MedLA: A Logic-Driven Multi-Agent Framework for Complex Medical Reasoning with Large Language Models

2025-09-28

AI Summary

Existing medical multi-agent systems suffer from rigid role assignment and superficial interaction, limiting their ability to detect and resolve fine-grained logical inconsistencies. To address this, we propose a logic-driven multi-agent framework that structures each agent’s reasoning as a syllogistic logical tree and introduces a graph-guided multi-round negotiation mechanism, enabling premise-level alignment and systematic contradiction resolution for interpretable, traceable consensus. Our approach integrates formal logic modeling, dynamic role allocation, and large language model (LLM) foundations to balance deductive rigor with collaborative flexibility. Evaluated on medical QA benchmarks—including MedDDx—our framework significantly outperforms static-role multi-agent systems and single-agent baselines. It is compatible with mainstream open-source and commercial LLMs and achieves state-of-the-art performance.

Technology Category

• Multiagent Systems: Agreement, Argumentation & Negotiation
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Knowledge Representation and Reasoning: Description Logics

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Answering complex medical questions requires not only domain expertise and patient-specific information, but also structured and multi-perspective reasoning. Existing multi-agent approaches often rely on fixed roles or shallow interaction prompts, limiting their ability to detect and resolve fine-grained logical inconsistencies. To address this, we propose extsc{MedLA}, a logic-driven multi-agent framework built on large language models. Each agent organizes its reasoning process into an explicit logical tree based on syllogistic triads (major premise, minor premise, and conclusion), enabling transparent inference and premise-level alignment. Agents engage in a multi-round, graph-guided discussion to compare and iteratively refine their logic trees, achieving consensus through error correction and contradiction resolution. We demonstrate that extsc{MedLA} consistently outperforms both static role-based systems and single-agent baselines on challenging benchmarks such as MedDDx and standard medical QA tasks. Furthermore, extsc{MedLA} scales effectively across both open-source and commercial LLM backbones, achieving state-of-the-art performance and offering a generalizable paradigm for trustworthy medical reasoning.

Problem

• Addressing complex medical reasoning requiring multi-perspective analysis
• Resolving fine-grained logical inconsistencies in medical decision-making
• Overcoming limitations of fixed-role agents in medical question answering

Innovation

• Logic-driven multi-agent framework for medical reasoning
• Explicit logical trees using syllogistic triads
• Multi-round graph-guided discussion for consensus