OMAC: A Broad Optimization Framework for LLM-Based Multi-Agent Collaboration
This work addresses the problem of inefficient handcrafted methods in multi-agent collaboration for researchers and practitioners, offering a systematic optimization framework that is incremental in improving existing capabilities.
The paper tackles the lack of systematic design and optimization in LLM-based multi-agent systems by introducing OMAC, a general framework for holistic optimization across five key dimensions, resulting in superior performance on code generation, arithmetic reasoning, and general reasoning tasks compared to state-of-the-art approaches.
Agents powered by advanced large language models (LLMs) have demonstrated impressive capabilities across diverse complex applications. Recently, Multi-Agent Systems (MAS), wherein multiple agents collaborate and communicate with each other, have exhibited enhanced capabilities in complex tasks, such as high-quality code generation and arithmetic reasoning. However, the development of such systems often relies on handcrafted methods, and the literature on systematic design and optimization of LLM-based MAS remains limited. In this work, we introduce OMAC, a general framework designed for holistic optimization of LLM-based MAS. Specifically, we identify five key optimization dimensions for MAS, encompassing both agent functionality and collaboration structure. Building upon these dimensions, we first propose a general algorithm, utilizing two actors termed the Semantic Initializer and the Contrastive Comparator, to optimize any single dimension. Then, we present an algorithm for joint optimization across multiple dimensions. Extensive experiments demonstrate the superior performance of OMAC on code generation, arithmetic reasoning, and general reasoning tasks against state-of-the-art approaches.