Memorization in Large Language Models in Medicine: Prevalence, Characteristics, and Implications

2025-09-10

AI Summary

This paper systematically investigates training data memorization—i.e., unintended verbatim reproduction of original medical texts—in large language models (LLMs) deployed in healthcare. We examine memorization across four dimensions: prevalence, content characteristics, scale, and clinical implications, comparing three training paradigms: continued pretraining, standard fine-tuning, and fine-tuning on real inpatient records containing sensitive health data. We introduce the first tripartite taxonomy—“beneficial,” “uninformative,” and “harmful”—to categorize memorized content, empirically demonstrating significantly higher memorization rates in medical LLMs than in general-domain counterparts. Integrating sensitive-content detection with multi-task evaluation, we quantify divergent impacts of memory types on model utility and patient privacy. Finally, we propose targeted mitigation strategies that jointly preserve clinical performance and safeguard protected health information, offering both theoretical foundations and actionable guidelines for secure, trustworthy deployment of medical AI.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Large Language Models (LLMs) have demonstrated significant potential in medicine. To date, LLMs have been widely applied to tasks such as diagnostic assistance, medical question answering, and clinical information synthesis. However, a key open question remains: to what extent do LLMs memorize medical training data. In this study, we present the first comprehensive evaluation of memorization of LLMs in medicine, assessing its prevalence (how frequently it occurs), characteristics (what is memorized), volume (how much content is memorized), and potential downstream impacts (how memorization may affect medical applications). We systematically analyze common adaptation scenarios: (1) continued pretraining on medical corpora, (2) fine-tuning on standard medical benchmarks, and (3) fine-tuning on real-world clinical data, including over 13,000 unique inpatient records from Yale New Haven Health System. The results demonstrate that memorization is prevalent across all adaptation scenarios and significantly higher than reported in the general domain. Memorization affects both the development and adoption of LLMs in medicine and can be categorized into three types: beneficial (e.g., accurate recall of clinical guidelines and biomedical references), uninformative (e.g., repeated disclaimers or templated medical document language), and harmful (e.g., regeneration of dataset-specific or sensitive clinical content). Based on these findings, we offer practical recommendations to facilitate beneficial memorization that enhances domain-specific reasoning and factual accuracy, minimize uninformative memorization to promote deeper learning beyond surface-level patterns, and mitigate harmful memorization to prevent the leakage of sensitive or identifiable patient information.

Problem

• Assessing memorization prevalence in medical LLMs across adaptation scenarios
• Evaluating characteristics and volume of memorized medical training content
• Analyzing downstream impacts of memorization on medical applications

Innovation

• Comprehensive evaluation of medical data memorization in LLMs
• Systematic analysis across three adaptation scenarios
• Classification into beneficial, uninformative, and harmful memorization types