A Dynamic LLM-Powered Agent Network for Task-Oriented Agent Collaboration
2023/10/03 by Zijun Liu, Liu, Zijun, Yanzhe Zhang +7 · 148 citations
Computer Science · #Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #FOS: Computer and information sciences #Multiagent Systems (cs.MA) #Multimodal Machine Learning Applications #Natural Language Processing Techniques #Topic Modeling
paper · pdf · doi:10.48550/arxiv.2310.02170
openalex publication_date 2023/10/03 · openalex created_date 2023/10/05 · openalex updated_date 2026/07/28
Abstract
Recent studies show that collaborating multiple large language model (LLM) powered agents is a promising way for task solving. However, current approaches are constrained by using a fixed number of agents and static communication structures. In this work, we propose automatically selecting a team of agents from candidates to collaborate in a dynamic communication structure toward different tasks and domains. Specifically, we build a framework named Dynamic LLM-Powered Agent Network (DyLAN) for LLM-powered agent collaboration, operating a two-stage paradigm: (1) Team Optimization and (2) Task Solving. During the first stage, we utilize an agent selection algorithm, based on an unsupervised metric called Agent Importance Score, enabling the selection of best agents according to their contributions in a preliminary trial, oriented to the given task. Then, in the second stage, the selected agents collaborate dynamically according to the query. Empirically, we demonstrate that DyLAN outperforms strong baselines in code generation, decision-making, general reasoning, and arithmetic reasoning tasks with moderate computational cost. On specific subjects in MMLU, selecting a team of agents in the team optimization stage improves accuracy by up to 25.0% in DyLAN.
Cited by
- UCAgents: Unidirectional Convergence for Visual Evidence Anchored Multi-Agent Medical Decision-Making
- From Cognitive Architectures to Language Agents: A Mechanism-Level Review of Lineage, Convergence, and Migration Gaps
- Focus Is All You Need: Adaptive Goal-aware Attention Orchestration for Multi-Agent Graph Systems
- Toward an Organizational Science of Multi-Agent LLM Systems: Decoupling Who, How, and Which Algorithm
- Gleam: Adaptive Network-Efficient CUDA API Remoting for Cross-Device GPU Sharing over LANs
- One Tool Is Enough: Reinforcement Learning for Repository-Level LLM Agents
- Rethinking Multi-Agent Intelligence Through the Lens of Small-World Networks
- FutureWeaver: Planning Test-Time Compute for Multi-Agent Systems with Modularized Collaboration
- Multi-LLM Collaboration for Medication Recommendation
- MedEinst: Benchmarking the Einstellung Effect in Medical LLMs through Counterfactual Differential Diagnosis
- RoCo: Role-Based LLMs Collaboration for Automatic Heuristic Design
- Automating modeling in mechanics: LLMs as designers of physics-constrained neural networks for constitutive modeling of materials
- Augmented Runtime Collaboration for Self-Organizing Multi-Agent Systems: A Hybrid Bi-Criteria Routing Approach
- Toward a Safe Internet of Agents
- Multi-Agent Systems for Dataset Adaptation in Software Engineering: Capabilities, Limitations, and Future Directions
- A2Flow: Automating Agentic Workflow Generation via Self-Adaptive Abstraction Operators
- Cost-Effective Communication: An Auction-based Method for Language Agent Interaction
- Designing LLM-based Multi-Agent Systems for Software Engineering Tasks: Quality Attributes, Design Patterns and Rationale
- S-DAG: A Subject-Based Directed Acyclic Graph for Multi-Agent Heterogeneous Reasoning
- Maestro: Learning to Collaborate via Conditional Listwise Policy Optimization for Multi-Agent LLMs
- OptiMA: A Transaction-Based Framework with Throughput Optimization for Very Complex Multi-Agent Systems
- A Criminology of Machines
- Optimal-Agent-Selection: State-Aware Routing Framework for Efficient Multi-Agent Collaboration
- The Geometry of Dialogue: Graphing Language Models to Reveal Synergistic Teams for Multi-Agent Collaboration
- Generalizing Test-time Compute-optimal Scaling as an Optimizable Graph
- Multi-Agent Debate Strategies: Survey, Taxonomy, and Challenges
- Enhancing Multi-Agent Communication through Attention Steering with Context Relevance
- Automated Composition of Agents: A Knapsack Approach for Agentic Component Selection
- MAS-ProVe: Understanding the Process Verification of Multi-Agent Systems
- MASPRM: Multi-Agent System Process Reward Model
- Towards AI as Colleagues: Multi-Agent System Improves Structured Professional Ideation
- Agentic AI Security: Threats, Defenses, Evaluation, and Open Challenges
- Alita-G: Self-Evolving Generative Agent for Agent Generation
