Biomedical reasoning in action: Multi-agent System for Auditable Biomedical Evidence Synthesis

TL;DR

The paper addresses the challenge of producing robust, auditable biomedical evidence synthesis using an auditable multi-agent LLM workflow. It proposes M-Reason, a modular platform where independent evidence streams are analyzed by dedicated agents and then integrated by a synthesis layer with consensus-based validation. Key contributions include a detailed architectural blueprint, specialized prompts to curb hallucinations, and an interactive interface that makes reasoning transparent and auditable. Evaluation demonstrates substantial time savings (e.g., ~135x faster than manual reading in large gene sets) and high output consistency, supporting M-Reason as a practical testbed for robust multi-agent LLM systems in biomedicine.

Abstract

We present M-Reason, a demonstration system for transparent, agent-based reasoning and evidence integration in the biomedical domain, with a focus on cancer research. M-Reason leverages recent advances in large language models (LLMs) and modular agent orchestration to automate evidence retrieval, appraisal, and synthesis across diverse biomedical data sources. Each agent specializes in a specific evidence stream, enabling parallel processing and fine-grained analysis. The system emphasizes explainability, structured reporting, and user auditability, providing complete traceability from source evidence to final conclusions. We discuss critical tradeoffs between agent specialization, system complexity, and resource usage, as well as the integration of deterministic code for validation. An open, interactive user interface allows researchers to directly observe, explore and evaluate the multi-agent workflow. Our evaluation demonstrates substantial gains in efficiency and output consistency, highlighting M-Reason's potential as both a practical tool for evidence synthesis and a testbed for robust multi-agent LLM systems in scientific research, available at https://m-reason.digitalecmt.com.

Figures (5)

• Figure 1: M-Reason system overview: visualization of agent orchestration, parallel processing of independent evidence modules, and extensible design enabling straightforward integration of new analytical agents.
• Figure 2: The M-Reason interface with seven message terminals, enabling users to observe the ongoing interpretation of evidence, structured report generation, and parallel feedback from validation agents.
• Figure 3: Prompt structure: Each agent receives a prompt that is context- and question-aware for every iteration. Prompts dynamically adapt to revision cycles by incorporating specific instructions to address reviewer feedback. Every prompt includes a detailed system message defining the agent’s role and responsibilities.
• Figure 4: Exemplary structured report generated by M-Reason, summarizing integrated biomedical evidence and highlighting novel findings, implications, and source citations.
• Figure 5: M-Reason report generation time compared to estimated human reading time for the same evidence sets. Y axis in log scale.