⌁ Awesome World Model Hub Papers · Datasets · Projects
中文 GitHub
← Back to papers

EvolvingWorld: An Open-Schema Framework for Co-Evolving Role-Play Agents and World Model in Interactive Literary World

arXiv 2026 43.2 system, benchmark, application

TLDR

EvolvingWorld introduces an open-schema framework for co-evolving characters and world models in interactive literary simulations, with a dataset and evaluation protocol.

Reasoning

The paper presents a novel approach to long-horizon literary simulation with an LLM-based world model, supported by a substantial dataset and multi-dimensional evaluation. However, its focus on text-based literary worlds limits relevance to video or reinforcement learning applications.

Read-first score

Read-first score 43.2, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 35.

Recency 6%
100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 18%
90

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=benchmark,dataset,evaluation,experiment,metric

Topical relevance 29%
50

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 18%
38

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=dataset

Citation impact 18%
0

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 12%
0

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 531.

Keyword Scores

world model
9
interactive world model
8
world simulator
7
generative world model
6
world dynamics prediction
5
video world model
0
model-based reinforcement learning world model
0

Deep Analysis

Innovations

  • Long-horizon co-evolution of characters and world with persistent state updates across interactions.
  • Open-schema framework that avoids fixed schemas, enabling simulation across diverse literary worlds.
  • Dual-module architecture: Character Agent for multi-agent role-play and profile evolution, and LLM-based World Model for global and entity-level state management.
  • Formulation of 7 trainable tasks for scene initialization, interaction generation, and state update, supporting supervised learning.
  • Benchmark dataset from 57 books with 138,596 training samples and 222 test snapshots, plus trajectory-level LLM-as-Judge evaluation with 10 dimensions and 20 metrics.

Methodology

EvolvingWorld consists of a Character Agent for multi-character role-play and persistent profile evolution, and an LLM-based World Model for maintaining global and location/entity-level states and driving scene progression. The framework defines 7 trainable tasks for scene initialization, interaction generation, and state update, trained on a dataset of 138,596 samples from 57 books. Evaluation uses a trajectory-level LLM-as-Judge protocol with 10 dimensions and 20 metrics.

Key Results

EvolvingWorld improves long-horizon simulation by maintaining persistent, coherent character and world development over time.

Resources

Paper

Tags