- Towards Scalable Oversight via Partitioned Human Supervision
- Agent-GSPO: Communication-Efficient Multi-Agent Systems via Group Sequence Policy Optimization
- Thought Communication in Multiagent Collaboration
- Multi-dimensional Data Analysis and Applications Basing on LLM Agents and Knowledge Graph Interactions
- Evolution in Simulation: AI-Agent School with Dual Memory for High-Fidelity Educational Dynamics
- Collaborative Shadows: Distributed Backdoor Attacks in LLM-Based Multi-Agent Systems
- HyperAgent: Leveraging Hypergraphs for Topology Optimization in Multi-Agent Communication
- Failure-Driven Workflow Refinement
- Effective Strategies for Asynchronous Software Engineering Agents
- BILLY: Steering Large Language Models via Merging Persona Vectors for Creative Generation
- ATLAS: Adaptive Trading with LLM AgentS Through Dynamic Prompt Optimization and Multi-Agent Coordination
- Measuring and Mitigating Identity Bias in Multi-Agent Debate via Anonymization
- Belief-Calibrated Multi-Agent Consensus Seeking for Complex NLP Tasks
- SPOGW: a Score-based Preference Optimization method via Group-Wise comparison for workflows
- LLM Chemistry Estimation for Multi-LLM Recommendation
- Adversarial Agent Collaboration for Correctness Improvements of C to Safe Rust Translation
- Cache-to-Cache: Direct Semantic Communication Between Large Language Models
- JoyAgent-JDGenie: Technical Report on the GAIA
- Stochastic Self-Organization in Multi-Agent Systems
- DyFlow: Dynamic Workflow Framework for Agentic Reasoning
- MAS2: Self-Generative, Self-Configuring, Self-Rectifying Multi-Agent Systems
- MedLA: A Logic-Driven Multi-Agent Framework for Complex Medical Reasoning with Large Language Models
- Escaping the Self-Confirmation Trap: An Execute-Distill-Verify Paradigm for Agentic Experience Learning
- AgentInit: Initializing LLM-based Multi-Agent Systems via Diversity and Expertise Orchestration for Effective and Efficient Collaboration
- LLM-based Agents Suffer from Hallucinations: A Survey of Taxonomy, Methods, and Directions
- Towards Transparent and Incentive-Compatible Collaboration in Decentralized LLM Multi-Agent Systems: A Blockchain-Driven Approach
- Debate or Vote: Which Yields Better Decisions in Multi-Agent Large Language Models?
- Generalizability of Large Language Model-Based Agents: A Comprehensive Survey
- (P)rior(D)yna(F)low: A Priori Dynamic Workflow Construction via Multi-Agent Collaboration
- Aegis: Automated Error Generation and Attribution for Multi-Agent Systems
- Abduct, Act, Predict: Scaffolding Causal Inference for Automated Failure Attribution in Multi-Agent Systems
- Leveraging AI Agents for Autonomous Networks: A Reference Architecture and Empirical Studies
- Murakkab: Resource-Efficient Agentic Workflow Orchestration in Cloud Platforms
- DRF: LLM-AGENT Dynamic Reputation Filtering Framework
- Psychologically Enhanced AI Agents
- Learning to Deliberate: Meta-policy Collaboration for Agentic LLMs with Multi-agent Reinforcement Learning
- LLM-Assisted Iterative Evolution with Swarm Intelligence Toward SuperBrain
- HiVA: Self-organized Hierarchical Variable Agent via Goal-driven Semantic-Topological Evolution
- Generative Interfaces for Language Models
- CausalMACE: Causality Empowered Multi-Agents in Minecraft Cooperative Tasks
- Self-Organizing Agent Network for LLM-based Workflow Automation
- SafeSieve: From Heuristics to Experience in Progressive Pruning for LLM-based Multi-Agent Communication
- AdaptFlow: Adaptive Workflow Optimization via Meta-Learning
- Mediator-Guided Multi-Agent Collaboration among Open-Source Models for Medical Decision-Making
- An Auditable Agent Platform For Automated Molecular Optimisation
- Polymath: A Self-Optimizing Agent with Dynamic Hierarchical Workflow
- GasAgent: A Multi-Agent Framework for Automated Gas Optimization in Smart Contracts
- Graph-Augmented Large Language Model Agents: Current Progress and Future Prospects
- "X of Information'' Continuum: A Survey on AI-Driven Multi-dimensional Metrics for Next-Generation Networked Systems
- EH-Benchmark Ophthalmic Hallucination Benchmark and Agent-Driven Top-Down Traceable Reasoning Workflow
- From Skills to Talent: Organising Heterogeneous Agents as a Real-World Company
- SVAgent: AI Agent for Hardware Security Verification Assertion
- Agentic Neural Networks: Self-Evolving Multi-Agent Systems via Textual Backpropagation
- MasHost Builds It All: Autonomous Multi-Agent System Directed by Reinforcement Learning
- Graph World Model
- Introspection of Thought Helps AI Agents
- AgentsNet: Coordination and Collaborative Reasoning in Multi-Agent LLMs
- Adaptive collaboration for online personalized distributed learning with heterogeneous clients
