Understanding Rollout Error in Graph World Models

2026-06-26

AI Summary

This work addresses the error accumulation and planning failure in long-horizon rollout predictions of graph-based world models, which stem from evolving topologies and dynamic edge structures. To mitigate these issues, the authors propose a unified modeling framework that jointly handles static and dynamic edges, incorporates action nodes to support multi-granular decision-making, and, for the first time, disentangles topology-induced from model-induced error amplification mechanisms. By decomposing rollout errors through graph-value bounds, they devise an error-aware training strategy integrating spectral regularization, rollout consistency constraints, critical node weighting, and joint dynamic edge modeling. The resulting approach significantly reduces both prediction errors and planning regret, outperforming existing specialized graph models in both synthetic and real-world heterogeneous multi-agent environments.

Abstract

World models are often used for planning by rolling learned dynamics forward. Many planning environments, however, are not vectors or images; they are graphs of agents, tools, skills, routes, and dependencies. In these settings, a local prediction error may stay local or spread through the graph, and the failure mode changes again when edges are predicted rather than fixed. This paper studies long-horizon rollout error in Graph World Models (GWMs). We formulate a unified fixed-edge and dynamic-edge GWM framework with action nodes for node-, edge-, and graph-level decisions. We develop graph-valued rollout bounds that separate topology-induced amplification from model-induced amplification, and we introduce a joint node-edge operator for dynamic-edge rollouts. Guided by the analysis, we propose Error-Aware GWM, which combines spectral regularization, rollout consistency, and critical-node weighting. Across synthetic topologies and heterogeneous agent-graph testbeds, rollout error and planning regret grow with horizon, dynamic-edge training is needed when structure evolves, and Error-Aware GWM prevents long-horizon divergence while preserving prediction accuracy. Real-world graph benchmarks clarify the scope of GWMs: they are most useful for dynamic graph rollout and agent planning, while specialized graph models remain strong on static or sparse prediction tasks.

Problem

• rollout error
• graph world models
• dynamic-edge graphs
• long-horizon planning
• error propagation

Innovation

• Graph World Models
• rollout error
• dynamic-edge graphs
• spectral regularization
• error-aware planning