🚀 ReactiveGWM: Bringing true autonomy to NPCs in Game World Models! Instead of treating NPCs as background pixels , we explicitly decouple player controls from NPC behaviors. </p>\n<p>✨ Highlights:<br>Strategy-Driven: NPCs autonomously execute high-level intents (Offense, Control, Defense).<br>Zero-Shot Transfer: The learned interactive logic is game-agnostic. Plug our modules directly into unannotated world models of different games to unlock steerable NPCs instantly—no domain-specific retraining needed!<br>High Fidelity: Preserves fine-grained player control and native visual dynamics. </p>\n<p>👇 Dive in:<br>📄 Arxiv: <a href=\"http://arxiv.org/abs/2605.15256\" rel=\"nofollow\">http://arxiv.org/abs/2605.15256</a><br>💻 Code: <a href=\"https://github.com/INV-WZQ/ReactiveGWM\" rel=\"nofollow\">https://github.com/INV-WZQ/ReactiveGWM</a><br>🏠 Project: <a href=\"https://inv-wzq.github.io/ReactiveGWM/\" rel=\"nofollow\">https://inv-wzq.github.io/ReactiveGWM/</a><br>🤗 Model: <a href=\"https://huggingface.co/INV-WZQ/ReactiveGWM-Models\">https://huggingface.co/INV-WZQ/ReactiveGWM-Models</a><br>🗄️ Dataset: <a href=\"https://huggingface.co/datasets/INV-WZQ/ReactiveGWM-Datasets\">https://huggingface.co/datasets/INV-WZQ/ReactiveGWM-Datasets</a></p>\n","updatedAt":"2026-05-18T01:48:52.355Z","author":{"_id":"668e740f1173ab43d9d9ed5e","avatarUrl":"/avatars/caa9b47c2a5f6d6d679759b8b234a0ab.svg","fullname":"Zeqing Wang","name":"INV-WZQ","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":5,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7337396144866943},"editors":["INV-WZQ"],"editorAvatarUrls":["/avatars/caa9b47c2a5f6d6d679759b8b234a0ab.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2605.15256","authors":[{"_id":"6a0a6d8475184a0d71e02587","user":{"_id":"668e740f1173ab43d9d9ed5e","avatarUrl":"/avatars/caa9b47c2a5f6d6d679759b8b234a0ab.svg","isPro":true,"fullname":"Zeqing Wang","user":"INV-WZQ","type":"user","name":"INV-WZQ"},"name":"Zeqing Wang","status":"claimed_verified","statusLastChangedAt":"2026-05-18T09:40:58.873Z","hidden":false},{"_id":"6a0a6d8475184a0d71e02588","name":"Danze Chen","hidden":false},{"_id":"6a0a6d8475184a0d71e02589","name":"Zhaohu Xing","hidden":false},{"_id":"6a0a6d8475184a0d71e0258a","name":"Zizhao Tong","hidden":false},{"_id":"6a0a6d8475184a0d71e0258b","name":"Yinhan Zhang","hidden":false},{"_id":"6a0a6d8475184a0d71e0258c","name":"Xingyi Yang","hidden":false},{"_id":"6a0a6d8475184a0d71e0258d","name":"Yeying Jin","hidden":false}],"mediaUrls":["https://cdn-uploads.huggingface.co/production/uploads/668e740f1173ab43d9d9ed5e/te7qYOGruGEQnIEzZj9BG.mp4"],"publishedAt":"2026-05-14T00:00:00.000Z","submittedOnDailyAt":"2026-05-18T00:00:00.000Z","title":"ReactiveGWM: Steering NPC in Reactive Game World Models","submittedOnDailyBy":{"_id":"668e740f1173ab43d9d9ed5e","avatarUrl":"/avatars/caa9b47c2a5f6d6d679759b8b234a0ab.svg","isPro":true,"fullname":"Zeqing Wang","user":"INV-WZQ","type":"user","name":"INV-WZQ"},"summary":"Current game world models simulate environments from a subjective, player-centric perspective. However, by treating the Non-Player Character (NPC) merely as background pixels, these models cannot capture interactions between the player and NPC. In that sense, they act as passive video renderers rather than real simulation engines, lacking the physical understanding needed to model action-induced NPC reactivities. We introduce ReactiveGWM, a reactive game world model that synthesizes dynamic interactions between the player and NPC. Instead of entangling all interaction dynamics, ReactiveGWM explicitly decouples