Pub-AI: AI in Science

JEPA · 2026-09-30

WorldGraph: Graph-Native World Modeling

Zezhong Ding, Yipeng Li, Xike Xie

arXiv:2609.34159PDF

Auto-summarized: this summary was generated by a language model from the paper’s text and has not been reviewed by an editor. Check the paper before relying on it. Benchmark numbers appear only after a human has verified them.

TL;DR

WorldGraph studies graph world modeling (GWM) where the evolving graph itself is the world dynamics. It formulates latent graph states and heterogeneous transition prediction over node-, edge-, and graph-level changes, builds the GWM-Zero benchmark (8 temporal graph datasets), and proposes WorldGraph with a state-aware graph transformer and a transition-aware GRPO (dynamic grouping, structure-aware verifiable rewards).

Why it matters

It targets environments where entities/relations/attributes evolve and are observed as graphs, proposing a graph-native world model that explicitly models latent world state and predicts heterogeneous future graph transitions at multiple granularities using a unified framework.

Method

  • Formulate GWM over observed graph evolution (g_t, Δg_t) and learn latent world states s_t plus a transition model predicting next graph change, instantiated at node-, edge-, and graph-level granularities.
  • State-aware graph transformer: fuse hop-level message passing and random-walk path sampling, then integrate history-aware temporal-distance and transition-conditioned attention to build s_t.
  • Transition-aware GRPO: use dynamic grouping biased to rare historical transitions and structure-aware verifiable rewards combining structural importance and historical variability for node-/edge-/graph-level tasks.

Limitation

K.1 Limitations: further work is needed to better integrate information from diverse modalities, to more fully realize the potential of graph world models.

Abstract (from arXiv)

World models infer latent states of an environment to capture its underlying dynamics and predict future evolution. Many real-world environments, however, are inherently relational and observed as evolving graphs, where entities, relations, and their properties change over time. Prior graph-related world models use graph structures to organize internal states or support task-specific reasoning, rather than treating an evolving graph itself as the modeled world. We instead study graph world modeling (GWM), where graph evolution itself constitutes the world dynamics. We formulate graph world modeling over observed graph evolution, latent graph states, and heterogeneous graph-transition predictions. Based on this formulation, we construct GWM-Zero, a benchmark covering node-, edge-, and graph-level transitions over eight temporal graph datasets. We propose WorldGraph, which combines a state-aware graph transformer for multi-granularity structural and transition-conditioned evolution modeling with transition-aware GRPO using dynamic grouping and structure-aware verifiable rewards. Extensive experiments on GWM-Zero show that WorldGraph consistently outperforms representative graph representation, temporal graph learning, graph pretraining, and graph world-model baselines across all three transition granularities.

Related papers