ResearchNovember 05, 2026
-41
🧊
MedGraph Beats Five Memory Systems
A graph-based memory system called MedGraph maintains accurate patient state across long-term medical dialogues.
#graph-memory#medical-AI#LLM#dialogue#benchmark

🔥 What happened
Donghua Zhang's team unveiled MedGraph, a graph memory for medical LLM assistants that tracks patient state across visits. It tops MedMemoryBench by 10.5 points over five baselines with a GPT backbone, and 10.4 with Qwen.
💡 Why it matters
Existing memory systems fail three ways: stale state, another person's facts, and missing safety data. MedGraph chains dated facts per patient and reads each chain at the question's timestamp — a real state tracker, not vector search.
⚡ Our take
Longitudinal care needs graphs, not embeddings. A 10-point gap isn't tuning — it's an architecture shift.
The title, summary and analysis of this item were produced automatically by an AI system and have not been editorially reviewed. They may contain errors, bias or omissions — when in doubt, read the linked original source.
Deep Dives & Similar Intelligence

Claude audits neural nets by weight
ResearchNovember 03, 2026

AI Learns to Lie in Mafia Game
ResearchNovember 05, 2026

One in three benchmark tasks is broken
ResearchNovember 05, 2026

Stanford: LLM Research Expires in 16 Months
ResearchNovember 03, 2026

GPT-6 Reads Neural Nets Like Plain Text
ResearchNovember 02, 2026