DynaSwarm: Dynamically Graph Structure Selection for LLM-based Multi-agent System

2025/07/31 by Hui Yi Leong, Yuqing Wu, Leong, Hui Yi +1
Computer Science · #Advanced Graph Neural Networks #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Graph Theory and Algorithms #Machine Learning (cs.LG) #Multiagent Systems (cs.MA) #Topic Modeling

paper · pdf · doi:10.48550/arxiv.2507.23261

openalex publication_date 2025/07/31 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Current multi-agent systems (MAS) frameworks often rely on manually designed and static collaboration graph structures, limiting adaptability and performance. To address these limitations, we propose DynaSwarm, a dynamic framework that enhances LLM-based MAS through two key innovations: (1) an actor-critic reinforcement learning (A2C) mechanism to optimize graph structures with improved stability over prior RL methods, and (2) a dynamic graph selector that adaptively chooses the optimal graph structure for each input sample via parameter-efficient LLM fine-tuning. DynaSwarm eliminates the need for rigid, one-fits-all graph architectures, instead leveraging sample-specific idiosyncrasies to dynamically route queries through specialized agent networks. (c) We propose to fine-tune the demonstration retriever to fully exploit the power of in-context learning (ICL). Extensive experiments on question answering, mathematical reasoning, and coding tasks demonstrate that DynaSwarm consistently outperforms state-of-the-art single-agent and MAS baselines across multiple LLM backbones. Our findings highlight the importance of sample-aware structural flexibility in LLM MAS designs.

Citations

• The Llama 3 Herd of Models
• PromptCBLUE: A Chinese Prompt Tuning Benchmark for the Medical Domain
• AgentVerse: Facilitating Multi-Agent Collaboration and Exploring Emergent Behaviors
• Graph of Thoughts: Solving Elaborate Problems with Large Language Models
• MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework
• Auto-GPT for Online Decision Making: Benchmarks and Additional Opinions
• Mindstorms in Natural Language-Based Societies of Mind
• Unified Demonstration Retriever for In-Context Learning
• Generative Agents: Interactive Simulacra of Human Behavior
• CAMEL: Communicative Agents for "Mind" Exploration of Large Language Model Society
• Is ChatGPT a General-Purpose Natural Language Processing Task Solver?
• Self-Instruct: Aligning Language Models with Self-Generated Instructions
• Late Prompt Tuning: A Late Prompt Could Be Better Than Many Prompts
• Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them
• Distributing Accountability, Not Capability: Phase Separation and the LLM Workflow Quadrant in Autonomous AI Agent Architectures
• Socratic Models: Composing Zero-Shot Multimodal Reasoning with Language
• Training language models to follow instructions with human feedback
• Learning To Retrieve Prompts for In-Context Learning
• Evaluating Large Language Models Trained on Code
• LoRA: Low-Rank Adaptation of Large Language Models
• Measuring Massive Multitask Language Understanding
• Language Models are Few-Shot Learners
• Asynchronous Methods for Deep Reinforcement Learning

Related

• MasHost Builds It All: Autonomous Multi-Agent System Directed by Reinforcement Learning
• AMAS: Adaptively Determining Communication Topology for LLM-based Multi-Agent System
• Enhancing Multi-Agent Systems via Reinforcement Learning with LLM-based Planner and Graph-based Policy
• Graph-Enhanced Policy Optimization in LLM Agent Training
• Graph Counselor: Adaptive Graph Exploration via Multi-Agent Synergy to Enhance LLM Reasoning