Specification-Driven Generation and Evaluation of Discrete-Event World Models via the DEVS Formalism

Researchers propose a new approach to world models that combines explicit simulators with learned models using the DEVS formalism. The method uses LLMs to generate discrete-event world models from natural language specifications, targeting environments with event-driven dynamics like queueing systems and multi-agent coordination.

• →New world modeling approach bridges gap between rigid hand-engineered simulators and flexible but unreliable neural models.
• →Uses DEVS formalism with LLM-based pipeline to generate executable discrete-event models from natural language specifications.
• →Targets environments governed by discrete events like queueing systems, task planning, and multi-agent coordination.
• →Models can be validated through structured event traces against specification-derived constraints for verification.
• →Enables consistent long-horizon rollouts while maintaining adaptability during online execution.