Agent World Model in the Wild: Computer-Use Agent Environment Synthesis
Zhaoyang Wang, Qianhui Wu, Baolin Peng, Jianfeng Gao, Huaxiu Yao and colleagues (UNC-Chapel Hill, Microsoft Research, Georgia Tech, Northeastern) introduce WildAWM, a pipeline built only on open-source models that synthesizes executable computer-use environments exposing GUI, MCP and CLI interfaces together, plus verified long-horizon tasks.
Ask this paper
Three interfaces in one world. Each synthesized environment supports visual interaction, MCP tool calls and shell commands at once, so a trained agent can choose between clicking, calling a tool or writing code for the same task.
Quality gates. Solvability, difficulty and verifier reliability are checked with heterogeneous agent rollouts, cross-judge consistency checks and counterfactual probes of each judge.
Difficulty. The pipeline yields 341 environment-task pairs, some needing more than 100 steps. GPT-6-Astra solves only 60% of them.
Transfer. Two Qwen models trained only in these synthetic worlds improve on 8 out-of-distribution benchmarks covering web, claw and computer-use agents.
Abstract
Agent learning is shifting from learning from human data toward learning through interaction with the world. However, scaling such learning directly in real-world computer systems is challenging and can introduce substantial security risks. We therefore propose Agent World Model in the Wild (WildAWM), a scalable pipeline for synthesizing controllable environments and tasks for training general computer-use agents (CUAs). Whereas existing approaches typically specialize in a single agent interaction interface, we construct executable environments that jointly support graphical user interfaces, model context protocol services, and command-line interfaces, allowing agents to flexibly combine visual interaction, tool use, and code execution. WildAWM is entirely powered by open-source models, demonstrating that high-quality computer-use environments can be generated without proprietary frontier models. Beyond environments, we address three recurring challenges in task synthesis: ensuring solvability, maintaining sufficient difficulty, and producing reliable verifiers, by introducing execution-based quality gates that combine heterogeneous agent rollouts, cross-judge consistency checks, and counterfactual probes of judgment robustness. The resulting pipeline synthesizes 341 long-horizon environment-task pairs that can require over 100 agent steps to solve and remain challenging even for frontier models, with GPT-6-Astra solving only 60% of them. We train two Qwen models of different sizes exclusively in these synthetic worlds. Both models achieve substantial gains across 8 out-of-distribution benchmarks spanning web, claw, and computer-use agents, demonstrating that learning in WildAWM's synthesized worlds generalizes broadly to unseen computer-use tasks. The code is publicly available at https://github.com/MSR-Orchard/WildAWM.