Jun 2026· ACM International Conference on Bioinformatics, Computational Biology and Biomedicine· 0 citations· 17 references
Computer Science
TL;DR
Findings highlight significant reliability and safety concerns when applying LLMs to medical classification tasks in their current basic form, and underline the need for improved models and methodologies before such systems can be responsibly deployed in medical informatics.
Abstract
This paper investigates the effectiveness of Large Language Models (LLMs) in handling medical classifications, focusing on two tasks: converting medical descriptions to codes (text-to-code) and generating descriptions from codes (code-to-text). We deliberately study these tasks in a probing setting, without employing fine-tuning or Retrieval-Augmented Generation (RAG) techniques, as access to external knowledge sources would make the tasks largely trivial and would not reflect the models’ intrinsic knowledge acquired during training. Our goal is therefore to evaluate the structured medical knowledge internalized by LLMs during pre-training. Experiments across several widely used medical classifications reveal substantial limitations: although larger models generally perform better, LLMs correctly identify only a limited fraction of codes and descriptions, while frequently producing hallucinated or incorrect outputs. Performance also varies substantially across classification systems, particularly for more complex ontologies. These findings highlight significant reliability and safety concerns when applying LLMs to medical classification tasks in their current basic form, and underline the need for improved models and methodologies before such systems can be responsibly deployed in medical informatics.
It is suggested that domain-adapted encoder models may be preferable for similar structured clinical NER settings, although larger and externally validated benchmarks are needed before generalizing to other languages, clinical corpora, model families, or deployment environments.
L. Elvas, Carolina Carvalho· Scientific Reports· 0 citations
This entry-level tutorial aims to equip healthcare professionals with the tools necessary to effectively integrate LLMs into clinical practice, ensuring that these powerful technologies are applied in a safe, reliable, and impactful manner.
Qiao Jin, Nicholas Wan, Robert Leaman et al.· Nature Protocols· 1 citation
Large Language Models (LLMs) have emerged as transformative tools across various domains, exhibiting remarkable capabilities in natural language processing and generation. However, their reliance on static pre-training data limits their ability to access up-to-date and domain-specific information. The existing research often treats augmentation strategies in isolation, and limited efforts have been made to systematically compare them through the lens of information integrity. This review focuses specifically on Retrieval-Augmented Generation (RAG) and fine-tuning, identifying them as the two dominant paradigms for integrating external knowledge: RAG for retrieval-based context injection and fine-tuning for parametric knowledge adaptation. While existing surveys predominantly focus on performance metrics like accuracy or latency, this paper addresses the critical gap of data fidelity—the preservation of truthfulness, integrity, and fairness during augmentation. We systematically synthesize empirical findings from diverse methodologies to determine how each approach mitigates hallucinations and bias. By comparing the trade-offs between retrieval-based context injection and parametric knowledge adaptation, this survey provides unique value to readers by providing a structured taxonomy, a unified evaluation framework, and actionable insights to guide future research and practical deployment of robust, high-fidelity LLMs.
Soham Mukherjee, John Le, Chau Nguyen· Knowledge· 0 citations
Findings show that clinical LLM explainability has shifted toward fluent generative rationales, but evidence that such explanations reflect model reasoning remains limited, and three regulatory priorities are highlighted: prioritizing explanations that enable independent verification or logic auditing over plausibility-only rationales; preferring inspectable models where regulatory documentation is required; and prospectively validating explanations in clinical workflows before scaling.
Factual reliability remains a critical bottleneck for Large Language Models (LLMs) in specialized medical domains like Systemic Lupus Erythematosus (SLE). While Retrieval-Augmented Generation (RAG) mitigates hallucinations, the qualitative impact of different underlying knowledge sources remains under-explored. This paper addresses this gap by evaluating three distinct repositories—Medical Textbooks, Academic Journals, and Clinical Guidelines—within a multi-path RAG framework using an open-weights, small-scale LLM (Gemma-3:4b). Based on a double-blind assessment of 45 stratified queries by three medical residents, we map the unique competency profiles of each source type. The core contributions of this study are twofold: (1) we provide an empirical characterization of medical source trade-offs, demonstrating that textbooks excel in foundational clarity and complex reasoning, guidelines ensure procedural precision, and journals provide technical depth at the cost of synthesis friction; and (2) we offer initial empirical insights that can help the development of future intent-aware RAG routing mechanisms to enhance clinical safety.
Since the widespread adoption of large language models (LLMs), AI-generated text has become increasingly prevalent in scientific communication and online content. As the proportion of LLM generated text grows, concerns have emerged regarding recursive self-training, where models are trained on data generated by earlier model versions. Prior work suggests that such training regimes can lead to model collapse, characterized by the loss of semantic diversity and degradation of learned data distributions. Although these effects have been studied primarily in general-domain English settings, their implications for domain-specific and high-stakes biomedical applications remain insufficiently understood. We investigate the impact of recursive self-training on German-language medical text by recursively fine-tuning a medically specialized language model over multiple iterations. For comparison purposes we present two alternative training pipelines. In the first pipeline, the model is recursively trained exclusively on synthetic German medical text generated by earlier model versions. In the second pipeline, synthetic German medical text is combined with human German medical text at each iteration to test whether data blending mitigates recursive degradation. The evaluation is based on answer accuracy achieved on a set of multiple-choice questions (MCQs) drawn from the German medical licensing examination, the Staatsexamen, as well as lexical diversity metrics such as distinct-n, and next-token probability concentration measures that capture distributional collapse. Across multiple training iterations, the purely synthetic pipeline exhibits progressive degradation in both question–answering performance and linguistic diversity, consistent with the effects of model collapse. In contrast, although minor degradation is still observable, the hybrid pipeline maintains substantially more stable performance over time. These results suggest that human-anchored recursive training (blending synthetic data with real, human-generated data) constitutes a promising mitigation strategy against recursive degradation.