AMAS: Adaptively Determining Communication Topology for LLM-based Multi-Agent System
Hui Yi Leong, Yuheng Li, Yuqing Wu, Wenwen Ouyang, Wei Zhu, Jiechao Gao, Wei Han
TL;DR
AMAS tackles inflexible graph topologies in LLM-based multi-agent systems by introducing a per-sample dynamic graph designer. The designer uses parameter-efficient adaptation via LoRA to score candidate graphs and uses a ranking loss to align predicted scores with observed performance. Across domains including open-domain question answering, formal mathematics, and code generation, AMAS consistently outperforms single-agent and static MAS architectures, with latency comparable to existing approaches. This work shows that context-sensitive structural adaptability is a foundational requirement for robust, scalable LLM-driven multi-agent systems.
Abstract
Although large language models (LLMs) have revolutionized natural language processing capabilities, their practical implementation as autonomous multi-agent systems (MAS) for industrial problem-solving encounters persistent barriers. Conventional MAS architectures are fundamentally restricted by inflexible, hand-crafted graph topologies that lack contextual responsiveness, resulting in diminished efficacy across varied academic and commercial workloads. To surmount these constraints, we introduce AMAS, a paradigm-shifting framework that redefines LLM-based MAS through a novel dynamic graph designer. This component autonomously identifies task-specific optimal graph configurations via lightweight LLM adaptation, eliminating the reliance on monolithic, universally applied structural templates. Instead, AMAS exploits the intrinsic properties of individual inputs to intelligently direct query trajectories through task-optimized agent pathways. Rigorous validation across question answering, mathematical deduction, and code generation benchmarks confirms that AMAS systematically exceeds state-of-the-art single-agent and multi-agent approaches across diverse LLM architectures. Our investigation establishes that context-sensitive structural adaptability constitutes a foundational requirement for high-performance LLM MAS deployments.