player controls from NPC behaviors. Player actions are injected into the diffusion backbone via a lightweight additive bias, while high-level NPC responses (e.g., Offense, Control, Defense) are grounded through cross-attention modules. Crucially, these modules learn a game-agnostic representation of interactive logic. This enables zero-shot strategy transfer: our learned modules can be plugged directly into off-the-shelf, unannotated world models of different games. This instantly unlocks steerable NPC interactions without any domain-specific retraining. Evaluated on two Street Fighter games, ReactiveGWM maintains fine-grain player controllability while achieving robust, prompt-aligned NPC strategy adherence, paving the way for scalable, strategy-rich interaction with the NPC.","upvotes":24,"discussionId":"6a0a6d8475184a0d71e0258e","projectPage":"https://inv-wzq.github.io/ReactiveGWM/","githubRepo":"https://github.com/INV-WZQ/ReactiveGWM","githubRepoAddedBy":"user","ai_summary":"ReactiveGWM enables dynamic player-NPC interactions in game worlds by decoupling player controls from NPC behaviors through diffusion models with cross-attention modules for game-agnostic strategy transfer.","ai_keywords":["diffusion models","cross-attention modules","game-agnostic representation","zero-shot strategy transfer","player controls","NPC behaviors","reactive game world model","interactive logic","strategy-rich interaction"],"githubStars":34},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"668e740f1173ab43d9d9ed5e","avatarUrl":"/avatars/caa9b47c2a5f6d6d679759b8b234a0ab.svg","isPro":true,"fullname":"Zeqing Wang","user":"INV-WZQ","type":"user"},{"_id":"662236b6774056984cc88670","avatarUrl":"/avatars/d6413765e3931d7b300e4e985ccd9c16.svg","isPro":false,"fullname":"auagcu","user":"auagcu","type":"user"},{"_id":"634cfebc350bcee9bed20a4d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/634cfebc350bcee9bed20a4d/fN47nN5rhw-HJaFLBZWQy.png","isPro":false,"fullname":"Xingyi Yang","user":"adamdad","type":"user"},{"_id":"6729d1fed3ec5370cb035901","avatarUrl":"/avatars/50f7ce9c635148df76d1c63ebf3efa38.svg","isPro":false,"fullname":"1","user":"DANNY621","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":"677e31bb114aeff62d62bc11","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/JA85qWXtPO_mnK7_Bu2HJ.png","isPro":false,"fullname":"ZIJUN LIN","user":"onepiece1999","type":"user"},{"_id":"6860fe55a1ab4d5c885c3edf","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/HNSwVGoTd33mgbyjxA-fc.jpeg","isPro":false,"fullname":"QIN ZHIBIN","user":"tuantuan0321","type":"user"},{"_id":"69f575206fd96fe49c905781","avatarUrl":"/avatars/cd20544d367b417e78c400df2df31a59.svg","isPro":false,"fullname":"Danchen Guan","user":"DestInies","type":"user"},{"_id":"670b862d64cab942c0f358be","avatarUrl":"/avatars/dfba41565499ef17e17d12d5f5fe9c97.svg","isPro":false,"fullname":"Fu Youquan","user":"FYQ12138","type":"user"},{"_id":"6a06afdcb72ebb013fb4ff6d","avatarUrl":"/avatars/46651e9cea0e22cfab1b4d090c2c376c.svg","isPro":false,"fullname":"slwy","user":"slsdfaf","type":"user"},{"_id":"68dbe140c7fc381a7ba96efa","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/6j5mp3UXaonecsS6575mc.png","isPro":false,"fullname":"杨宇昂","user":"yyaOnions","type":"user"},{"_id":"68dbcbd15ee1009ceaedb052","avatarUrl":"/avatars/23093768b8bc5b085f446307daa9b9a5.svg","isPro":false,"fullname":"kyle","user":"kyle012","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2605/2605.15256.md"}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2605.15256 in a Space README.md to link it from this page.