GSEM: Graph-based Self-Evolving Memory for Experience Augmented Clinical Reasoning

TL;DR

GSEM introduces a graph-based memory system for clinical reasoning that organizes experiences relationally, improving retrieval accuracy and decision quality in clinical decision-making agents.

Contribution

The paper presents GSEM, a novel graph-structured memory framework that captures relational dependencies among clinical experiences for enhanced reasoning.

Findings

Achieves highest accuracy on MedR-Bench and MedAgentsBench datasets.

Outperforms baseline methods with 70.90% and 69.24% accuracy.

Supports applicability-aware retrieval and online calibration.

Abstract

Clinical decision-making agents can benefit from reusing prior decision experience. However, many memory-augmented methods store experiences as independent records without explicit relational structure, which may introduce noisy retrieval, unreliable reuse, and in some cases even hurt performance compared to direct LLM inference. We propose GSEM (Graph-based Self-Evolving Memory), a clinical memory framework that organizes clinical experiences into a dual-layer memory graph, capturing both the decision structure within each experience and the relational dependencies across experiences, and supporting applicability-aware retrieval and online feedback-driven calibration of node quality and edge weights. Across MedR-Bench and MedAgentsBench with two LLM backbones, GSEM achieves the highest average accuracy among all baselines, reaching 70.90\% and 69.24\% with DeepSeek-V3.2 and Qwen3.5-35B, respectively. Code is available at https://github.com/xhan1022/gsem.