Multi-Agent LLM System for Complex Task Solving

This paper has presented the design of a Multi-Agent LLM System that addresses well-documented shortcomings of single-model approaches to complex, multi-step tasks. The proposed architecture distributes responsibility across a Planner Agent, Executor Agents, an Evaluator Agent, and a Coordinator, each contributing a specialised function to the task completion pipeline. A review of recent literature confirms that while multi- agent LLM research is advancing rapidly, general-purpose systems with built-in validation and error correction remain relatively scarce. The present work directly targets this gap with a practical, modular design amenable to extension and domain-specific adaptation. When tested against standard NLP benchmarks, the sys- tem is expected to demonstrate measurable improvements in accuracy, reasoning quality, and task reliability over single- LLM baselines. The architecture also positions the work well for future exploration of dynamic agent allocation, persistent memory, and multi-modal input.

[1]     S. Han, Q. Zhang, Y. Yao, W. Jin, Z. Xu, and C. He, “LLM Multi-Agent Systems: Challenges and Open Problems,” arXiv:2402.03578, 2024. [2]     S. Gandhi, D. Goldie, and P. Abbeel, “BudgetMLAgent: A Cost- Effective LLM Multi-Agent System for Automating Machine Learning Tasks,” arXiv preprint, 2024. [3]     X. Li, Z. Liu, and L. Bing, “Survey on LLM-based Multi-Agent Systems: Workflow, Infrastructure, and Challenges,” arXiv preprint, 2024. [4]     T. Guo et al., “Large Language Model based Multi-Agents: A Survey of Progress and Challenges,” arXiv:2402.01680, 2024. [5]     W. Shen et al., “Small LLMs Are Weak Tool Learners: A Multi-LLM Agent,” arXiv preprint, 2024. [6]     Q. Wu et al., “AutoGen: Enabling Next-Generation LLM Applications via Multi-Agent Conversation,” in Proc. ICLR Workshop on LLM Agents, 2024. [7]     X. Liu et al., “AgentBench: Evaluating LLMs as Agents,” in Proc. ICLR, 2024. [8]     C. Qian et al., “ChatDev: Communicative Agents for Software Devel- opment,” in Proc. ACL, 2024. [9]     W. Chen et al., “AgentVerse: Facilitating Multi-Agent Collaboration and Exploring Emergent Behaviors,” in Proc. ICLR, 2024. [10]     G. Chen et al., “AutoAgents: A Framework for Automatic Agent Generation,” in Proc. IJCAI, 2024.