CoAct: A Global-Local Hierarchy for Autonomous Agent Collaboration

TL;DR

The paper introduces CoAct, a hierarchical framework for LLMs that mimics human-like global and local planning to improve performance on complex, long-horizon web tasks, especially when facing failures.

Contribution

CoAct is the first framework to incorporate hierarchical global-local planning and collaboration in LLM systems for complex task execution.

Findings

Outperforms baseline methods on WebArena benchmark

Re-arranges process trajectory effectively when facing failures

Achieves superior performance on long-horizon web tasks

Abstract

Existing LLMs exhibit remarkable performance on various NLP tasks, but still struggle with complex real-world tasks, even equipped with advanced strategies like CoT and ReAct. In this work, we propose the CoAct framework, which transfers the hierarchical planning and collaboration patterns in human society to LLM systems. Specifically, our CoAct framework involves two agents: (1) A global planning agent, to comprehend the problem scope, formulate macro-level plans and provide detailed sub-task descriptions to local execution agents, which serves as the initial rendition of a global plan. (2) A local execution agent, to operate within the multi-tier task execution structure, focusing on detailed execution and implementation of specific tasks within the global plan. Experimental results on the WebArena benchmark show that CoAct can re-arrange the process trajectory when facing failures, and achieves superior performance over baseline methods on long-horizon web tasks. Code is available at https://github.com/xmhou2002/CoAct.