- EvoAgentX: An Automated Framework for Evolving Agentic Workflows
- Optimas: Optimizing Compound AI Systems with Globally Aligned Local Rewards
- Exploring Advanced LLM Multi-Agent Systems Based on Blackboard Architecture
- M3MAD-Bench: Multi-Dimensional Evaluation of Multi-Agent Debate Across Domains and Modalities
- G-Memory: Tracing Hierarchical Memory for Multi-Agent Systems
- Compound AI Systems Optimization: A Survey of Methods, Challenges, and Future Directions
- MAM: Modular Multi-Agent Framework for Multi-Modal Medical Diagnosis via Role-Specialized Collaboration
- Graphs Meet AI Agents: Taxonomy, Progress, and Future Opportunities
- AnyMAC: Cascading Flexible Multi-Agent Collaboration via Next-Agent Prediction
- SwarmAgentic: Towards Fully Automated Agentic System Generation via Swarm Intelligence
- OAgents: An Empirical Study of Building Effective Agents
- United Minds or Isolated Agents? Exploring Coordination of LLMs under Cognitive Load Theory
- MedSyn: Enhancing Diagnostics with Human-AI Collaboration
- AgentSGEN: Multi-Agent LLM in the Loop for Semantic Collaboration and GENeration of Synthetic Data
- Adaptive Graph Pruning for Multi-Agent Communication
- Goal-Aware Identification and Rectification of Misinformation in Multi-Agent Systems
- RMoA: Optimizing Mixture-of-Agents through Diversity Maximization and Residual Compensation
- An Adversary-Resistant Multi-Agent LLM System via Credibility Scoring
- Enhancing LLM-Based Code Generation with Complexity Metrics: A Feedback-Driven Approach
- Cross-Task Experiential Learning on LLM-based Multi-Agent Collaboration
- Literature Review Of Multi-Agent Debate For Problem-Solving
- MermaidFlow: Redefining Agentic Workflow Generation via Safety-Constrained Evolutionary Programming
- Topological Structure Learning Should Be A Research Priority for LLM-Based Multi-Agent Systems
- Efficient Leave-one-out Approximation in LLM Multi-agent Debate Based on Introspection
- Seven Security Challenges in Cross-domain Multi-agent LLM Systems
- Scaling External Knowledge Input Beyond Context Windows of LLMs via Multi-Agent Collaboration
- Long Context Scaling: Divide and Conquer via Multi-Agent Question-driven Collaboration
- Multi-Agent Collaboration via Evolving Orchestration
- Multi-View Encoders for Performance Prediction in LLM-Based Agentic Workflows
- MA-RAG: Multi-Agent Retrieval-Augmented Generation via Collaborative Chain-of-Thought Reasoning
- CoTGuard: Using Chain-of-Thought Triggering for Copyright Protection in Multi-Agent LLM Systems
- GUARDIAN: Safeguarding LLM Multi-Agent Collaborations with Temporal Graph Modeling
- Global Optimization and Inference-Time Region Grafting for Agentic Workflows
- Optimizing LLM-Based Multi-Agent System with Textual Feedback: A Case Study on Software Development
- X-MAS: Towards Building Multi-Agent Systems with Heterogeneous LLMs
- MASLab: A Unified and Comprehensive Codebase for LLM-based Multi-Agent Systems
- MAS-ZERO: Designing Multi-Agent Systems with Zero Supervision
- LAMeTA: Intent-Aware Agentic Network Optimization via a Large AI Model-Empowered Two-Stage Approach
- Do Code LLMs Do Static Analysis?
- HALO: Hierarchical Autonomous Logic-Oriented Orchestration for Multi-Agent LLM Systems
- BELLE: A Bi-Level Multi-Agent Reasoning Framework for Multi-Hop Question Answering
- Time Travel is Cheating: Going Live with DeepFund for Real-Time Fund Investment Benchmarking
- Systematic Failures in Collective Reasoning under Distributed Information in Multi-Agent LLMs
- MASS: Muli-agent simulation scaling for portfolio construction
- Streaming Communication in Multi-Agent Reasoning
- Internet of Agents: Fundamentals, Applications, and Challenges
- Multi-agent Embodied AI: Advances and Future Directions
- Exploring Silicon-Based Societies: An Early Study of the Moltbook Agent Community
- TheBotCompany: Self-Organizing Multi-agent Systems for Continuous Software Development
- Brain-Inspired Graph Multi-Agent Systems for LLM Reasoning
- Efficient and Adaptable Overlapping for Computation and Communication via Signaling and Reordering
- Co-RedTeam: Orchestrated Security Discovery and Exploitation with LLM Agents
- LLM-Powered GUI Agents in Phone Automation: Surveying Progress and Prospects
- HELENA:Hierarchical Sparse Coordination over a Union of Complementary Topologies for MAS
- Debate Only When Necessary: Adaptive Multiagent Collaboration for Efficient LLM Reasoning
- OptimAI: Optimization from Natural Language Using LLM-Powered AI Agents
- FlowReasoner: Reinforcing Query-Level Meta-Agents
- Meta-Thinking in LLMs via Multi-Agent Reinforcement Learning: A Survey
- Weak-for-Strong: Training Weak Meta-Agent to Harness Strong Executors
Related