MedReseacher-R1: Expert-Level Medical Deep Researcher via A Knowledge-Informed Trajectory Synthesis Framework

2025-08-20

AI Summary

Current LLM-driven general-purpose deep research agents face dual bottlenecks in healthcare: sparse medical knowledge impairs clinical reasoning, and the absence of domain-specific retrieval tools hinders performance on complex medical benchmarks. This paper introduces a knowledge-guided trajectory synthesis framework to build an expert-level medical deep researcher. First, it synthesizes long-chain, multi-hop medical QA data grounded in a curated medical knowledge graph. Second, it develops a dedicated private medical retrieval engine enabling domain-customized tool orchestration and multi-source information fusion. Third, it employs a two-stage training paradigm: supervised fine-tuning followed by online reinforcement learning with a composite reward function. The framework achieves state-of-the-art performance across multiple medical benchmarks—outperforming large proprietary systems—while preserving general research capabilities. It generates over 2,100 high-quality reasoning trajectories spanning 12 medical specialties.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Search and Optimization: Learning to Search
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Recent developments in Large Language Model (LLM)-based agents have shown impressive capabilities spanning multiple domains, exemplified by deep research systems that demonstrate superior performance on complex information-seeking and synthesis tasks. While general-purpose deep research agents have shown impressive capabilities, they struggle significantly with medical domain challenges, as evidenced by leading proprietary systems achieving limited accuracy on complex medical benchmarks. The key limitations are: (1) the model lacks sufficient dense medical knowledge for clinical reasoning, and (2) the framework is constrained by the absence of specialized retrieval tools tailored for medical contexts.We present a medical deep research agent that addresses these challenges through two core innovations. First, we develop a novel data synthesis framework using medical knowledge graphs, extracting the longest chains from subgraphs around rare medical entities to generate complex multi-hop question-answer pairs. Second, we integrate a custom-built private medical retrieval engine alongside general-purpose tools, enabling accurate medical information synthesis. Our approach generates 2100+ diverse trajectories across 12 medical specialties, each averaging 4.2 tool interactions.Through a two-stage training paradigm combining supervised fine-tuning and online reinforcement learning with composite rewards, our MedResearcher-R1-32B model demonstrates exceptional performance, establishing new state-of-the-art results on medical benchmarks while maintaining competitive performance on general deep research tasks. Our work demonstrates that strategic domain-specific innovations in architecture, tool design, and training data construction can enable smaller open-source models to outperform much larger proprietary systems in specialized domains.

Problem

• Addresses medical domain limitations in deep research agents
• Overcomes insufficient medical knowledge for clinical reasoning
• Integrates specialized retrieval tools for medical information synthesis

Innovation

• Medical knowledge graph data synthesis framework
• Custom private medical retrieval engine integration
• Two-stage training with fine-tuning and reinforcement learning