AgentStore: Scalable Integration of Heterogeneous Agents As Specialized Generalist Computer Assistant
Chengyou Jia, Minnan Luo, Zhuohang Dang, Qiushi Sun, Fangzhi Xu, Junlin Hu, Tianbao Xie, Zhiyong Wu
2024-10-29
Summary
This paper introduces AgentStore, a platform designed to integrate various digital agents that can automate complex computer tasks, making them more efficient and adaptable.
What's the problem?
Current digital agents often struggle to perform well in diverse and complex tasks because they lack the ability to generalize their skills across different situations. This limitation makes it difficult for them to handle open-ended tasks in real-world environments, which can lead to inefficiencies in human-computer interactions.
What's the solution?
The authors present AgentStore, a scalable platform that allows users to integrate different types of agents, each with specialized skills. This system is inspired by app stores, where new functionalities can be added easily. They introduce a core component called MetaAgent, which uses an AgentToken strategy to manage these diverse agents effectively. Through extensive testing on various benchmarks, AgentStore has shown significant improvements in performance, particularly in completing complex tasks more efficiently than previous systems.
Why it matters?
This research is important because it represents a step forward in creating more capable and flexible digital assistants. By allowing for the integration of specialized agents, AgentStore can enhance productivity and streamline workflows in various applications, making technology more accessible and effective for users.
Abstract
Digital agents capable of automating complex computer tasks have attracted considerable attention due to their immense potential to enhance human-computer interaction. However, existing agent methods exhibit deficiencies in their generalization and specialization capabilities, especially in handling open-ended computer tasks in real-world environments. Inspired by the rich functionality of the App store, we present AgentStore, a scalable platform designed to dynamically integrate heterogeneous agents for automating computer tasks. AgentStore empowers users to integrate third-party agents, allowing the system to continuously enrich its capabilities and adapt to rapidly evolving operating systems. Additionally, we propose a novel core MetaAgent with the AgentToken strategy to efficiently manage diverse agents and utilize their specialized and generalist abilities for both domain-specific and system-wide tasks. Extensive experiments on three challenging benchmarks demonstrate that AgentStore surpasses the limitations of previous systems with narrow capabilities, particularly achieving a significant improvement from 11.21\% to 23.85\% on the OSWorld benchmark, more than doubling the previous results. Comprehensive quantitative and qualitative results further demonstrate AgentStore's ability to enhance agent systems in both generalization and specialization, underscoring its potential for developing the specialized generalist computer assistant. All our codes will be made publicly available in https://chengyou-jia.github.io/AgentStore-Home.