Sunday, 11 October 2026 · 7 min read
Trace2Env simulates environments from interaction traces without original systems
Trace2Env reconstructs environment knowledge from historical interaction traces into a structured worldbook and uses an agentic language world model to simulate faithful, stateful environment responses, improving next-observation fidelity and long-horizon consistency across nine environments.
Based on the paper From Traces to Agentic Worlds: Agentic Language World Models for Interactive Environment Simulation by Quanyu Long, Xiao Chen, Jianda Chen, Haozhen Zhang and 3 others
Challenges in simulating environments without original systems
Interactive agents require environments that respond consistently to their actions over time, but reproducing original systems like terminals or enterprise software may be impossible due to unavailable code or infrastructure. However, historical interaction traces recording sequences of actions and observations often remain accessible, offering a potential source to reconstruct environment behavior.
Simulating environments faithfully demands preserving long-term state continuity, since some actions have delayed effects not immediately observable. Conventional language world models predict the next observation from a flattened prompt containing environment description and interaction history, which struggles to maintain consistent state across multiple turns, limiting their ability to simulate realistic multi-turn interactions.
Sign in to read the whole paper
You have read the opening. The rest is free with a Bezen account, and no card is needed.
In the rest of this paper
- How the method works, step by step, with a flow chart
- Three key formulas from the paper, explained symbol by symbol
- Three figures from the paper, and how to read them
- How it was tested, what was found, and where it falls short