Hugging Face researchers have unveiled EvolvingWorld, a novel framework and benchmark designed to model the co-evolution of characters and worlds in interactive literary simulations. Unlike prior systems that treat literary simulation as static persona imitation or isolated scene generation, EvolvingWorld captures how characters and worlds evolve together over time through a long-horizon process.
The framework adopts an open-schema approach, enabling support for diverse literary worlds without relying on fixed schemas. It comprises two coupled modules: a Character Agent for multi-character role-play and persistent profile evolution, and an LLM-based World Model for global and location/entity-level state maintenance and scene progression.
Based on this architecture, the researchers formulated 7 trainable tasks covering scene initialization, interaction generation, and state update. A dataset was constructed from 57 books, yielding 138,596 supervised training samples and 222 snapshots for testing. Additionally, a trajectory-level LLM-as-Judge evaluation protocol was introduced, spanning 10 dimensions and 20 metrics.
Experiments demonstrate that EvolvingWorld improves long-horizon simulation by effectively maintaining persistent, coherent character and world development. The work addresses a key gap in interactive literary simulation, moving beyond static persona imitation to dynamic co-evolution.