MAGIC: Transition-Aware Generation of Navigable Multi-Scene Game Worlds with Large Language Models
Summary
MAGIC is a system that uses large language models to generate connected, navigable multi-scene game worlds from a single natural-language prompt, addressing cross-scene consistency, in-scene navigability, and transition evaluation. It achieves high precision and recall on a benchmark of 100 multi-scene cases.
View Cached Full Text
Cached at: 07/16/26, 01:40 AM
Paper page - MAGIC: Transition-Aware Generation of Navigable Multi-Scene Game Worlds with Large Language Models
Source: https://huggingface.co/papers/2607.11594
Abstract
Multi-scenenavigation(clearinganobjectiveinoneboundedspaceandthencrossingaportalintothenext)isadefiningfeatureofcontemporary3Dgames,butauthoringitislaborious:everyportalmusthaveconsistentendpointsonbothsides,eachinteriormustremainnavigableonceitisfurnished,andtheresultingconnectivitymustbekeptconsistentacrossmanyfiles.Recentlargelanguagemodel(LLM)andmultimodalLLM(MLLM)scenegeneratorshavemadesingle-interiorsynthesisdramaticallycheaper,yettheyproduceonesceneatatimeandcannot,bynaiverepetition,yieldaconnectedmulti-sceneworld.Weidentifythreeobstaclesthatsingle-scenemethodsleaveunsolved:cross-sceneconsistency,in-scenenavigability,andtheevaluationofwhetheratransitionactuallyworks.WepresentMAGIC,aprompt-to-projectsystemthataddressesallthree.MAGICisafour-stagepipelinethatturnsasinglenatural-languagepromptintoarunnablemulti-scenegameproject:itplansasharedtransition-awareintermediaterepresentation,specifieseachscenewhileenforcingportalreachabilitywithaflood-fillvalidator,generatesthescenestogetherwiththeirtransitionscripts,andcombinesthemintooneproject.Becauseexistingsingle-scenefidelitymetricsneverexecuteatransition,wefurtherintroduceatransition-focusedevaluationagentthatrunseachtransitioninplay.Onanewbenchmarkof100multi-scenecases,MAGICproducesanexecutableprojectforeverycaseandreaches0.99precision,0.95recall,and0.96F1onend-to-endtransitionidentification;stagebystage,itrecoversmoreground-truthportalsandyieldsmarkedlymorenavigablelayoutsthananLLMbaselineandHolodeck.Ourcodeisavailableathttps://github.com/sereneee1201/MAGIC/.
View arXiv pageView PDFProject pageGitHub0Add to collection
Get this paper in your agent:
hf papers read 2607\.11594
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2607.11594 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2607.11594 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2607.11594 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
WM^MLLM: Turning MLLMs into World Models
WM^MLLM 将多模态大语言模型改造为交互式世界模型,通过外部全景状态与修订历史实现从单张图像出发的持久、可编辑世界生成,支持精确重访与局部编辑,并在 MBench-A、CompBench、WBench 等基准上显著超越现有世界模型基线。
Beyond the Current Observation: Evaluating Multimodal Large Language Models in Controllable Non-Markov Games
This paper introduces RNG-Bench, a benchmark suite for evaluating multimodal foundation models' ability to reconstruct past observations and use them for decision-making in multi-step interactions, featuring two games (Matching Pairs and 3D Maze) with controlled difficulty parameters and a memory gap metric to distinguish forgetting from poor decision-making.
Right Screen, Wrong Transition: World Models as Verifiers for GUI Agents
The paper introduces LGWM, a decoder-free, action-conditioned world model that predicts next-screen representations in latent space to verify GUI transitions and detect hijacked screens, achieving 0.987 AUC at 17ms per decision on their RSWT-BENCH benchmark, outperforming generative GUI world models with vastly lower latency.
SceneMosaic: Efficient and Diverse Simulation-Ready Scene Generation via Hybrid Agentic Layout Evolution
SceneMosaic combines learned image priors with vision-language agents to efficiently generate diverse, physically valid indoor scenes, achieving a 24x speedup and improved physical validity over existing methods.
Why We Need World Models for AGI: Where LLMs Fail and How World Models May Outperform
This paper argues that large language models struggle with causal reasoning and long-horizon planning due to a mismatch between sequence prediction and reasoning over latent environment dynamics, and introduces the Latent Dynamics Inference perspective along with the Flux environment to study these limitations.