From Traces to Agentic Worlds: Agentic Language World Models for Interactive Environment Simulation

2026-10-05

AI Summary

This study addresses the challenge of constructing faithful replicas of real-world environments for training and evaluating LLM agents when the original systems are inaccessible. To this end, it proposes Trace2Env, a learning-free language world modeling framework. By leveraging trajectory reconstruction techniques, Trace2Env extracts schemas and evidence from historical interactions to generate a "world book." Combined with persistent episodic memory, this mechanism enables agents to simulate environment dynamics and state transitions, achieving high-fidelity, stateful interactive simulation without rebuilding executable systems. Experimental results demonstrate that the proposed approach significantly improves next-observation prediction fidelity and long-horizon consistency across nine diverse environments. Furthermore, it substantially enhances the validity of multi-turn interactive actions when replayed in real-world settings.

Abstract

Realistic environment replicas are increasingly valuable for training and evaluating LLM agents, yet the original systems may be inaccessible or impractical to reproduce. We explore agentic language world modeling: rather than rebuilding an executable environment, a world model agent serves as the environment for a task agent and supports faithful and stateful simulation. We instantiate this paradigm with Trace2Env, a learning-free framework for settings where the original system is unavailable but historical interaction traces remain accessible. Trace2Env reconstructs these traces into a reusable environment worldbook containing environment schemas, grounded evidence, and induced behavioral knowledge. At runtime, the world model agent actively consults the worldbook together with persistent episodic state to infer each action's observation and lasting state effects. Across nine environments, Trace2Env improves both next-observation fidelity and long-horizon interaction consistency over conventional prompt-based LWMs. In multi-turn interaction, task agent actions generated against Trace2Env remain valid more often when replayed in the real environment, indicating that its simulated dynamics better preserve the consequences of earlier actions across successive turns. These results establish agentic language world modeling as an alternative direction for building realistic environment replicas without reconstructing the original executable system.

Problem

• Language World Models
• Environment Simulation
• LLM Agents
• Interaction Traces

Innovation

• Agentic Language World Models
• Trace2Env
• Environment Simulation
• Worldbook
• Learning-free Framework