CompoWorld scales AI agents by linking multiple service environments
CompoWorld: Compositional Environment Scaling for General Agents
Artificial Intelligence
Summary
Most AI training uses tasks within just one kind of environment, but real-world problems often need connecting different services. The authors create CompoWorld, which builds complex tasks by combining many small, reusable services. Their method verifies that information flows correctly across tasks and trains AI agents that can successfully complete these linked tasks. This approach improved AI performance on various benchmarks and outperformed other leading models on a specific automation test.
What this means in practice
- •For automation engineers: Create AI systems that automate workflows spanning multiple software services by composing verified subtasks into overall tasks.
- •For software developers: Develop AI assistants that reliably use multiple APIs and tools together to complete complex user requests.
- •For robotic process automation (rpa) vendors: Build AI-powered products that improve cross-service task handling by training agents on compositional, verified task trajectories.$Commercial implications: Enables selling more capable intelligent automation products that reliably integrate multiple IT services into workflows.
Authors
Xiao-Wen Yang, Weiyi Xu, Wen Da, Hang Xu, Canwei Li, Hong-Jie You, Pusen Dong, Yucheng Zeng, Zhaokai Luo, Yu-Feng Li, Yao Hu, Mu Chuan
Abstract
Automatically generated environments provide a scalable source of interaction data for training general agents. However, existing approaches mainly generate tasks within a single environment, while real-world workflows require agents to connect information and actions across multiple services. We introduce Compositional Environment Scaling (\textbf{CompoWorld}), which expands the task space by composing a finite library of reusable services. Coding agents turn tool specifications into verified services with typed states and shared interfaces, while a world model handles tools that cannot be reliably implemented. A random-walk procedure connects services through dependency graphs, enabling the generation and verification of tasks that require information to flow across services. Verified trajectories support supervised fine-tuning (SFT), while our Completion-Focused Rubric Reward guides reinforcement learning (RL) toward full task completion by emphasizing criteria with lower pass rates within each rollout group. We construct 448 services exposing 10,130 tools and use 3K SFT trajectories and 1K RL tasks to train Qwen3.6-35B-A3B. Experimental results show that CompoWorld improves on its backbone by 9.17 points on average across eight benchmarks. On AutomationBench, it surpasses frontier models such as Claude Opus 4.6 and leads all compared agent-specialized 35B-A3B models.