Summary
Interactive worlds, like video games, change based on rules about how things act and respond. The authors address how to keep these rules working correctly when saving and reopening worlds, and how to make sure rule effects appear properly even on things not yet seen on screen. They created WorldRuler, a system that uses 3D rule graphs to keep rules, objects, and world states connected so the world behaves consistently and changes look correct. They also made a dataset called RuleVerse to help train and test these rule-based world models. Tests show WorldRuler keeps rules and world states intact across sessions and realistically applies rule effects to specific objects and features.
What this means in practice
- •For game developers: Create interactive 3D game worlds that consistently preserve and apply custom gameplay rules across play sessions.
- •For virtual environment designers: Build customizable simulations where rule-driven interactions are visually accurate even when affecting unseen objects or hidden states.
Authors
Mingliang Zhai, Chuanhao Li, Jiaming Tan, Zhen Li, Yuwei Wu, Yunde Jia, Zian Meng, Yuanyang Yin, Kaipeng Zhang
Abstract
Rules govern how player actions change interactive worlds. Different rules create distinct gameplay experiences within similar scenes and action spaces. For customizable gameplay, generative world models must follow user-defined rules beyond real-world behaviors. However, realizing user-defined rules presents two challenges: Rule-State Persistence and Rule-to-Visual Realization. The former requires preserving rule definitions, dependencies, and world states across saving and loading, so that restored worlds follow the established rules. The latter requires accurately realizing rule-induced changes on designated objects and attributes, particularly when affected objects are unseen. In this paper, we propose WorldRuler, which organizes rule definitions, dependencies, and object states in an executable 3D rule graph and connects rule execution, video generation, and perceptual feedback in a closed loop. The graph supports joint saving and restoration of rules and world states, while dynamic object binding grounds rule-specified entities in visual instances and uses stored states to constrain generation from their first appearance. We further introduce object attribute state replacement and counterfactual reconstruction training, enabling accurate attribute control and compatible appearance and motion changes while limiting interference with unrelated content. We develop an automated data collection and annotation pipeline using GTA-V and Habitat to construct RuleVerse, a dataset of rule-driven interactions with detailed state and event annotations, including cases where rules take effect before affected objects become visible. Extensive experiments demonstrate the effectiveness of WorldRuler in preserving rule-state consistency across saving and loading and accurately realizing rule outcomes on designated objects and attributes.