Multi-agent system improves goal tracking in complex tasks
Goal-Aware Dynamic Multi-Agent Collaboration for Long-Horizon Tasks
Artificial Intelligence
Summary
When computers try to do big, complicated jobs that require many steps, they often get stuck checking their work too late or mixing up important details. The authors studied how single computers handle different duties like planning and checking their tasks, and found that checking too late causes mistakes. They then tested using several computers that each focus on different duties but found that without clear coordination, they lose track of the overall goal and repeat work. To fix this, they created a system where multiple computers share a clear structure of tasks and goals, communicate dependencies carefully, and check each other's work to catch problems early.
What this means in practice
- •For robotics development teams: Coordinate multiple robots to divide complex tasks and verify progress to ensure long-term goals are met efficiently.
- •For software automation engineers: Design software agents that dynamically allocate planning, verification, and recovery duties to improve automated task completion.
Authors
Xu Shen, Lingyong Yan, Shuaiqiang Wang, Jianmin Wu, Zhikai Chen, Jiliang Tang, Jie Ren
Abstract
LLM-based agents are increasingly used to complete complex, long-horizon tasks without human intervention. To make progress toward the task goal, agents need to perform different duties, such as planning and verification, when needed. To understand how single agents manage these duties during long-horizon execution, we conduct a preliminary study. Our analysis shows that verification has the strongest association with task success, but often occurs late, leaving limited opportunities for correction. We also find that the proportion of duty-relevant information decreases as records from different duties accumulate in a shared context. These findings lead us to examine whether multi-agent systems (MAS) can address these problems by assigning duties to agents with separate contexts. Our analysis shows that this separation alone is insufficient. Agents can lose track of the overall goal and repeat prior investigations. They may also lack necessary background while receiving irrelevant information from others. To maintain goal awareness and coordinate collaboration as task needs change, we propose Goal-Aware Dynamic Multi-Agent Collaboration (GAD). It uses a growing execution graph to organize subtasks and their dependencies. To keep agents focused on their goals, each context retains the original instruction and the assigned task's goal and acceptance conditions. To adapt collaboration and information sharing during execution, agents can divide tasks into subtasks and specify dependencies based on execution evidence. These dependencies guide when agents work and which results they receive. Verification and recovery check subtask results and full task completion against their goals, providing feedback for repair when checks fail.