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

2026-03-23 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, quick

Summary

GSEM (Graph-based Self-Evolving Memory) is a novel clinical memory framework designed to enhance clinical decision-making agents by organizing prior experiences into a dual-layer memory graph. Unlike traditional memory-augmented methods that store experiences as independent records, GSEM explicitly captures decision structures within individual experiences and relational dependencies across multiple experiences. This approach aims to mitigate issues like noisy retrieval and unreliable reuse, which can sometimes degrade performance compared to direct Large Language Model (LLM) inference. GSEM incorporates applicability-aware retrieval and online feedback-driven calibration for node quality and edge weights. Evaluated across MedR-Bench and MedAgentsBench, GSEM achieved the highest average accuracy among baselines, reaching 70.90% with DeepSeek-V3.2 and 69.24% with Qwen3.5-35B.

Key takeaway

For AI Scientists developing clinical decision-making agents, GSEM demonstrates that structuring prior experiences into a dual-layer memory graph significantly boosts accuracy. You should consider adopting graph-based memory frameworks with online calibration to overcome limitations of independent record storage, potentially improving diagnostic support and treatment recommendations in real-world applications.

Key insights

GSEM organizes clinical experiences into a dual-layer memory graph for improved decision-making and retrieval.

Principles

• Explicit relational structure enhances memory retrieval.
• Online feedback improves memory graph quality.

Method

GSEM constructs a dual-layer memory graph for clinical experiences, enabling applicability-aware retrieval and calibrating node quality and edge weights via online feedback.

In practice

• Implement dual-layer graphs for complex data.
• Integrate feedback loops for memory refinement.

Topics

• Clinical Reasoning
• Graph-based Memory
• Large Language Models
• Memory Augmentation
• Clinical Decision Support

Code references

• xhan1022/gsem

Best for: AI Scientist, Research Scientist, AI Researcher, AI Engineer, AI Data Scientist

Related on AIssential

• Experience Makes Skillful: Enabling Generalizable Medical Agent Reasoning via Self-Evolving Skill Memory — SkeMex improves medical agents by distilling interaction trajectories into self-evolving, skill-based memory for generalizable reasoning.
• Psy-Chronicle:A Structured Pipeline for Synthesizing Long-Horizon Campus Psychological Counseling Dialogues — Long-horizon psychological counseling data can be synthesized by modeling student profiles, temporal stress events, and cross-session memory.
• Memory is Reconstructed, Not Retrieved: Graph Memory for LLM Agents — Memory retrieval for LLM agents is more effective as an active, reconstructive process over a graph.
• APEX-MEM: Agentic Semi-Structured Memory with Temporal Reasoning for Long-Term Conversational AI — APEX-MEM enhances long-term conversational AI through structured memory, temporal grounding, and intelligent retrieval.
• ActMem: Bridging the Gap Between Memory Retrieval and Reasoning in LLM Agents — ActMem integrates memory retrieval with active causal reasoning via a structured knowledge graph to enhance LLM agent decision-making.
• Reinforcement Learning Improves LLM Accuracy and Reasoning in Disease Classification from Radiology Reports — Reinforcement learning with GRPO enhances LLM accuracy and reasoning in medical text classification without explicit reasoning supervision.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.