c46f6ceb915cc790275","avatarUrl":"/avatars/c30541fd8ea55d479740f534a49e6248.svg","isPro":false,"fullname":"Yihua Du","user":"Duyh","type":"user"},{"_id":"620783f24e28382272337ba4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/620783f24e28382272337ba4/zkUveQPNiDfYjgGhuFErj.jpeg","isPro":false,"fullname":"GuoLiangTang","user":"Tommy930","type":"user"},{"_id":"686b40c60ad06652bc59c8c8","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/686b40c60ad06652bc59c8c8/eqQl2IQ4QvEjz3jxvp1gl.jpeg","isPro":false,"fullname":"Yin","user":"Sleeepyin","type":"user"},{"_id":"64049ae20ab5e22719f35103","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1678023295407-noauth.jpeg","isPro":false,"fullname":"Dongyu Yan","user":"StarYDY","type":"user"},{"_id":"65d5aa45dca2a85f0fe895f3","avatarUrl":"/avatars/a3cbcade6ea101e99f58641aa409fdfe.svg","isPro":false,"fullname":"Guibao SHEN","user":"PaulSHEN1","type":"user"},{"_id":"6418554a0956be7233a1023e","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6418554a0956be7233a1023e/9EKN0GoOpcDbvBDmAQEJf.png","isPro":false,"fullname":"zhang yuechen","user":"julianjuaner","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"67d15cca6e2cf0e062dbfb54","name":"AlibabaTongyiLab","fullname":"TongyiLab","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/67d1502bfabfe9974d1f77bb/XdUSVf6HqBzE7zFBfSDQP.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.23565.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2608.23565 in a model README.md to link it from this page.

Cite arxiv.org/abs/2608.23565 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2608.23565 in a Space README.md to link it from this page.