arXiv：cs.CL· Manjushree B. Aithal, Ph. D., Alexander Kotz, James Mitchell, Ph. D·· 4 小时前
PLACID：用端侧 2B-10B 小模型实现临床缩写消歧的隐私保护级联方案
PLACID: Privacy-preserving Large language models for Acronym Clinical Inference and Disambiguation
AI 导读
PLACID 提出一种隐私保护级联流程，用完全端侧部署的 2B-10B 小模型完成临床缩写检测与消歧：通用指令模型检测准确率约 0.988，但扩展准确率仅约 0.655；接入领域生物医学模型后，扩展准确率提升至约 0.81，避免向外部服务器传输受保护健康信息。
正文
Abstract:Large Language Models (LLMs) offer transformative solutions across many domains, but healthcare integration is hindered by strict data privacy constraints. Clinical narratives are dense with ambiguous acronyms, misinterpretation these abbreviations can precipitate severe outcomes like life-threatening medication errors. While cloud-dependent LLMs excel at Acronym Disambiguation, transmitting Protected Health Information to external servers violates privacy frameworks. To bridge this gap, this study pioneers the evaluation of small-parameter models deployed entirely on-device to ensure privacy preservation. We introduce a privacy-preserving cascaded pipeline leveraging general-purpose local models to detect clinical acronyms, routing them to domain-specific biomedical models for context-relevant expansions. Results reveal that while general instruction-following models achieve high detection accuracy (~0.988), their expansion capabilities plummet (~0.655). Our cascaded approach utilizes domain-specific medical models to increase expansion accuracy to (~0.81). This novel work demonstrates that privacy-preserving, on-device (2B-10B) models deliver high-fidelity clinical acronym disambiguation support.
| Comments: | 10 pages, 2 figures, Under review AMIA Symposium |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2603.23678 [cs.CL] |
| (or arXiv:2603.23678v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2603.23678 arXiv-issued DOI via DataCite |
Submission history
From: Manjushree Aithal [view email]
[v1]
Tue, 24 Mar 2026 19:33:44 UTC (687 KB)
[v2]
Tue, 13 Oct 2026 21:12:12 UTC (687 KB)
来源：arXiv：cs.CL · arxiv.org