Search papers, labs, and topics across Lattice.
This paper conducts a systematic literature review of 141 studies on Multi-Agent Debate (MAD), identifying a narrow design pattern that dominates current research, characterized by static topologies and short-term memory strategies. The authors propose a comprehensive three-dimensional taxonomy that categorizes MAD based on participants, interaction mechanisms, and agreement protocols, addressing the fragmented nature of existing research. Their findings highlight the need for a more diverse exploration of MAD configurations to improve the robustness and accuracy of LLM-based systems, paving the way for future advancements in this area.
Current Multi-Agent Debate research is stuck in a narrow design rut, risking the potential of more innovative configurations that could enhance LLM performance.
Multi-Agent Debate (MAD) is a promising paradigm for improving the accuracy and robustness of Large Language Model (LLM)-based agentic systems. It enables multiple agents to exchange arguments, critique each other's outputs, and iteratively converge towards a solution. However, research remains fragmented, with inconsistent terminology and no rigorous synthesis of MAD design dimensions. We present a systematic literature review characterizing 141 primary studies on MAD. We derive a three-dimensional taxonomy covering debate participants, the interaction mechanisms structuring the exchange, and the agreement protocols governing debate resolution, supported by formal notations to render MAD configurations. Our analysis reveals that the field has implicitly converged on a narrow design pattern - static, fully connected topologies, verbatim exchange, short-term memory and voting resolution strategies - adopted by convention rather than systematic comparison, while promising alternatives remain marginal. Because any MAD setting reflects roughly a dozen interacting design decisions, cross-study comparison is unreliable when these are left implicit. We position the taxonomy as a descriptive map of the research landscape, a framework for controlled benchmarking, and potentially as a schema for machine-readable MAD specifications. As future work, we propose formalizing it into an executable specification, enabling cost-aware benchmarking and automated tuning of debate configurations.
