While multi-agent debate (MAD) frameworks have shown significant potential in general reasoning, their effectiveness in highly structured, knowledge-heavy legal domains remains under-explored. This work introduces the Legal Multi-Agent Debate (L-MAD) framework to systematically evaluate different debate structures and aggregation methods within Legal Textual Entailment. By assigning distinct expert personas to multiple agents, L-MAD improves upon strong single-agent baselines by up to 8%.
Moreover, analyzing how debate scales reveals a clear trade-off: increasing the agent population reduces consistency and improves accuracy, whereas extending discussion rounds induces a detrimental \textit{over-deliberation drift} where agents reinforce each other's mistakes. Ultimately, our findings outline the practical boundaries and safety margins of deploying collaborative multi-agent systems in high-stakes legal reasoning environments.
Blogger's Review: The L-MAD framework showcases the potential of multi-agent systems in legal reasoning, particularly in tackling complex, knowledge-intensive tasks. However, the phenomenon of over-deliberation highlights the need for careful design of agent interaction mechanisms to avoid reinforcing errors, ensuring the accuracy and reliability of legal reasoning.