ArticleCommunications medicine2025
Multi-model assurance analysis showing large language models are highly vulnerable to adversarial hallucination attacks during clinical decision support.
Article in Communications medicine, 2025. The graph could read no effect estimate from its abstract, so it casts no vote on the map. Cited by 84 papers, 3 of them syntheses that pooled it.
What it found
Each row is one number read from the abstract, on the scale the paper reported it, with its interval. Left of the dashed line favours the treatment, right favours the comparator. Under each row is the sentence it came from. New to these charts? A ten-minute tutorial.
The abstract states no effect estimate the extractor could read, or names no intervention and outcome on the map, so this paper lights no cell and moves no belief. It is still indexed, cited and linked below.
The trial behind it
Trials whose registry record cites this paper, or whose number appears in the abstract. A trial that started after this paper was published is citing it as background, not reporting it.
Neither the registry nor the abstract names a trial number. If this is a trial report, that itself is worth knowing.
Who cites it
84 citing papers in PubMed, 3 syntheses or guidelines pooled it.
- Fine-Tuning, Retrieval-Augmented Generation, and Hybrid Adaptation of Language Models for Clinical Decision-Making in Health Care: Systematic Review.Journal of medical Internet research · 2026Pooled it
- Generative large language models in the clinical management of Alzheimer's disease and mild cognitive impairment.Neurological sciences : official journal of the Italian Neurological Society and of the Italian Society of Clinical Neurophysiology · 2026Pooled it
- Generative AI in Precision Nutrition: A Review of Current Developments and Future Directions.Nutrients · 2026Pooled it
- The missing links in agentic AI autonomy.Nature medicine · 2026Article
- Large Language Models and Retrieval-Augmented Platforms for the Diagnosis and Management of Periodontal Diseases: A Blinded Expert-Rated Comparative Study of 11 Systems.Journal of clinical periodontology · 2026Article
- Performance of large language models in ambiguous treatment scenarios of early breast cancer.Breast (Edinburgh, Scotland) · 2026Article
- Article
- The Patient-AI Relationship in Obsessive-Compulsive and Related Disorders: A Cognitive-Behavioral Framework.Journal of clinical medicine · 2026Article
- AI Agents for Multimodal Oncology Diagnosis: Toward Transparent and Traceable Clinical Decision Support.JMIR cancer · 2026Review
- Evaluating Large Language Models in Clinical Audiology (AUDIOLOGYBENCH): Benchmark Development and Validation Study.Journal of medical Internet research · 2026Article
- Multimodal Large Language Models for Dental Chart Image Interpretation: Cross-Sectional Benchmarking Study With Students and Clinicians.JMIR medical education · 2026Article
- Evaluating the Performance and Fragility of Large Language Models on the Self-Assessment for Neurological Surgeons.Neurosurgery · 2026Article
- The Performance of Large Language Models in Extracting Intestinal Symptoms From Electronic Health Records: Retrospective Observational Study.Journal of medical Internet research · 2026Observational
- Large Language Models for Patient Education in Cardiovascular Imaging: Prospective Observational Comparative Study.JMIR formative research · 2026Observational
- Clinical AI Generators and Reviewers must be Tested Together.Journal of medical systems · 2026Article
- A Self-Controlled Benchmark of Retrieval-Augmented Generation for Large Language Models on Clinical Guideline Questions.Diagnostics (Basel, Switzerland) · 2026Article
- Review
- Mitigating hallucinations in synthesized clinical texts to improve multimodal deep learning for dermatology.Journal of biomedical informatics · 2026Article
- From Static Outputs to Living Evidence: AI for Integrated Knowledge Translation in Canadian Health Research.JMIR AI · 2026Article
- A Multiagent Large Language Model Framework for Emergency Treatment Recommendation in Acute Ischemic Stroke: Development and Validation Study.Journal of medical Internet research · 2026Article
24 more citing papers are in PubMed but not listed here.
Corrections and comments
PubMed lists nothing against this paper. Absence here is not a guarantee, only a check that was made.
Authors and funding
11 authors.
Funding
No grant is acknowledged in the PubMed record.
Abstract
backgroundLarge language models (LLMs) show promise in clinical contexts but can generate false facts (often referred to as "hallucinations"). One subset of these errors arises from adversarial attacks, in which fabricated details embedded in prompts lead the model to produce or elaborate on the false information. We embedded fabricated content in clinical prompts to elicit adversarial hallucination attacks in multiple large language models. We quantified how often they elaborated on false details and tested whether a specialized mitigation prompt or altered temperature settings reduced errors.
methodsWe created 300 physician-validated simulated vignettes, each containing one fabricated detail (a laboratory test, a physical or radiological sign, or a medical condition). Each vignette was presented in short and long versions-differing only in word count but identical in medical content. We tested six LLMs under three conditions: default (standard settings), mitigating prompt (designed to reduce hallucinations), and temperature 0 (deterministic output with maximum response certainty), generating 5,400 outputs. If a model elaborated on the fabricated detail, the case was classified as a "hallucination".
resultsHallucination rates range from 50 % to 82 % across models and prompting methods. Prompt-based mitigation lowers the overall hallucination rate (mean across all models) from 66 % to 44 % (p < 0.001). For the best-performing model, GPT-4o, rates decline from 53 % to 23 % (p < 0.001). Temperature adjustments offer no significant improvement. Short vignettes show slightly higher odds of hallucination.
conclusionsLLMs are highly susceptible to adversarial hallucination attacks, frequently generating false clinical details that pose risks when used without safeguards. While prompt engineering reduces errors, it does not eliminate them.
Identifiers
What OpenQuestion holds
Registered trials
Read under generation 80e0d062 · epoch 390. Bibliography from PubMed, PubMed Central and OpenAlex; grants from NIH RePORTER; trial links from ClinicalTrials.gov; estimates, votes and beliefs from the OpenQuestion graph.