Language models tested on keeping virtual worlds running correctly
World Agent: Can Language Models Keep a World Running?
Artificial Intelligence
Summary
Creating virtual worlds is easier than making sure they keep working over time. The authors introduce a new way to test if language models can manage ongoing events and their consequences in these worlds. They built a benchmark with two tasks: one for maintaining the world's state correctly, and another for predicting how the world changes with partial information. Their tests show that current models struggle most with understanding cause-and-effect between events over time. This benchmark helps measure whether a world model just completes small steps well or can handle long-term world operation.
What this means in practice
- •For game developers: Evaluate and improve AI systems that control ongoing in-game worlds by testing how well they maintain world consistency over time.
- •For simulation engineers: Test AI-driven virtual environments where maintaining correct event sequences under partial information is critical for realistic outcomes.
Authors
Weixing Chen, Weipeng Zhang, Nan An, Yang Liu, Liang Lin
Abstract
World models are moving from generating realistic frames to generating playable worlds, yet whether a delivered world can keep running is not tested anywhere. Existing evaluations stop at generation, at delivery, or at single-step transitions, and each stops at a different point along the way. Correct local state transitions or intermediate outcomes do not guarantee a correctly organized causal event flow. We propose the world agent task, which moves the evaluation point of world generation from the moment of delivery to the continued operation that follows. In this task, a model is not asked to generate a world. It is held responsible for keeping the world running, which requires coordinating events and carrying forward their consequences to constrain subsequent evolution. We instantiate the task in WorldAgent-Benchmark with two complementary tracks. In the maintenance track, the model must ground the events of a continuous narrative into correct transitions of the explicit world state while respecting causal, temporal, and concurrency constraints. In the deduction track, the model must predict how the world will evolve under partial observations and act toward a goal. The maintenance track combines LLM-assisted semantic judgments with programmatic validation and scoring, while the deduction track is evaluated entirely programmatically. Individual judgments are auditable against world states and execution logs, and scores can be recomputed from the saved judgments and execution records. Across 8 models, scores decline steadily as pre-built structure is removed from the world, and causal-relation checking is the weakest component for every model. The benchmark makes the continued operation of a world measurable and distinguishes local completion from failures in event organization. Code and dataset will be released on https://github.com/HCPLab-SYSU/WorldAgent-Benchmark.