Paper page - Agent World Model in the Wild: Computer-Use Agent Environment Synthesis

Source: https://huggingface.co/papers/2610.26828

Abstract

WildAWM synthesizes controllable computer-use environments and tasks to train general agents via interaction, improving out-of-distribution performance across web, CLI, and GUI benchmarks.

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 proposeAgent World Model in the Wild(WildAWM), a scalable pipeline for synthesizing controllable environments and tasks for training generalcomputer-use agents(CUAs). Whereas existing approaches typically specialize in a single agent interaction interface, we construct executable environments that jointly supportgraphical user interfaces,model context protocolservices, andcommand-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 introducingexecution-based quality gatesthat combineheterogeneous agent rollouts,cross-judge consistency checks, andcounterfactual probesof 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 twoQwen modelsof different sizes exclusively in these synthetic worlds. Both models achieve substantial gains across 8 out-of-distribution benchmarks spanning web, claw, andcomputer-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.

View arXiv pageView PDFProject pageGitHub69Add to collection

Get this paper in your agent:

hf papers read 2610\.26828

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper2

#### ChilleD/WildAWM-35B-A3B 36B• Updated3 days ago • 42 #### ChilleD/WildAWM-4B 5B• Updated3 days ago • 38

Datasets citing this paper1

#### ChilleD/WildAWM Preview• Updated2 days ago • 274 • 1

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2610.26828 in a Space README.md to link it from this page.

Collections including this paper1