Privacy-Preserving Retrieval for Auditable Clinical Language Modeling on Real-World Radiology Data
Abstract
Abstract Clinical large language models adapted for real-world use are commonly fine-tuned on patient data, embedding confidential information within model parameters and limiting auditability and privacy. We evaluate a retrieval-based framework separating clinical data from the language model by storing patient records in externally governed memory. On a radiology report summarisation task, retrieval recovers 32–67% of fine-tuning gains (perplexity, ROUGE-L; p < 0.0001), while hallucination reduction remains a future focus.
Fields of Study & Subjects
Artificial Intelligence in Healthcare and Education
Privacy-Preserving Technologies in Data
Machine Learning in Healthcare
Publication details
- Journal / source
- Research Square
- Publication date
- 8 March 2026
- Language
- EN
- Document type
- Preprint